<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:dc="http://purl.org/dc/elements/1.1/">
  <channel>
    <title>DEV Community: Seller Tools</title>
    <description>The latest articles on DEV Community by Seller Tools (@yougkxm).</description>
    <link>https://dev.to/yougkxm</link>
    <image>
      <url>https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F4248501%2F7b4132bf-6173-493f-bc73-87dbc9cb9321.png</url>
      <title>DEV Community: Seller Tools</title>
      <link>https://dev.to/yougkxm</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://dev.to/feed/yougkxm"/>
    <language>en</language>
    <item>
      <title>ArtStation Marketplace Closes Dec 18: The 4 Gumroad Migration Steps Most Sellers Miss</title>
      <dc:creator>Seller Tools</dc:creator>
      <pubDate>Thu, 19 Nov 2026 12:07:32 +0000</pubDate>
      <link>https://dev.to/yougkxm/artstation-marketplace-closes-dec-18-the-4-gumroad-migration-steps-most-sellers-miss-1nea</link>
      <guid>https://dev.to/yougkxm/artstation-marketplace-closes-dec-18-the-4-gumroad-migration-steps-most-sellers-miss-1nea</guid>
      <description>&lt;p&gt;ArtStation's Marketplace and Prints shop stop taking orders on &lt;strong&gt;December 18, 2026, 3:59 PM PT&lt;/strong&gt; (December 19, 08:59 KST). ArtStation has said the date is final.&lt;/p&gt;

&lt;p&gt;The migration to Gumroad is not a single button. It is a queue, then a set of drafts you have to review and publish yourself — and four of those steps are the ones sellers skip. Here is what each one actually costs when it goes wrong.&lt;/p&gt;

&lt;p&gt;If you would rather tick them off one by one, there is a free interactive version of this list: &lt;strong&gt;&lt;a href="https://yougkxm.github.io/artstation-to-gumroad/" rel="noopener noreferrer"&gt;ArtStation → Gumroad migration checklist&lt;/a&gt;&lt;/strong&gt;. It saves your progress in the browser and has the full category table with a filter.&lt;/p&gt;

&lt;h2&gt;
  
  
  1. Connecting your Gumroad account is not migrating
&lt;/h2&gt;

&lt;p&gt;Connecting is half of it. The other half is the &lt;strong&gt;"Join the queue"&lt;/strong&gt; button on the seller banner or in account settings. If you connect and never queue, your listings do not move. ArtStation also allows one Gumroad store per ArtStation account, and that store can be connected to only one ArtStation account — so pick the store you actually want buyers to land on.&lt;/p&gt;

&lt;p&gt;Queuing does not take your listings down early. They keep selling on ArtStation until December 18.&lt;/p&gt;

&lt;h2&gt;
  
  
  2. Publishing is locked until a payout method is saved
&lt;/h2&gt;

&lt;p&gt;Payout details do not come over from ArtStation. On the new Gumroad account you add a payout method in Payments settings — and &lt;strong&gt;saving it is what unlocks publishing&lt;/strong&gt;. That is why it comes before you touch a single draft.&lt;/p&gt;

&lt;p&gt;Worth knowing before you price anything, because it changes when money actually arrives:&lt;/p&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Rule&lt;/th&gt;
&lt;th&gt;Default&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;Minimum balance&lt;/td&gt;
&lt;td&gt;$100 USD&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Hold&lt;/td&gt;
&lt;td&gt;7 days&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Sales counted through&lt;/td&gt;
&lt;td&gt;the previous Friday (UTC)&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;PayPal fee&lt;/td&gt;
&lt;td&gt;2%&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Transfer time&lt;/td&gt;
&lt;td&gt;1–3 business days&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;Selling from India? Bank payouts cannot be set up on a newly created account there. PayPal is the method, and saving it is also what unlocks publishing.&lt;/p&gt;

&lt;h2&gt;
  
  
  3. Category decides whether your ArtStation Store tab survives
&lt;/h2&gt;

&lt;p&gt;Migrated products arrive as &lt;strong&gt;unpublished drafts&lt;/strong&gt;. Nothing sells until you publish each one. A draft you never publish never goes live, and this is where "I joined the queue" sellers end up with a Store tab full of drafts on the 18th.&lt;/p&gt;

&lt;p&gt;The category is the visibility switch. A Gumroad product with no category, or one outside ArtStation's list, will not appear on your ArtStation Store tab. It still sells on Gumroad — you just lose the storefront that was sending buyers to it.&lt;/p&gt;

&lt;p&gt;Only products in these categories stay visible:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;3D&lt;/strong&gt; — 3D Assets (3ds Max, Blender, Cinema 4D, Houdini, Maya, Modo, SketchUp, Unity, Unreal, ZBrush), Accessories / Clothing / Footwear and Headwear, Props, Textures (including PBR and MatCap), Unity animations / particle systems / shaders, 3D Modeling / Character Design / Rigging / Animating&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Drawing &amp;amp; Painting&lt;/strong&gt; — Digital Illustration, Illustration Brushes (Photoshop, Procreate, Blender), Traditional Art&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Design&lt;/strong&gt; — Entertainment Design, Industrial Design, Textures &amp;amp; Patterns (2D)&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Photography&lt;/strong&gt; — Reference Photos, Stock Photos, Photo Courses&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Comics &amp;amp; Graphic Novels&lt;/strong&gt;&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;Two more things in this step:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Duplicates.&lt;/strong&gt; The migration copies your ArtStation products and does not merge them with listings you already had on Gumroad. If the product already exists in your store, delete the imported draft and keep the original.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Files and prices.&lt;/strong&gt; Download the draft's file and confirm it is the version you want to sell. ArtStation prices carry over, but check the currency and any launch discount. Description, licence and cover edits you make on ArtStation after migration do not reach Gumroad.&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  4. The cover is a different shape, and your reviews do not come with you
&lt;/h2&gt;

&lt;p&gt;ArtStation listing art is not shaped like a Gumroad card. A 16:9 cover exported at &lt;strong&gt;1280×720&lt;/strong&gt; sits correctly in the Gumroad storefront and in the ArtStation Store tab feed. Keep the subject inside the middle third — both surfaces crop the edges, and not identically.&lt;/p&gt;

&lt;p&gt;Ratings, reviews and sales history do &lt;strong&gt;not&lt;/strong&gt; transfer. A product with hundreds of reviews on ArtStation starts at zero social proof on Gumroad. Cheap mitigations, all free: keep a screenshot or PDF of your best reviews before the old pages retire, add a short "previously on ArtStation" line to the description, and point people at your Gumroad profile rather than at old product URLs — those links now redirect to your Store tab or profile.&lt;/p&gt;

&lt;p&gt;For the cover half, there is a free browser tool with Gumroad 1280×720 and 600×600 presets, safe-zone guides and PNG export — no install, nothing uploaded: &lt;strong&gt;&lt;a href="https://yougkxm.github.io/listing-kit/" rel="noopener noreferrer"&gt;Listing &amp;amp; Mockup Frame Kit&lt;/a&gt;&lt;/strong&gt;.&lt;/p&gt;

&lt;h2&gt;
  
  
  Work backwards from December 18
&lt;/h2&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Date&lt;/th&gt;
&lt;th&gt;What to do&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;Now – Nov 24&lt;/td&gt;
&lt;td&gt;Connect, queue, save a payout method, publish drafts, fix categories&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Nov 25 – Dec 4&lt;/td&gt;
&lt;td&gt;ArtStation's final Black Friday sale — your ArtStation listings can still sell&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Dec 18, 3:59 PM PT&lt;/td&gt;
&lt;td&gt;Selling stops on ArtStation; your Store tab shows Gumroad products&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;By Feb 28, 2027&lt;/td&gt;
&lt;td&gt;Last chance to migrate digital products (selling nowhere in between)&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;That last row is the trap. The digital migration window runs to February 28, 2027, but between December 19 and then your listings sell nowhere. The gap is the cost of being late — not the loss of the products.&lt;/p&gt;

&lt;h2&gt;
  
  
  The checklist, if you want it in order
&lt;/h2&gt;

&lt;p&gt;The interactive version is free and needs no account: &lt;strong&gt;&lt;a href="https://yougkxm.github.io/artstation-to-gumroad/" rel="noopener noreferrer"&gt;yougkxm.github.io/artstation-to-gumroad&lt;/a&gt;&lt;/strong&gt; — 22 tickable steps, a saved progress bar, and a category filter for the "does my product type stay visible?" question.&lt;/p&gt;

&lt;p&gt;Dates, rules and category names here come from ArtStation's official &lt;a href="https://help.artstation.com/en/articles/18442391-artstation-marketplace-migration-seller-faq" rel="noopener noreferrer"&gt;Marketplace Migration Seller FAQ&lt;/a&gt; (updated November 18, 2026) and Gumroad's own migration help article (revised November 16, 2026), both checked on November 19, 2026. Platforms change terms — confirm on their pages before you act. This is an independent guide, not affiliated with ArtStation or Gumroad, and it promises no income or ranking.&lt;/p&gt;

</description>
      <category>artstation</category>
      <category>gumroad</category>
      <category>gamedev</category>
      <category>3d</category>
    </item>
    <item>
      <title>The sign is where bank statement parsers break: -64.28 DR, (123.45), 100.00 CR</title>
      <dc:creator>Seller Tools</dc:creator>
      <pubDate>Thu, 19 Nov 2026 04:42:04 +0000</pubDate>
      <link>https://dev.to/yougkxm/the-sign-is-where-bank-statement-parsers-break-6428-dr-12345-10000-cr-1pl</link>
      <guid>https://dev.to/yougkxm/the-sign-is-where-bank-statement-parsers-break-6428-dr-12345-10000-cr-1pl</guid>
      <description>&lt;p&gt;Bank statements are the one PDF most people are told to 'just convert', and the one they should be most careful about handing to a website. A statement is a complete account history: account number, every payer and vendor, every balance. I wanted to see how far a good parser could get with the file never leaving the machine, so I built one as a single page and kept the parsing rules explicit.&lt;/p&gt;

&lt;p&gt;This is what actually decides whether the CSV is right or wrong.&lt;/p&gt;

&lt;h2&gt;
  
  
  Why parse in the browser
&lt;/h2&gt;

&lt;p&gt;The file goes: FileReader → ArrayBuffer → pdfjsLib.getDocument({data}) → page.getTextContent() → my parser → Blob → download. There is no fetch, no upload endpoint, no analytics event. The trade-off is honest: no OCR, no server-side PDF toolbox, and any really broken layout stays broken. In exchange, 'what happens to my statement?' has a one-word answer.&lt;/p&gt;

&lt;h2&gt;
  
  
  Five spellings of the same debit
&lt;/h2&gt;

&lt;p&gt;A statement writes money going out in at least five ways, and they are not interchangeable to a spreadsheet:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;code&gt;-64.28&lt;/code&gt; — a minus sign&lt;/li&gt;
&lt;li&gt;
&lt;code&gt;64.28 DR&lt;/code&gt; — the bank debit marker, no minus&lt;/li&gt;
&lt;li&gt;
&lt;code&gt;(123.45)&lt;/code&gt; — accounting parentheses&lt;/li&gt;
&lt;li&gt;
&lt;code&gt;−64.28&lt;/code&gt; — the Unicode minus (U+2212), which is not the hyphen&lt;/li&gt;
&lt;li&gt;
&lt;code&gt;100.00 CR&lt;/code&gt; — a credit, so positive&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;The classic bug is inverting a sign twice: treat &lt;code&gt;DR&lt;/code&gt; as negative and then honour a minus that is already there, or flip parentheses and then negate again. One $64 purchase becomes a $64 deposit, and reconciliation is off by $128 for that row. So the rule is: decide the sign once.&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight javascript"&gt;&lt;code&gt;&lt;span class="kd"&gt;function&lt;/span&gt; &lt;span class="nf"&gt;cleanAmount&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;value&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt;
  &lt;span class="k"&gt;if&lt;/span&gt; &lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;value&lt;/span&gt; &lt;span class="o"&gt;==&lt;/span&gt; &lt;span class="kc"&gt;null&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="k"&gt;return&lt;/span&gt; &lt;span class="kc"&gt;null&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
  &lt;span class="kd"&gt;let&lt;/span&gt; &lt;span class="n"&gt;s&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="nb"&gt;String&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;value&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;trim&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
  &lt;span class="n"&gt;s&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="n"&gt;s&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;split&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="s1"&gt;'$'&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;join&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="s1"&gt;''&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;split&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="s1"&gt;','&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;join&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="s1"&gt;''&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;
       &lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;split&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="nb"&gt;String&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;fromCharCode&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="mh"&gt;0x2212&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;join&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="s1"&gt;'-'&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
  &lt;span class="kd"&gt;let&lt;/span&gt; &lt;span class="n"&gt;negative&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="kc"&gt;false&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;

  &lt;span class="k"&gt;if&lt;/span&gt; &lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;s&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;startsWith&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="s1"&gt;'('&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="o"&gt;&amp;amp;&amp;amp;&lt;/span&gt; &lt;span class="n"&gt;s&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;endsWith&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="s1"&gt;')'&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="n"&gt;negative&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="kc"&gt;true&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt; &lt;span class="n"&gt;s&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="n"&gt;s&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;slice&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="mi"&gt;1&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="o"&gt;-&lt;/span&gt;&lt;span class="mi"&gt;1&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;trim&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt; &lt;span class="p"&gt;}&lt;/span&gt;

  &lt;span class="kd"&gt;const&lt;/span&gt; &lt;span class="n"&gt;upper&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="n"&gt;s&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;toUpperCase&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
  &lt;span class="k"&gt;if&lt;/span&gt; &lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;upper&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;endsWith&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="s1"&gt;'DR'&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="n"&gt;negative&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="kc"&gt;true&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt; &lt;span class="n"&gt;s&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="n"&gt;s&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;slice&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="mi"&gt;0&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="o"&gt;-&lt;/span&gt;&lt;span class="mi"&gt;2&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;trim&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt; &lt;span class="p"&gt;}&lt;/span&gt;
  &lt;span class="k"&gt;else&lt;/span&gt; &lt;span class="k"&gt;if&lt;/span&gt; &lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;upper&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;endsWith&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="s1"&gt;'CR'&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="n"&gt;s&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="n"&gt;s&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;slice&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="mi"&gt;0&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="o"&gt;-&lt;/span&gt;&lt;span class="mi"&gt;2&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;trim&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt; &lt;span class="p"&gt;}&lt;/span&gt;

  &lt;span class="k"&gt;if&lt;/span&gt; &lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;s&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;startsWith&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="s1"&gt;'-'&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="n"&gt;negative&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="kc"&gt;true&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt; &lt;span class="n"&gt;s&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="n"&gt;s&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;slice&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="mi"&gt;1&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;trim&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt; &lt;span class="p"&gt;}&lt;/span&gt;
  &lt;span class="k"&gt;else&lt;/span&gt; &lt;span class="k"&gt;if&lt;/span&gt; &lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;s&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;startsWith&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="s1"&gt;'+'&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="n"&gt;s&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="n"&gt;s&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;slice&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="mi"&gt;1&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;trim&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt; &lt;span class="p"&gt;}&lt;/span&gt;

  &lt;span class="k"&gt;if&lt;/span&gt; &lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;s&lt;/span&gt; &lt;span class="o"&gt;===&lt;/span&gt; &lt;span class="s1"&gt;''&lt;/span&gt; &lt;span class="o"&gt;||&lt;/span&gt; &lt;span class="nf"&gt;isNaN&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="nb"&gt;Number&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;s&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="k"&gt;return&lt;/span&gt; &lt;span class="kc"&gt;null&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
  &lt;span class="kd"&gt;const&lt;/span&gt; &lt;span class="n"&gt;n&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="nb"&gt;parseFloat&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;s&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
  &lt;span class="k"&gt;return&lt;/span&gt; &lt;span class="n"&gt;negative&lt;/span&gt; &lt;span class="o"&gt;?&lt;/span&gt; &lt;span class="o"&gt;-&lt;/span&gt;&lt;span class="n"&gt;n&lt;/span&gt; &lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="n"&gt;n&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
&lt;span class="p"&gt;}&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Note the order: strip the marker, then the explicit sign, then apply the negative flag once at the end.&lt;/p&gt;

&lt;h2&gt;
  
  
  PDF text is not lines
&lt;/h2&gt;

&lt;p&gt;&lt;code&gt;getTextContent()&lt;/code&gt; gives you items with a transform matrix; each item has an x (matrix[4]) and a y (matrix[5]). Rebuild lines yourself:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight javascript"&gt;&lt;code&gt;&lt;span class="kd"&gt;const&lt;/span&gt; &lt;span class="n"&gt;buckets&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
&lt;span class="k"&gt;for&lt;/span&gt; &lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="kd"&gt;const&lt;/span&gt; &lt;span class="n"&gt;it&lt;/span&gt; &lt;span class="k"&gt;of&lt;/span&gt; &lt;span class="n"&gt;items&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt;
  &lt;span class="k"&gt;if&lt;/span&gt; &lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="o"&gt;!&lt;/span&gt;&lt;span class="n"&gt;it&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="n"&gt;str&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="k"&gt;continue&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
  &lt;span class="kd"&gt;const&lt;/span&gt; &lt;span class="n"&gt;x&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="n"&gt;it&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="n"&gt;transform&lt;/span&gt;&lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="mi"&gt;4&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
  &lt;span class="kd"&gt;const&lt;/span&gt; &lt;span class="n"&gt;y&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="n"&gt;it&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="n"&gt;transform&lt;/span&gt;&lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="mi"&gt;5&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
  &lt;span class="kd"&gt;let&lt;/span&gt; &lt;span class="n"&gt;b&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="n"&gt;buckets&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;find&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;b&lt;/span&gt; &lt;span class="o"&gt;=&amp;gt;&lt;/span&gt; &lt;span class="nb"&gt;Math&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;abs&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;b&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="n"&gt;y&lt;/span&gt; &lt;span class="o"&gt;-&lt;/span&gt; &lt;span class="n"&gt;y&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="o"&gt;&amp;lt;=&lt;/span&gt; &lt;span class="mf"&gt;2.5&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
  &lt;span class="k"&gt;if&lt;/span&gt; &lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="o"&gt;!&lt;/span&gt;&lt;span class="n"&gt;b&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="n"&gt;b&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="n"&gt;y&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;items&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt; &lt;span class="p"&gt;}&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt; &lt;span class="n"&gt;buckets&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;push&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;b&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt; &lt;span class="p"&gt;}&lt;/span&gt;
  &lt;span class="n"&gt;b&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="n"&gt;items&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;push&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="n"&gt;x&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;s&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="n"&gt;it&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="n"&gt;str&lt;/span&gt; &lt;span class="p"&gt;}&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
&lt;span class="p"&gt;}&lt;/span&gt;
&lt;span class="n"&gt;buckets&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;sort&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;a&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;b&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="o"&gt;=&amp;gt;&lt;/span&gt; &lt;span class="n"&gt;b&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="n"&gt;y&lt;/span&gt; &lt;span class="o"&gt;-&lt;/span&gt; &lt;span class="n"&gt;a&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="n"&gt;y&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;          &lt;span class="c1"&gt;// top to bottom&lt;/span&gt;
&lt;span class="kd"&gt;const&lt;/span&gt; &lt;span class="n"&gt;lines&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="n"&gt;buckets&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;map&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;b&lt;/span&gt; &lt;span class="o"&gt;=&amp;gt;&lt;/span&gt; &lt;span class="n"&gt;b&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="n"&gt;items&lt;/span&gt;
  &lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;sort&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;a&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;b&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="o"&gt;=&amp;gt;&lt;/span&gt; &lt;span class="n"&gt;a&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="n"&gt;x&lt;/span&gt; &lt;span class="o"&gt;-&lt;/span&gt; &lt;span class="n"&gt;b&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="n"&gt;x&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;                &lt;span class="c1"&gt;// left to right&lt;/span&gt;
  &lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;map&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;i&lt;/span&gt; &lt;span class="o"&gt;=&amp;gt;&lt;/span&gt; &lt;span class="n"&gt;i&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="n"&gt;s&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;join&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="s1"&gt;' '&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;The 2.5 tolerance is the part that needs patience: too tight and one line splits in two, too loose and two transactions merge.&lt;/p&gt;

&lt;h2&gt;
  
  
  Two layouts, one parser
&lt;/h2&gt;

&lt;p&gt;Some banks print one transaction per line:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight plaintext"&gt;&lt;code&gt;10/02/2026  ONLINE PAYMENT TO ADOBE INC   -54.99   4,945.01
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Others print one column per line:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight plaintext"&gt;&lt;code&gt;10/07/2026
CARD PURCHASE STAPLES #1123
64.28 DR
7,380.73
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Both are handled by grouping lines into blocks that start at a date, and stopping collection once the block already holds two amounts. Without that stop, the footer closing balance gets appended to the last transaction, and the running balance is wrong for exactly the rows people check first.&lt;/p&gt;

&lt;p&gt;Within a block, if there are two or more numbers, the last is the balance and the first non-zero one before it is the transaction. The description is whatever is left after removing the matched amounts.&lt;/p&gt;

&lt;h2&gt;
  
  
  Fail loudly on scans
&lt;/h2&gt;

&lt;p&gt;An image-only PDF produces zero characters of text. The friendly-looking bug is to show an empty table, which reads as 'your statement has no transactions'. The tool checks the extracted character count and says what is actually true: no selectable text layer — run OCR first, or download the PDF version from your bank.&lt;/p&gt;

&lt;h2&gt;
  
  
  The regex bug worth naming
&lt;/h2&gt;

&lt;p&gt;My amount pattern was:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight javascript"&gt;&lt;code&gt;&lt;span class="o"&gt;/&lt;/span&gt;&lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="mi"&gt;0&lt;/span&gt;&lt;span class="o"&gt;-&lt;/span&gt;&lt;span class="mi"&gt;9&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt;&lt;span class="p"&gt;{&lt;/span&gt;&lt;span class="mi"&gt;1&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;&lt;span class="mi"&gt;3&lt;/span&gt;&lt;span class="p"&gt;}&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="o"&gt;?:&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;&lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="mi"&gt;0&lt;/span&gt;&lt;span class="o"&gt;-&lt;/span&gt;&lt;span class="mi"&gt;9&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt;&lt;span class="p"&gt;{&lt;/span&gt;&lt;span class="mi"&gt;3&lt;/span&gt;&lt;span class="p"&gt;}&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="o"&gt;*&lt;/span&gt;&lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt;&lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="mi"&gt;0&lt;/span&gt;&lt;span class="o"&gt;-&lt;/span&gt;&lt;span class="mi"&gt;9&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt;&lt;span class="p"&gt;{&lt;/span&gt;&lt;span class="mi"&gt;2&lt;/span&gt;&lt;span class="p"&gt;}&lt;/span&gt;&lt;span class="o"&gt;/&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;For &lt;code&gt;1,850.00&lt;/code&gt; it matches all of it. For &lt;code&gt;1850.00&lt;/code&gt; — an ordinary way to print four digits — it matches &lt;code&gt;850.00&lt;/code&gt;, silently. No error, no empty row: a wrong number that looks right.&lt;/p&gt;

&lt;p&gt;The fix is an explicit alternation:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight javascript"&gt;&lt;code&gt;&lt;span class="o"&gt;/&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="o"&gt;?:&lt;/span&gt;&lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="mi"&gt;0&lt;/span&gt;&lt;span class="o"&gt;-&lt;/span&gt;&lt;span class="mi"&gt;9&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt;&lt;span class="p"&gt;{&lt;/span&gt;&lt;span class="mi"&gt;1&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;&lt;span class="mi"&gt;3&lt;/span&gt;&lt;span class="p"&gt;}&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="o"&gt;?:&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;&lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="mi"&gt;0&lt;/span&gt;&lt;span class="o"&gt;-&lt;/span&gt;&lt;span class="mi"&gt;9&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt;&lt;span class="p"&gt;{&lt;/span&gt;&lt;span class="mi"&gt;3&lt;/span&gt;&lt;span class="p"&gt;}&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="o"&gt;+&lt;/span&gt;&lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt;&lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="mi"&gt;0&lt;/span&gt;&lt;span class="o"&gt;-&lt;/span&gt;&lt;span class="mi"&gt;9&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt;&lt;span class="p"&gt;{&lt;/span&gt;&lt;span class="mi"&gt;2&lt;/span&gt;&lt;span class="p"&gt;}&lt;/span&gt;&lt;span class="o"&gt;|&lt;/span&gt;&lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="mi"&gt;0&lt;/span&gt;&lt;span class="o"&gt;-&lt;/span&gt;&lt;span class="mi"&gt;9&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt;&lt;span class="o"&gt;+&lt;/span&gt;&lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt;&lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="mi"&gt;0&lt;/span&gt;&lt;span class="o"&gt;-&lt;/span&gt;&lt;span class="mi"&gt;9&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt;&lt;span class="p"&gt;{&lt;/span&gt;&lt;span class="mi"&gt;2&lt;/span&gt;&lt;span class="p"&gt;}&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="o"&gt;/&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;The grouped form now requires at least one comma; everything else falls back to plain digits. I only found it by writing a test with an ungrouped amount, not by reading the code.&lt;/p&gt;

&lt;h2&gt;
  
  
  The tests that matter
&lt;/h2&gt;

&lt;p&gt;If you maintain anything like this, this list catches most of the damage:&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;
&lt;code&gt;-64.28&lt;/code&gt;, &lt;code&gt;64.28 DR&lt;/code&gt;, &lt;code&gt;(123.45)&lt;/code&gt;, &lt;code&gt;100.00 CR&lt;/code&gt;, &lt;code&gt;−64.28&lt;/code&gt; resolve to the right sign&lt;/li&gt;
&lt;li&gt;a minus plus DR does not flip twice&lt;/li&gt;
&lt;li&gt;a split-line block keeps its own balance&lt;/li&gt;
&lt;li&gt;amounts with and without thousands separators&lt;/li&gt;
&lt;li&gt;CSV quoting for a description containing a comma or a quote&lt;/li&gt;
&lt;li&gt;an image-only PDF reports no text layer, not an empty table&lt;/li&gt;
&lt;/ol&gt;

&lt;h2&gt;
  
  
  Try it
&lt;/h2&gt;

&lt;p&gt;The free page converts up to 20 rows per file and still shows the full transaction count with an in/out/net summary: &lt;a href="https://yougkxm.github.io/bank-to-csv/" rel="noopener noreferrer"&gt;https://yougkxm.github.io/bank-to-csv/&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;The desktop version exports every row, writes a formatted .xlsx workbook and batches a folder of statements. Early-bird link: &lt;a href="https://8299181964131.gumroad.com/l/bank-statement-to-excel/EARLY50" rel="noopener noreferrer"&gt;https://8299181964131.gumroad.com/l/bank-statement-to-excel/EARLY50&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;No OCR, no categorising, no tax logic. It reads electronic statements, gets the sign right once, and tells you when it cannot read the file at all.&lt;/p&gt;

</description>
      <category>javascript</category>
      <category>privacy</category>
      <category>finance</category>
      <category>webdev</category>
    </item>
    <item>
      <title>The listing image size is the part of a product photo nobody checks until Etsy crops it</title>
      <dc:creator>Seller Tools</dc:creator>
      <pubDate>Thu, 19 Nov 2026 03:44:54 +0000</pubDate>
      <link>https://dev.to/yougkxm/the-listing-image-size-is-the-part-of-a-product-photo-nobody-checks-until-etsy-crops-it-2641</link>
      <guid>https://dev.to/yougkxm/the-listing-image-size-is-the-part-of-a-product-photo-nobody-checks-until-etsy-crops-it-2641</guid>
      <description>&lt;p&gt;Most "my photos look bad" problems in a shop are not photography problems. They are the wrong aspect ratio, or a label sitting exactly where the marketplace draws its own button.&lt;/p&gt;

&lt;p&gt;Every marketplace publishes a size for each image slot, and they are all different:&lt;/p&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Slot&lt;/th&gt;
&lt;th&gt;Pixels&lt;/th&gt;
&lt;th&gt;Ratio&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;Etsy listing photo&lt;/td&gt;
&lt;td&gt;2700 x 2025&lt;/td&gt;
&lt;td&gt;4:3&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Gumroad cover&lt;/td&gt;
&lt;td&gt;1280 x 720&lt;/td&gt;
&lt;td&gt;16:9&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Gumroad library thumbnail&lt;/td&gt;
&lt;td&gt;600 x 600&lt;/td&gt;
&lt;td&gt;1:1&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Pinterest standard pin&lt;/td&gt;
&lt;td&gt;1000 x 1500&lt;/td&gt;
&lt;td&gt;2:3&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Shopify product master&lt;/td&gt;
&lt;td&gt;2048 x 2048&lt;/td&gt;
&lt;td&gt;1:1&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Instagram post / story&lt;/td&gt;
&lt;td&gt;1080 x 1080 / 1080 x 1920&lt;/td&gt;
&lt;td&gt;1:1, 9:16&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Amazon main image&lt;/td&gt;
&lt;td&gt;2000 x 2000&lt;/td&gt;
&lt;td&gt;1:1&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;eBay listing&lt;/td&gt;
&lt;td&gt;1600 x 1600&lt;/td&gt;
&lt;td&gt;1:1&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;KDP eBook cover&lt;/td&gt;
&lt;td&gt;1600 x 2560&lt;/td&gt;
&lt;td&gt;1:1.6&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;A4 print at 300 dpi&lt;/td&gt;
&lt;td&gt;2480 x 3508&lt;/td&gt;
&lt;td&gt;1:1.414&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;Two things go wrong with almost every one of them.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;1. The crop is not centred.&lt;/strong&gt; A 16:9 photo dropped into a 4:3 slot loses a full quarter of its width, and the marketplace decides which quarter. Text near the edge goes first.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;2. Text lands in the overlay zone.&lt;/strong&gt; Instagram puts the caption and the action icons over the bottom of a feed image, a story covers roughly the top 250 px and bottom 400 px with its own UI, and an Etsy shop banner has the shop icon sitting in the middle. A safe zone is the fix: an inset where text and the product edge can live without being covered.&lt;/p&gt;

&lt;p&gt;So I built a small tool for that instead of opening a design app every time.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;&lt;a href="https://yougkxm.github.io/listing-kit/" rel="noopener noreferrer"&gt;Listing &amp;amp; Mockup Frame Kit&lt;/a&gt;&lt;/strong&gt; - free, in the browser, no account.&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;Drop or pick one product photo (or press "Use sample product" to try it with nothing at hand).&lt;/li&gt;
&lt;li&gt;Pick a marketplace size.&lt;/li&gt;
&lt;li&gt;Turn on the safe-zone guide, and optionally a badge, a callout box or a price tag.&lt;/li&gt;
&lt;li&gt;Export a PNG at the exact pixel size.&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;The whole thing is one HTML file with a Canvas 2D renderer. No build step, no framework, no server: &lt;code&gt;drawImage&lt;/code&gt; places the photo with a cover crop, the guides are drawn with &lt;code&gt;setLineDash&lt;/code&gt;, the overlays are measured with &lt;code&gt;measureText&lt;/code&gt; and scaled by &lt;code&gt;Math.min(w, h) / 1200&lt;/code&gt; so a badge looks the same on a 600 px thumbnail and a 3508 px A4 page. Export is &lt;code&gt;canvas.toBlob("image/png")&lt;/code&gt; and an &lt;code&gt;&amp;lt;a download&amp;gt;&lt;/code&gt;. The photo never leaves the browser - there is nothing to upload it to.&lt;/p&gt;

&lt;p&gt;Etsy 2700 x 2025, Gumroad 1280 x 720 and 600 x 600 are free, along with the guides and overlays. The other twelve sizes and a one-page size sheet are a $12 one-time unlock with a Gumroad license key, verified against &lt;code&gt;api.gumroad.com/v2/licenses/verify&lt;/code&gt; and cached in &lt;code&gt;localStorage&lt;/code&gt; - the free part keeps working either way.&lt;/p&gt;

&lt;p&gt;If you sell on more than one marketplace, the useful habit is to export the widest ratio first and crop from that file, rather than exporting from the original each time. It keeps the framing consistent across a shop.&lt;/p&gt;

&lt;p&gt;There is also a free &lt;a href="https://yougkxm.github.io/" rel="noopener noreferrer"&gt;profit and break-even calculator&lt;/a&gt; on the same domain if the next problem after the image is the price.&lt;/p&gt;

&lt;p&gt;&lt;em&gt;Sizes follow each marketplace's published guidance as of November 2026; check your own seller dashboard before a large batch. I build both tools.&lt;/em&gt;&lt;/p&gt;

</description>
      <category>design</category>
      <category>ecommerce</category>
    </item>
    <item>
      <title>Three fees that quietly drain your Etsy and Gumroad margin (and the break-even formula that counts them)</title>
      <dc:creator>Seller Tools</dc:creator>
      <pubDate>Thu, 19 Nov 2026 02:33:22 +0000</pubDate>
      <link>https://dev.to/yougkxm/three-fees-that-quietly-drain-your-etsy-and-gumroad-margin-and-the-break-even-formula-that-counts-569a</link>
      <guid>https://dev.to/yougkxm/three-fees-that-quietly-drain-your-etsy-and-gumroad-margin-and-the-break-even-formula-that-counts-569a</guid>
      <description>&lt;p&gt;A $25 Etsy order. Product cost $8, shipping $4, 10% of orders come back as refunds, $1 of ad spend per order.&lt;/p&gt;

&lt;p&gt;A normal price sheet says you make &lt;strong&gt;$10.18&lt;/strong&gt;.&lt;/p&gt;

&lt;p&gt;Your payout says you made &lt;strong&gt;$7.41&lt;/strong&gt;.&lt;/p&gt;

&lt;p&gt;That is 37% more than you actually made, or $276 per 100 orders, and the gap comes from three places a per-unit calculation never looks: the fixed part of card processing, Offsite Ads charged as a percentage of the whole order, and the money already spent on the parcels that came back.&lt;/p&gt;

&lt;p&gt;Here is each one, then the formula that puts them back where they belong.&lt;/p&gt;

&lt;h2&gt;
  
  
  1. The fixed part of payment processing does not come back
&lt;/h2&gt;

&lt;p&gt;Card processing is quoted as a percentage plus cents. On Etsy it is 3% + $0.25. On Gumroad Direct the platform takes 10% + $0.50 and the card processor about 2.9% + $0.30, so roughly 12.9% + $0.80 per sale. On a $5 digital download that is $1.45, or 29% of the price, before you have paid for anything else.&lt;/p&gt;

&lt;p&gt;The percentage part sometimes gets credited back when you refund. The cents almost never do, and the $0.20 Etsy listing fee does not come back either, because the listing was published whether or not the sale completed. Those amounts are small per order and charged on refunded orders all the same, which is why they never turn up in a price sheet: invisible on a single sale, constant across a thousand.&lt;/p&gt;

&lt;p&gt;Whether the percentage part returns on a refund depends on the platform, payment method and country, and it changes. The model below assumes percentage fees apply to the orders you kept and fixed fees to every order.&lt;/p&gt;

&lt;h2&gt;
  
  
  2. Offsite Ads is 12% or 15% of the order, not a marketing budget
&lt;/h2&gt;

&lt;p&gt;Etsy charges 12% or 15% on orders that an offsite ad brought in. Once a shop passes the platform's sales threshold, participation is not optional, so this is not an expense you can skip in a slow month: it arrives as a percentage of the order total whenever an outside ad produced the sale.&lt;/p&gt;

&lt;p&gt;Same order: net profit $7.41 at a $25 price. If it was attributed to an offsite ad at 15%, $3.75 goes to the ad fee and $3.66 is left. At 12% it is $4.41.&lt;/p&gt;

&lt;p&gt;Now a thinner product: $20 price, same $8 cost, $4 shipping, $1 ad spend and a 10% return rate yields $3.34, a 16.7% margin. A single 15% offsite-attributed sale on it loses money. The listing looked healthy in the spreadsheet because the spreadsheet never asked how the buyer arrived.&lt;/p&gt;

&lt;p&gt;A percentage-of-order fee is not a fixed cost you amortise across a month. It lands on the sales that were most expensive to acquire, which are often the only ones you would not have had otherwise.&lt;/p&gt;

&lt;h2&gt;
  
  
  3. Returns cost the outbound leg and the handling, not just the refund
&lt;/h2&gt;

&lt;p&gt;A refund returns the buyer's money. It does not return the postage you paid to send the parcel, the return label you pay for, or the repacking, and the goods come back to sit in stock. On a $25 order with a 10% return rate and $3 of return handling, that is $0.30 per order of handling plus $0.40 of outbound shipping spread across the orders that came back.&lt;/p&gt;

&lt;p&gt;For physical products this is the second-biggest number after product cost, and the one most often left out. For digital products it is small, but the fixed fees from section 1 still apply, and so does the ad spend on an order that was not kept.&lt;/p&gt;

&lt;h2&gt;
  
  
  The formula
&lt;/h2&gt;

&lt;p&gt;Let:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;code&gt;P&lt;/code&gt; price&lt;/li&gt;
&lt;li&gt;
&lt;code&gt;C&lt;/code&gt; product cost per unit&lt;/li&gt;
&lt;li&gt;
&lt;code&gt;S&lt;/code&gt; shipping you pay per dispatched order&lt;/li&gt;
&lt;li&gt;
&lt;code&gt;r&lt;/code&gt; return rate&lt;/li&gt;
&lt;li&gt;
&lt;code&gt;K&lt;/code&gt; cost per return (label, repack, restocking)&lt;/li&gt;
&lt;li&gt;
&lt;code&gt;A&lt;/code&gt; ad spend per order (ad budget divided by orders, not per click)&lt;/li&gt;
&lt;li&gt;
&lt;code&gt;f&lt;/code&gt; platform percentage fee&lt;/li&gt;
&lt;li&gt;
&lt;code&gt;p&lt;/code&gt; payment processing percentage&lt;/li&gt;
&lt;li&gt;
&lt;code&gt;F&lt;/code&gt; fixed per-order fees (card cents, platform cents)&lt;/li&gt;
&lt;li&gt;
&lt;code&gt;L&lt;/code&gt; listing or fixed fee per order&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;Then the profit that actually reaches your account, averaged across kept and refunded orders:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight plaintext"&gt;&lt;code&gt;net = P*(1-r)*(1-f-p) - C*(1-r) - S - A - F - L - K*r
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Two things in there matter. &lt;code&gt;(1-r)&lt;/code&gt; multiplies both the revenue and the product cost, because a returned item is resold: you lose the revenue, not the goods. The other charges are not multiplied, because shipping, ads, listing and fixed fees were all spent on the order that came back.&lt;/p&gt;

&lt;p&gt;Set &lt;code&gt;net = 0&lt;/code&gt; and you get the break-even price:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight plaintext"&gt;&lt;code&gt;P_breakeven = ( C*(1-r) + S + A + F + L + K*r ) / ( (1-r)*(1-f-p) )
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;For a target net margin &lt;code&gt;m&lt;/code&gt;, divide by &lt;code&gt;(1-r)*(1-f-p) - m&lt;/code&gt; instead:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight plaintext"&gt;&lt;code&gt;P_target = ( C*(1-r) + S + A + F + L + K*r ) / ( (1-r)*(1-f-p) - m )
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Note the denominator: a target margin at or above &lt;code&gt;(1-r)*(1-f-p)&lt;/code&gt; has no solution. No price reaches it, and the honest answer is that the cost structure, not the price, has to change.&lt;/p&gt;

&lt;h2&gt;
  
  
  The same order, worked through
&lt;/h2&gt;

&lt;p&gt;Etsy preset: price $25, cost $8, shipping $4, return rate 10%, return handling $3, ad spend $1, target margin 20%.&lt;/p&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Line&lt;/th&gt;
&lt;th&gt;Per order&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;Revenue kept (90 of 100 orders)&lt;/td&gt;
&lt;td&gt;$22.50&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Platform fee 6.5%&lt;/td&gt;
&lt;td&gt;$1.46&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Payment processing 3%&lt;/td&gt;
&lt;td&gt;$0.68&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Fixed card fee&lt;/td&gt;
&lt;td&gt;$0.25&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Listing fee&lt;/td&gt;
&lt;td&gt;$0.20&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Product cost&lt;/td&gt;
&lt;td&gt;$7.20&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Shipping you pay&lt;/td&gt;
&lt;td&gt;$4.00&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Return handling&lt;/td&gt;
&lt;td&gt;$0.30&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Ad spend&lt;/td&gt;
&lt;td&gt;$1.00&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;strong&gt;Net profit&lt;/strong&gt;&lt;/td&gt;
&lt;td&gt;&lt;strong&gt;$7.41&lt;/strong&gt;&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;ul&gt;
&lt;li&gt;Net margin: 29.7%&lt;/li&gt;
&lt;li&gt;Break-even price: &lt;strong&gt;$15.90&lt;/strong&gt;
&lt;/li&gt;
&lt;li&gt;Price for a 20% net margin: &lt;strong&gt;$21.07&lt;/strong&gt;
&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;Those are the numbers this model produces. A price sheet computing price minus fees minus cost produces $10.18 for the same order.&lt;/p&gt;

&lt;p&gt;A discount is where this bites. At $20 the order still clears break-even and makes $3.34, a 16.7% margin. At $18 it makes $1.71, 9.5%. At $15.90 it makes nothing. Volume at $18 is not a growth strategy if it does not cover the returns and the ads that produced the volume.&lt;/p&gt;

&lt;h2&gt;
  
  
  What I would check this quarter
&lt;/h2&gt;

&lt;ol&gt;
&lt;li&gt;Pull one month of your platform's fee report and compare the percentages you assumed with the ones actually withheld. Offsite-attributed orders are the usual gap.&lt;/li&gt;
&lt;li&gt;Compute break-even for your three cheapest products. If your sale price is within 10% of it, one return undoes several orders.&lt;/li&gt;
&lt;li&gt;Watch the channel mix on digital products. Gumroad Discover takes a flat 30%; Direct takes 10% + $0.50 plus card processing. The same file is profitable on one channel and thin on the other, and the product page does not show the difference.&lt;/li&gt;
&lt;li&gt;Recheck rates when the platform changes them. Every number above is the standard rate as of November 2026 and every one of them is editable, because your country, shop and payment method may differ.&lt;/li&gt;
&lt;/ol&gt;

&lt;h2&gt;
  
  
  If you want the arithmetic done for you
&lt;/h2&gt;

&lt;p&gt;I build small tools for sellers, so two of these are mine:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;&lt;a href="https://yougkxm.github.io/" rel="noopener noreferrer"&gt;Seller Profit &amp;amp; Break-Even Calculator&lt;/a&gt;&lt;/strong&gt; — free, no signup, no account. Fee presets for Etsy, Gumroad Direct, Gumroad Discover and Shopify, all editable, a 12%/15% Offsite Ads switch, return rate and return cost, and it solves the break-even and target-margin price. It runs entirely in your browser and works offline once loaded; no numbers leave the page.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;&lt;a href="https://8299181964131.gumroad.com/l/seller-profit-tax-calculator/LAUNCH25" rel="noopener noreferrer"&gt;Seller Profit &amp;amp; Tax Calculator 2026&lt;/a&gt;&lt;/strong&gt; ($19.99, LAUNCH25 takes it to $14.99) — the same fee logic across several SKUs in Excel or Google Sheets, plus a monthly ledger and a quarterly tax set-aside for Schedule C filers. There is a free single-product &lt;strong&gt;&lt;a href="https://8299181964131.gumroad.com/l/seller-profit-calculator-lite" rel="noopener noreferrer"&gt;Lite edition&lt;/a&gt;&lt;/strong&gt; if you want to try the fee model first.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;If you only want the formula, it is above and it is free. That is the part that matters most.&lt;/p&gt;

</description>
      <category>etsy</category>
      <category>ecommerce</category>
      <category>pricing</category>
      <category>gumroad</category>
    </item>
    <item>
      <title>Turning 300 reviews into a ranked list of the five things buyers actually hate</title>
      <dc:creator>Seller Tools</dc:creator>
      <pubDate>Mon, 16 Nov 2026 21:10:29 +0000</pubDate>
      <link>https://dev.to/yougkxm/turning-300-reviews-into-a-ranked-list-of-the-five-things-buyers-actually-hate-c6f</link>
      <guid>https://dev.to/yougkxm/turning-300-reviews-into-a-ranked-list-of-the-five-things-buyers-actually-hate-c6f</guid>
      <description>&lt;h1&gt;
  
  
  Turning 300 reviews into a ranked list of the five things buyers actually hate
&lt;/h1&gt;




&lt;p&gt;Star ratings are a terrible product research tool. A 4.6 average can hide the fact that&lt;br&gt;
every third review mentions the same broken zipper, because the people who liked the product&lt;br&gt;
wrote four words and the people who didn't wrote a paragraph.&lt;/p&gt;

&lt;p&gt;If you sell anything online — on Etsy, Shopify, Amazon, or your own store — the useful signal&lt;br&gt;
is in the text of the 1–3 star reviews, and the useful output is not "sentiment: negative".&lt;br&gt;
It is a ranked list of recurring complaints, each with a count and a representative quote.&lt;/p&gt;

&lt;p&gt;You can build that in an afternoon with a keyword lexicon, a bit of regex, and a Counter.&lt;/p&gt;
&lt;h2&gt;
  
  
  1. Decide what to collect before you collect anything
&lt;/h2&gt;

&lt;p&gt;Three rules that decide whether the output is a signal or noise:&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;
&lt;strong&gt;1–3 stars only.&lt;/strong&gt; 4–5 star reviews contain gold too (the phrases buyers use), but they
answer a different question. Keep them out of the complaint count.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Last 6–12 months.&lt;/strong&gt; Reviews from 2023 describe a product that may no longer exist.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Reviews with at least one sentence of detail.&lt;/strong&gt; "Bad quality" is a label, not a
complaint. "The stitching came apart at the seam after one wash" tells you what to fix.&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;Fifty to a hundred detailed negative reviews per competitor is usually enough to see the&lt;br&gt;
pattern. If you only get twelve, treat the ranking as a hypothesis, not a finding.&lt;/p&gt;
&lt;h2&gt;
  
  
  2. Stop treating it as sentiment analysis
&lt;/h2&gt;

&lt;p&gt;A sentiment model will tell you that 78% of the reviews are negative. You already knew that:&lt;br&gt;
you filtered by star rating. What you need is &lt;strong&gt;aspect extraction&lt;/strong&gt; — which &lt;em&gt;part&lt;/em&gt; of the&lt;br&gt;
experience the review is about.&lt;/p&gt;

&lt;p&gt;Start with a flat taxonomy of about ten aspects that covers most physical and digital products:&lt;/p&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Aspect&lt;/th&gt;
&lt;th&gt;Words reviewers use&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;Quality / durability&lt;/td&gt;
&lt;td&gt;broke, flimsy, cheap, fell apart, cheaply made, sturdy, well made&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Sizing / fit&lt;/td&gt;
&lt;td&gt;small, large, runs big, runs small, too tight, sizing chart, fit&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Product accuracy&lt;/td&gt;
&lt;td&gt;not as pictured, different colour, color, smaller than, misleading, as described&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Shipping&lt;/td&gt;
&lt;td&gt;late, delayed, never arrived, tracking, took weeks, fast shipping&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Packaging&lt;/td&gt;
&lt;td&gt;box, crushed, damaged, wrapped, dented, leaked&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Customer service&lt;/td&gt;
&lt;td&gt;no reply, ignored, rude, helpful, refund, return&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Value&lt;/td&gt;
&lt;td&gt;overpriced, not worth, expensive, for the price&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Instructions / usability&lt;/td&gt;
&lt;td&gt;instructions, confusing, no guide, setup, figured it out&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Missing features&lt;/td&gt;
&lt;td&gt;wish it had, would be perfect if, doesn't come with&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Digital file issues&lt;/td&gt;
&lt;td&gt;can't open, file, format, download, editable, layers, fonts&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;This taxonomy is not a model. It is a list, and that is the point: you can read it, argue with&lt;br&gt;
it, and add the words your buyers actually use.&lt;/p&gt;
&lt;h2&gt;
  
  
  3. Negation is where naive keyword counting lies
&lt;/h2&gt;

&lt;p&gt;"The box arrived crushed" is a packaging complaint. "The box was not crushed at all" is a&lt;br&gt;
compliment that keyword counting will file under &lt;em&gt;packaging problems&lt;/em&gt;. English reviews are full&lt;br&gt;
of "no", "not", "never", "wasn't", and they flip the meaning of the phrase that follows.&lt;/p&gt;

&lt;p&gt;A cheap, effective fix: look at the 40 characters before the keyword and check for a negator.&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight python"&gt;&lt;code&gt;&lt;span class="no"&gt;NEGATORS&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="s2"&gt;"not "&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="s2"&gt;"n't"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="s2"&gt;"never"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="s2"&gt;"no "&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="s2"&gt;"without"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="s2"&gt;"hardly"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="s2"&gt;"barely"&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;

&lt;span class="k"&gt;def&lt;/span&gt; &lt;span class="nf"&gt;is_negated&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;text&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;start&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt;
    &lt;span class="n"&gt;window&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="n"&gt;text&lt;/span&gt;&lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="nb"&gt;max&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="mi"&gt;0&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;start&lt;/span&gt; &lt;span class="o"&gt;-&lt;/span&gt; &lt;span class="mi"&gt;40&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt;&lt;span class="n"&gt;start&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;lower&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;
    &lt;span class="k"&gt;return&lt;/span&gt; &lt;span class="nb"&gt;any&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;n&lt;/span&gt; &lt;span class="k"&gt;in&lt;/span&gt; &lt;span class="n"&gt;window&lt;/span&gt; &lt;span class="k"&gt;for&lt;/span&gt; &lt;span class="n"&gt;n&lt;/span&gt; &lt;span class="k"&gt;in&lt;/span&gt; &lt;span class="no"&gt;NEGATORS&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;It is not grammatical analysis, and it will occasionally be wrong. It removes the most&lt;br&gt;
embarrassing errors, which is all you need for a ranked list a human will read.&lt;/p&gt;
&lt;h2&gt;
  
  
  4. Count, then sort by how much it would cost you
&lt;/h2&gt;

&lt;p&gt;Frequency alone is a trap: "shipping was slow" appears in every category because shipping is&lt;br&gt;
everyone's complaint. Weight the aspects by how much they matter to a buyer deciding whether to&lt;br&gt;
keep the product:&lt;/p&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Weight&lt;/th&gt;
&lt;th&gt;Aspects&lt;/th&gt;
&lt;th&gt;Why&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;3&lt;/td&gt;
&lt;td&gt;quality, accuracy, sizing/fit&lt;/td&gt;
&lt;td&gt;the buyer cannot use the product&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;2&lt;/td&gt;
&lt;td&gt;missing features, digital file issues, instructions&lt;/td&gt;
&lt;td&gt;the product is incomplete&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;1&lt;/td&gt;
&lt;td&gt;shipping, packaging, service, value&lt;/td&gt;
&lt;td&gt;friction, not failure&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;Then rank by &lt;code&gt;count × weight&lt;/code&gt;. A category with 40 shipping complaints and 8 that say "fell&lt;br&gt;
apart in a week" is telling you something different from what the raw counts suggest.&lt;/p&gt;
&lt;h2&gt;
  
  
  5. A working script, no dependencies
&lt;/h2&gt;


&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight python"&gt;&lt;code&gt;&lt;span class="k"&gt;import&lt;/span&gt; &lt;span class="n"&gt;csv&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;re&lt;/span&gt;
&lt;span class="k"&gt;from&lt;/span&gt; &lt;span class="n"&gt;collections&lt;/span&gt; &lt;span class="k"&gt;import&lt;/span&gt; &lt;span class="nc"&gt;Counter&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;defaultdict&lt;/span&gt;

&lt;span class="no"&gt;ASPECTS&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt;
    &lt;span class="s2"&gt;"quality"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="s2"&gt;"broke"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="s2"&gt;"flimsy"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="s2"&gt;"cheaply made"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="s2"&gt;"fell apart"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="s2"&gt;"cheap material"&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
    &lt;span class="s2"&gt;"sizing"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="s2"&gt;"too small"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="s2"&gt;"too large"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="s2"&gt;"runs small"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="s2"&gt;"runs big"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="s2"&gt;"sizing chart"&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
    &lt;span class="s2"&gt;"accuracy"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="s2"&gt;"not as pictured"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="s2"&gt;"different color"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="s2"&gt;"different colour"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
                 &lt;span class="s2"&gt;"misleading"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="s2"&gt;"smaller than expected"&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
    &lt;span class="s2"&gt;"shipping"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="s2"&gt;"late"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="s2"&gt;"delayed"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="s2"&gt;"never arrived"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="s2"&gt;"tracking"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="s2"&gt;"took weeks"&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
    &lt;span class="s2"&gt;"packaging"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="s2"&gt;"crushed"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="s2"&gt;"damaged"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="s2"&gt;"dented"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="s2"&gt;"leaked"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="s2"&gt;"box was"&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
    &lt;span class="s2"&gt;"service"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="s2"&gt;"no reply"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="s2"&gt;"ignored"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="s2"&gt;"rude"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="s2"&gt;"refund"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="s2"&gt;"return"&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
    &lt;span class="s2"&gt;"value"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="s2"&gt;"overpriced"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="s2"&gt;"not worth"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="s2"&gt;"for the price"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="s2"&gt;"expensive"&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
    &lt;span class="s2"&gt;"instructions"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="s2"&gt;"instructions"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="s2"&gt;"confusing"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="s2"&gt;"no guide"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="s2"&gt;"setup"&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
    &lt;span class="s2"&gt;"missing"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="s2"&gt;"wish it had"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="s2"&gt;"would be perfect if"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="s2"&gt;"doesn't come with"&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
&lt;span class="p"&gt;}&lt;/span&gt;
&lt;span class="no"&gt;WEIGHT&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt;&lt;span class="s2"&gt;"quality"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="mi"&gt;3&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="s2"&gt;"accuracy"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="mi"&gt;3&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="s2"&gt;"sizing"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="mi"&gt;3&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="s2"&gt;"missing"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="mi"&gt;2&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
          &lt;span class="s2"&gt;"instructions"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="mi"&gt;2&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="s2"&gt;"shipping"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="mi"&gt;1&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="s2"&gt;"packaging"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="mi"&gt;1&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="s2"&gt;"service"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="mi"&gt;1&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="s2"&gt;"value"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="mi"&gt;1&lt;/span&gt;&lt;span class="p"&gt;}&lt;/span&gt;
&lt;span class="no"&gt;NEGATORS&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="s2"&gt;"not "&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="s2"&gt;"n't"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="s2"&gt;"never"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="s2"&gt;"no "&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="s2"&gt;"without"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="s2"&gt;"hardly"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="s2"&gt;"barely"&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;


&lt;span class="k"&gt;def&lt;/span&gt; &lt;span class="nf"&gt;score_review&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;text&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt;
    &lt;span class="n"&gt;lowered&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="n"&gt;text&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;lower&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;
    &lt;span class="n"&gt;hits&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt;&lt;span class="p"&gt;}&lt;/span&gt;
    &lt;span class="k"&gt;for&lt;/span&gt; &lt;span class="n"&gt;aspect&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;words&lt;/span&gt; &lt;span class="k"&gt;in&lt;/span&gt; &lt;span class="no"&gt;ASPECTS&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;items&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt;
        &lt;span class="k"&gt;for&lt;/span&gt; &lt;span class="n"&gt;w&lt;/span&gt; &lt;span class="k"&gt;in&lt;/span&gt; &lt;span class="n"&gt;words&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt;
            &lt;span class="k"&gt;for&lt;/span&gt; &lt;span class="n"&gt;m&lt;/span&gt; &lt;span class="k"&gt;in&lt;/span&gt; &lt;span class="n"&gt;re&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;finditer&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;re&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;escape&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;w&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;lowered&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt;
                &lt;span class="k"&gt;if&lt;/span&gt; &lt;span class="ow"&gt;not&lt;/span&gt; &lt;span class="nb"&gt;any&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;n&lt;/span&gt; &lt;span class="k"&gt;in&lt;/span&gt; &lt;span class="n"&gt;lowered&lt;/span&gt;&lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="nb"&gt;max&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="mi"&gt;0&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;m&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;start&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="o"&gt;-&lt;/span&gt; &lt;span class="mi"&gt;40&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt;&lt;span class="n"&gt;m&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;start&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt; &lt;span class="k"&gt;for&lt;/span&gt; &lt;span class="n"&gt;n&lt;/span&gt; &lt;span class="k"&gt;in&lt;/span&gt; &lt;span class="no"&gt;NEGATORS&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt;
                    &lt;span class="n"&gt;hits&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;setdefault&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;aspect&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;append&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;w&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;
    &lt;span class="k"&gt;return&lt;/span&gt; &lt;span class="n"&gt;hits&lt;/span&gt;


&lt;span class="k"&gt;def&lt;/span&gt; &lt;span class="nf"&gt;analyse&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;rows&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt;
    &lt;span class="n"&gt;counts&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;examples&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="nc"&gt;Counter&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="nf"&gt;defaultdict&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="kt"&gt;list&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;
    &lt;span class="k"&gt;for&lt;/span&gt; &lt;span class="n"&gt;r&lt;/span&gt; &lt;span class="k"&gt;in&lt;/span&gt; &lt;span class="n"&gt;rows&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt;
        &lt;span class="n"&gt;hits&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="nf"&gt;score_review&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;r&lt;/span&gt;&lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="s2"&gt;"text"&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;
        &lt;span class="k"&gt;for&lt;/span&gt; &lt;span class="n"&gt;aspect&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;words&lt;/span&gt; &lt;span class="k"&gt;in&lt;/span&gt; &lt;span class="n"&gt;hits&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;items&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt;
            &lt;span class="n"&gt;counts&lt;/span&gt;&lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="n"&gt;aspect&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt; &lt;span class="o"&gt;+=&lt;/span&gt; &lt;span class="mi"&gt;1&lt;/span&gt;
            &lt;span class="k"&gt;if&lt;/span&gt; &lt;span class="nb"&gt;len&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;examples&lt;/span&gt;&lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="n"&gt;aspect&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="o"&gt;&amp;lt;&lt;/span&gt; &lt;span class="mi"&gt;3&lt;/span&gt; &lt;span class="ow"&gt;and&lt;/span&gt; &lt;span class="nb"&gt;len&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;r&lt;/span&gt;&lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="s2"&gt;"text"&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="o"&gt;&amp;gt;&lt;/span&gt; &lt;span class="mi"&gt;60&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt;
                &lt;span class="n"&gt;examples&lt;/span&gt;&lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="n"&gt;aspect&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;append&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;r&lt;/span&gt;&lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="s2"&gt;"text"&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt;&lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt;&lt;span class="mi"&gt;180&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;
    &lt;span class="n"&gt;ranked&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="nb"&gt;sorted&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;counts&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;items&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;key&lt;/span&gt;&lt;span class="o"&gt;=&lt;/span&gt;&lt;span class="k"&gt;lambda&lt;/span&gt; &lt;span class="n"&gt;kv&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="o"&gt;-&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;kv&lt;/span&gt;&lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="mi"&gt;1&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt; &lt;span class="o"&gt;*&lt;/span&gt; &lt;span class="no"&gt;WEIGHT&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;get&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;kv&lt;/span&gt;&lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="mi"&gt;0&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="mi"&gt;1&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;
    &lt;span class="k"&gt;return&lt;/span&gt; &lt;span class="n"&gt;ranked&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;examples&lt;/span&gt;


&lt;span class="k"&gt;if&lt;/span&gt; &lt;span class="n"&gt;__name__&lt;/span&gt; &lt;span class="o"&gt;==&lt;/span&gt; &lt;span class="s2"&gt;"__main__"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt;
    &lt;span class="k"&gt;with&lt;/span&gt; &lt;span class="nb"&gt;open&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="s2"&gt;"reviews.csv"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;encoding&lt;/span&gt;&lt;span class="o"&gt;=&lt;/span&gt;&lt;span class="s2"&gt;"utf-8"&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="k"&gt;as&lt;/span&gt; &lt;span class="n"&gt;fh&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt;
        &lt;span class="n"&gt;rows&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="n"&gt;r&lt;/span&gt; &lt;span class="k"&gt;for&lt;/span&gt; &lt;span class="n"&gt;r&lt;/span&gt; &lt;span class="k"&gt;in&lt;/span&gt; &lt;span class="n"&gt;csv&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nc"&gt;DictReader&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;fh&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="k"&gt;if&lt;/span&gt; &lt;span class="n"&gt;r&lt;/span&gt;&lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="s2"&gt;"stars"&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt; &lt;span class="k"&gt;in&lt;/span&gt; &lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="s2"&gt;"1"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="s2"&gt;"2"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="s2"&gt;"3"&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt;
    &lt;span class="n"&gt;ranked&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;examples&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="nf"&gt;analyse&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;rows&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;
    &lt;span class="n"&gt;total&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="nb"&gt;len&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;rows&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;
    &lt;span class="k"&gt;for&lt;/span&gt; &lt;span class="n"&gt;aspect&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;n&lt;/span&gt; &lt;span class="k"&gt;in&lt;/span&gt; &lt;span class="n"&gt;ranked&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt;
        &lt;span class="k"&gt;print&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;f&lt;/span&gt;&lt;span class="s2"&gt;"{aspect:12s} {n:3d} reviews ({n / total:.0%})  weight {WEIGHT.get(aspect, 1)}"&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;
        &lt;span class="k"&gt;for&lt;/span&gt; &lt;span class="n"&gt;ex&lt;/span&gt; &lt;span class="k"&gt;in&lt;/span&gt; &lt;span class="n"&gt;examples&lt;/span&gt;&lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="n"&gt;aspect&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt;&lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt;&lt;span class="mi"&gt;2&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt;
            &lt;span class="k"&gt;print&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;f&lt;/span&gt;&lt;span class="s2"&gt;"   - {ex}"&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;


&lt;p&gt;Run against a spreadsheet of reviews with two columns (&lt;code&gt;stars&lt;/code&gt;, &lt;code&gt;text&lt;/code&gt;), it prints something&lt;br&gt;
like:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight plaintext"&gt;&lt;code&gt;quality       41 reviews (34%)  weight 3
shipping      38 reviews (32%)  weight 1
sizing        22 reviews (18%)  weight 3
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Forty-one quality complaints will not stay in the shop's hands forever. That is the list.&lt;/p&gt;

&lt;h2&gt;
  
  
  6. The failure modes worth knowing
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Small n.&lt;/strong&gt; Below about thirty reviews, one person writing a long angry review can move the
ranking on their own. Read the reviews yourself before acting.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Review gating.&lt;/strong&gt; Some shops only solicit reviews from buyers they expect to be happy. Their
negative reviews will look random and unrelated — a sign the sample is not representative.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Category-wide complaints.&lt;/strong&gt; If every competitor has 40 shipping complaints, that is your
opportunity to differentiate, not a defect of one product. Compare across shops before you
change your own listing.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Translated reviews.&lt;/strong&gt; A lexicon built from English will miss complaints written in other
languages or in heavy slang. Either extend the list or keep those reviews separate.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Overfitting to your own product.&lt;/strong&gt; The taxonomy you start with shapes what you find. Add
the "none of the above" reviews to a fifth bucket and read them once a month.&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  7. From list to listing
&lt;/h2&gt;

&lt;p&gt;The useful part of this output is not the chart. It is that each top complaint maps to a line&lt;br&gt;
of copy or a photo:&lt;/p&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Complaint&lt;/th&gt;
&lt;th&gt;What you change&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;"Runs small"&lt;/td&gt;
&lt;td&gt;a real size chart with garment measurements, not a conversion table&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;"Arrived crushed"&lt;/td&gt;
&lt;td&gt;a rigid mailer, and a photo of the packaging in the listing&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;"Not as pictured"&lt;/td&gt;
&lt;td&gt;a colour-variance note beside the photo, taken in daylight&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;"Needed a guide"&lt;/td&gt;
&lt;td&gt;a one-page instructions insert, and a link in the digital download&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;None of that requires knowing your competitor's strategy. It requires reading what their&lt;br&gt;
customers already wrote.&lt;/p&gt;

&lt;p&gt;If you would rather not maintain a lexicon, I built a small local tool in this shape:&lt;br&gt;
&lt;a href="https://8299181964131.gumroad.com/l/review-analyzer-free" rel="noopener noreferrer"&gt;the free Lite analyzer&lt;/a&gt; takes a CSV&lt;br&gt;
export or up to 30 pasted reviews, keeps the negative ones, sorts them into five complaint&lt;br&gt;
buckets and prints the counts. &lt;a href="https://8299181964131.gumroad.com/l/ecom-review-painpoint-analyzer" rel="noopener noreferrer"&gt;The full&lt;br&gt;
version&lt;/a&gt; does the same for a&lt;br&gt;
whole export and writes an Excel dashboard: the five buckets with frequency and share, a&lt;br&gt;
suggested fix for each, and every review kept row by row so the quotes are already next to the&lt;br&gt;
counts. It runs offline too.&lt;/p&gt;

&lt;p&gt;One honest note on scope: the tool ranks by how often a bucket comes up, not by the weights in&lt;br&gt;
section 4 — the weighting is the judgement call you make after reading the output. Run one&lt;br&gt;
competitor at a time and keep the sheets side by side if you are comparing shops.&lt;/p&gt;

&lt;p&gt;The script above is forty lines and free. Either way, the ranking is only useful if you go and&lt;br&gt;
change the listing — that part nobody can automate for you.&lt;/p&gt;




&lt;p&gt;&lt;em&gt;Related, from the same small store: once a complaint turns into a product change, the next&lt;br&gt;
question is whether the new version still makes money — &lt;a href="https://8299181964131.gumroad.com/l/seller-profit-tax-calculator" rel="noopener noreferrer"&gt;Seller Profit &amp;amp; Tax Calculator&lt;br&gt;
2026&lt;/a&gt; takes marketplace fees,&lt;br&gt;
payment fees, COGS, shipping and refunds out of gross sales and shows profit per unit and the&lt;br&gt;
quarterly tax reserve (&lt;a href="https://8299181964131.gumroad.com/l/seller-profit-calculator-lite" rel="noopener noreferrer"&gt;free Lite&lt;br&gt;
edition&lt;/a&gt;).&lt;/em&gt;&lt;/p&gt;

</description>
      <category>python</category>
      <category>ecommerce</category>
      <category>productivity</category>
    </item>
    <item>
      <title>Getting a bank statement into a spreadsheet without losing a cent</title>
      <dc:creator>Seller Tools</dc:creator>
      <pubDate>Mon, 16 Nov 2026 21:06:45 +0000</pubDate>
      <link>https://dev.to/yougkxm/getting-a-bank-statement-into-a-spreadsheet-without-losing-a-cent-n5b</link>
      <guid>https://dev.to/yougkxm/getting-a-bank-statement-into-a-spreadsheet-without-losing-a-cent-n5b</guid>
      <description>&lt;h1&gt;
  
  
  Getting a bank statement into a spreadsheet without losing a cent
&lt;/h1&gt;




&lt;p&gt;Every month, thousands of small-business owners do the same thing: open a bank statement PDF,&lt;br&gt;
open a spreadsheet, and start typing. Sixty transactions means about three hundred typed&lt;br&gt;
values. Human entry error runs somewhere around 1% per value in careful conditions, and nobody&lt;br&gt;
is careful at 11pm on a Sunday.&lt;/p&gt;

&lt;p&gt;The good news is that most bank statement PDFs are not images. They contain a real text layer,&lt;br&gt;
which means you do not need OCR, a subscription, or a cloud upload to get the rows out. You&lt;br&gt;
need to understand how the text is positioned on the page.&lt;/p&gt;

&lt;p&gt;This post is the practical version of that: how the file is structured, how to pull the rows&lt;br&gt;
out with Python, how to know whether the result is right, and where the approach stops working.&lt;/p&gt;
&lt;h2&gt;
  
  
  1. A PDF is not a table, it is a pile of labels on a canvas
&lt;/h2&gt;

&lt;p&gt;A PDF does not know what a row is. It knows that at coordinate (312.4, 688.1) there is the&lt;br&gt;
string &lt;code&gt;-64.28&lt;/code&gt;, and at (72.0, 688.1) there is &lt;code&gt;2026-10-03&lt;/code&gt;. Everything else — which of those&lt;br&gt;
belongs to which column, whether a line is a transaction or the "Page 2 of 4" footer — is&lt;br&gt;
something a human eyeball infers.&lt;/p&gt;

&lt;p&gt;That is why naive extraction fails:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;code&gt;pdfplumber&lt;/code&gt;'s &lt;code&gt;.extract_table()&lt;/code&gt; uses ruling lines. Many statements have no vertical lines
between columns, so it returns one wide column or nothing.&lt;/li&gt;
&lt;li&gt;
&lt;code&gt;.extract_text()&lt;/code&gt; returns lines in reading order, but a statement page often has a header
block, then the transaction table, then a summary block, and the reading order can interleave
them.&lt;/li&gt;
&lt;li&gt;Debit and credit are frequently two separate columns with only one of them filled per row.
If you concatenate them you get empty cells in the middle of otherwise clean data.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;The reliable approach is &lt;strong&gt;positional clustering&lt;/strong&gt;: take every word with its bounding box,&lt;br&gt;
group the words by their vertical centre into lines, then assign each word to a column based on&lt;br&gt;
its x position.&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight python"&gt;&lt;code&gt;&lt;span class="k"&gt;import&lt;/span&gt; &lt;span class="n"&gt;pdfplumber&lt;/span&gt;
&lt;span class="k"&gt;from&lt;/span&gt; &lt;span class="n"&gt;collections&lt;/span&gt; &lt;span class="k"&gt;import&lt;/span&gt; &lt;span class="n"&gt;defaultdict&lt;/span&gt;

&lt;span class="k"&gt;def&lt;/span&gt; &lt;span class="nf"&gt;page_rows&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;page&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;tolerance&lt;/span&gt;&lt;span class="o"&gt;=&lt;/span&gt;&lt;span class="mf"&gt;3.0&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt;
    &lt;span class="n"&gt;words&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="n"&gt;page&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;extract_words&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;use_text_flow&lt;/span&gt;&lt;span class="o"&gt;=&lt;/span&gt;&lt;span class="kc"&gt;False&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;keep_blank_chars&lt;/span&gt;&lt;span class="o"&gt;=&lt;/span&gt;&lt;span class="kc"&gt;False&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;
    &lt;span class="n"&gt;lines&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="nf"&gt;defaultdict&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="kt"&gt;list&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;
    &lt;span class="k"&gt;for&lt;/span&gt; &lt;span class="n"&gt;w&lt;/span&gt; &lt;span class="k"&gt;in&lt;/span&gt; &lt;span class="n"&gt;words&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt;
        &lt;span class="n"&gt;key&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="nb"&gt;round&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;w&lt;/span&gt;&lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="s2"&gt;"top"&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt; &lt;span class="o"&gt;/&lt;/span&gt; &lt;span class="n"&gt;tolerance&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;
        &lt;span class="n"&gt;lines&lt;/span&gt;&lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="n"&gt;key&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;append&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;w&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;
    &lt;span class="k"&gt;for&lt;/span&gt; &lt;span class="n"&gt;key&lt;/span&gt; &lt;span class="k"&gt;in&lt;/span&gt; &lt;span class="nb"&gt;sorted&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;lines&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt;
        &lt;span class="k"&gt;yield&lt;/span&gt; &lt;span class="nb"&gt;sorted&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;lines&lt;/span&gt;&lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="n"&gt;key&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;key&lt;/span&gt;&lt;span class="o"&gt;=&lt;/span&gt;&lt;span class="k"&gt;lambda&lt;/span&gt; &lt;span class="n"&gt;w&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="n"&gt;w&lt;/span&gt;&lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="s2"&gt;"x0"&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Three words that sit within three points of each other vertically are on the same visual line.&lt;br&gt;
That single heuristic handles 90% of statement layouts, and it fails loudly rather than&lt;br&gt;
silently: when a row is broken, you see it as a row with two cells instead of four.&lt;/p&gt;
&lt;h2&gt;
  
  
  2. Find the column anchors once, not per row
&lt;/h2&gt;

&lt;p&gt;Do not hard-code x coordinates per file — statements get re-issued with slightly different&lt;br&gt;
margins. Instead, detect the header line (&lt;code&gt;Date&lt;/code&gt;, &lt;code&gt;Description&lt;/code&gt;, &lt;code&gt;Amount&lt;/code&gt;, &lt;code&gt;Balance&lt;/code&gt;, or a&lt;br&gt;
local equivalent), take the x position of each header word, and use the midpoints between them&lt;br&gt;
as column boundaries for the rest of the document.&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight python"&gt;&lt;code&gt;&lt;span class="k"&gt;def&lt;/span&gt; &lt;span class="nf"&gt;column_bounds&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;header_words&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt;
    &lt;span class="n"&gt;xs&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="n"&gt;w&lt;/span&gt;&lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="s2"&gt;"x0"&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt; &lt;span class="k"&gt;for&lt;/span&gt; &lt;span class="n"&gt;w&lt;/span&gt; &lt;span class="k"&gt;in&lt;/span&gt; &lt;span class="n"&gt;header_words&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt;
    &lt;span class="n"&gt;xs&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;sort&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;
    &lt;span class="k"&gt;return&lt;/span&gt; &lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;a&lt;/span&gt; &lt;span class="o"&gt;+&lt;/span&gt; &lt;span class="n"&gt;b&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="o"&gt;/&lt;/span&gt; &lt;span class="mi"&gt;2&lt;/span&gt; &lt;span class="k"&gt;for&lt;/span&gt; &lt;span class="n"&gt;a&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;b&lt;/span&gt; &lt;span class="k"&gt;in&lt;/span&gt; &lt;span class="nb"&gt;zip&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;xs&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;xs&lt;/span&gt;&lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="mi"&gt;1&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt;

&lt;span class="k"&gt;def&lt;/span&gt; &lt;span class="nf"&gt;cell_for&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;word&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;bounds&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt;
    &lt;span class="k"&gt;for&lt;/span&gt; &lt;span class="n"&gt;i&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;edge&lt;/span&gt; &lt;span class="k"&gt;in&lt;/span&gt; &lt;span class="nb"&gt;enumerate&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;bounds&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt;
        &lt;span class="k"&gt;if&lt;/span&gt; &lt;span class="n"&gt;word&lt;/span&gt;&lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="s2"&gt;"x0"&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt; &lt;span class="o"&gt;&amp;lt;&lt;/span&gt; &lt;span class="n"&gt;edge&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt;
            &lt;span class="k"&gt;return&lt;/span&gt; &lt;span class="n"&gt;i&lt;/span&gt;
    &lt;span class="k"&gt;return&lt;/span&gt; &lt;span class="nb"&gt;len&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;bounds&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;If your bank prints &lt;code&gt;Debit&lt;/code&gt; and &lt;code&gt;Credit&lt;/code&gt; as separate columns, keep them separate in the output&lt;br&gt;
too. Collapsing them into one signed amount is a decision, and it should be made after you see&lt;br&gt;
what is actually in each column — not before.&lt;/p&gt;
&lt;h2&gt;
  
  
  3. The sign problem is the expensive one
&lt;/h2&gt;

&lt;p&gt;Statements express money out in at least four ways:&lt;/p&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Form&lt;/th&gt;
&lt;th&gt;What you must not do&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;&lt;code&gt;-64.28&lt;/code&gt;&lt;/td&gt;
&lt;td&gt;keep it&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;code&gt;(64.28)&lt;/code&gt;&lt;/td&gt;
&lt;td&gt;parentheses mean negative, not "note"&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;code&gt;64.28 DR&lt;/code&gt;&lt;/td&gt;
&lt;td&gt;trailing &lt;code&gt;DR&lt;/code&gt; means money out&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;
&lt;code&gt;64.28&lt;/code&gt; in a Debit column, blank in Credit&lt;/td&gt;
&lt;td&gt;the column decides, not the number&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;A converter that drops the &lt;code&gt;DR&lt;/code&gt; suffix, or that treats &lt;code&gt;(64.28)&lt;/code&gt; as 64.28, does not produce a&lt;br&gt;
slightly wrong spreadsheet. It produces one where every refund looks like a payment, and the&lt;br&gt;
closing balance never reconciles. If your output has a &lt;code&gt;-&lt;/code&gt; where the source had &lt;code&gt;(&lt;/code&gt; or &lt;code&gt;DR&lt;/code&gt;,&lt;br&gt;
that is the first thing to test.&lt;/p&gt;

&lt;p&gt;Dates have the same class of problem. &lt;code&gt;07/05/2026&lt;/code&gt; is 5 July on a US statement and 7 May on a&lt;br&gt;
UK one. Your statement's own header usually tells you; if it does not, look at the statement&lt;br&gt;
period on page one — a statement covering 1 April to 30 April cannot contain a 5 July&lt;br&gt;
transaction.&lt;/p&gt;
&lt;h2&gt;
  
  
  4. The only validation that matters
&lt;/h2&gt;

&lt;p&gt;Row count is a weak check: two errors that cancel out keep the count the same. The check that&lt;br&gt;
catches nearly everything is the printed balance:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight plaintext"&gt;&lt;code&gt;opening balance + sum(money in) - sum(money out) = closing balance
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;If that identity holds to the cent on a statement with 200 transactions, the signs, the&lt;br&gt;
decimal separators and the page breaks were all read correctly. If it is off by 0.01, you have&lt;br&gt;
found a rounding problem worth understanding before you hand the file to an accountant. This is&lt;br&gt;
also the reason to keep the &lt;strong&gt;running balance column&lt;/strong&gt; even if you do not need it: it is a&lt;br&gt;
per-row checksum. When the running balance stops matching after row 57, you know exactly which&lt;br&gt;
row to look at.&lt;/p&gt;

&lt;h2&gt;
  
  
  5. Where this approach stops
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Scanned statements.&lt;/strong&gt; No text layer, no positional clustering. You are in OCR territory,
and OCR misreads digits in a way that positional extraction does not: &lt;code&gt;8&lt;/code&gt; for &lt;code&gt;3&lt;/code&gt;, &lt;code&gt;0&lt;/code&gt; for
&lt;code&gt;6&lt;/code&gt;. Never trust an OCR result that has not passed the balance check.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Multi-line descriptions.&lt;/strong&gt; A description wrapped onto a second line has no date and no
amount. Decide the rule up front: attach it to the row above, or emit it as a continuation
row. Silently dropping it loses the merchant name, which is the one field people actually
read.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Page breaks.&lt;/strong&gt; The transaction list runs across pages and the first row of page 2 often has
no repeated header. If you only look for the header once per page, you lose that row.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Statements with sub-tables.&lt;/strong&gt; Some business accounts print a "pending" block above the
transaction list. Both blocks look identical to a row-grouping function; only the section
heading distinguishes them.&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  6. A complete, small implementation
&lt;/h2&gt;



&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight python"&gt;&lt;code&gt;&lt;span class="k"&gt;import&lt;/span&gt; &lt;span class="n"&gt;csv&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;re&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;sys&lt;/span&gt;
&lt;span class="k"&gt;import&lt;/span&gt; &lt;span class="n"&gt;pdfplumber&lt;/span&gt;

&lt;span class="no"&gt;NUM&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="n"&gt;re&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nb"&gt;compile&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;r&lt;/span&gt;&lt;span class="s2"&gt;"^\(?-?[\d,]+\.\d{2}\)?(?:\s?(?:DR|CR))?$"&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;

&lt;span class="k"&gt;def&lt;/span&gt; &lt;span class="nf"&gt;to_amount&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;text&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt;
    &lt;span class="n"&gt;neg&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="n"&gt;text&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;strip&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;startswith&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="s2"&gt;"("&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="ow"&gt;or&lt;/span&gt; &lt;span class="n"&gt;text&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;strip&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;endswith&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="s2"&gt;"DR"&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;
    &lt;span class="n"&gt;value&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="kt"&gt;float&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;re&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;sub&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;r&lt;/span&gt;&lt;span class="s2"&gt;"[^\d.]"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="s2"&gt;""&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;text&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;
    &lt;span class="k"&gt;return&lt;/span&gt; &lt;span class="o"&gt;-&lt;/span&gt;&lt;span class="n"&gt;value&lt;/span&gt; &lt;span class="k"&gt;if&lt;/span&gt; &lt;span class="n"&gt;neg&lt;/span&gt; &lt;span class="k"&gt;else&lt;/span&gt; &lt;span class="n"&gt;value&lt;/span&gt;

&lt;span class="k"&gt;def&lt;/span&gt; &lt;span class="nf"&gt;convert&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;pdf_path&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;out_csv&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt;
    &lt;span class="n"&gt;rows&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt;
    &lt;span class="k"&gt;with&lt;/span&gt; &lt;span class="n"&gt;pdfplumber&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nb"&gt;open&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;pdf_path&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="k"&gt;as&lt;/span&gt; &lt;span class="n"&gt;pdf&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt;
        &lt;span class="k"&gt;for&lt;/span&gt; &lt;span class="n"&gt;page&lt;/span&gt; &lt;span class="k"&gt;in&lt;/span&gt; &lt;span class="n"&gt;pdf&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="n"&gt;pages&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt;
            &lt;span class="k"&gt;for&lt;/span&gt; &lt;span class="n"&gt;line&lt;/span&gt; &lt;span class="k"&gt;in&lt;/span&gt; &lt;span class="nf"&gt;page_rows&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;page&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt;
                &lt;span class="n"&gt;cells&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="n"&gt;w&lt;/span&gt;&lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="s2"&gt;"text"&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt; &lt;span class="k"&gt;for&lt;/span&gt; &lt;span class="n"&gt;w&lt;/span&gt; &lt;span class="k"&gt;in&lt;/span&gt; &lt;span class="n"&gt;line&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt;
                &lt;span class="k"&gt;if&lt;/span&gt; &lt;span class="nb"&gt;len&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;cells&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="o"&gt;&amp;lt;&lt;/span&gt; &lt;span class="mi"&gt;3&lt;/span&gt; &lt;span class="ow"&gt;or&lt;/span&gt; &lt;span class="ow"&gt;not&lt;/span&gt; &lt;span class="n"&gt;re&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="k"&gt;match&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;r&lt;/span&gt;&lt;span class="s2"&gt;"\d{1,2}[/-]\d{1,2}[/-]\d{2,4}"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;cells&lt;/span&gt;&lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="mi"&gt;0&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt;
                    &lt;span class="k"&gt;continue&lt;/span&gt;
                &lt;span class="n"&gt;amounts&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="n"&gt;c&lt;/span&gt; &lt;span class="k"&gt;for&lt;/span&gt; &lt;span class="n"&gt;c&lt;/span&gt; &lt;span class="k"&gt;in&lt;/span&gt; &lt;span class="n"&gt;cells&lt;/span&gt; &lt;span class="k"&gt;if&lt;/span&gt; &lt;span class="no"&gt;NUM&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="k"&gt;match&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;c&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt;
                &lt;span class="k"&gt;if&lt;/span&gt; &lt;span class="ow"&gt;not&lt;/span&gt; &lt;span class="n"&gt;amounts&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt;
                    &lt;span class="k"&gt;continue&lt;/span&gt;
                &lt;span class="n"&gt;rows&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;append&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="p"&gt;{&lt;/span&gt;
                    &lt;span class="s2"&gt;"date"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="n"&gt;cells&lt;/span&gt;&lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="mi"&gt;0&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
                    &lt;span class="s2"&gt;"description"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="s2"&gt;" "&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;join&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;
                        &lt;span class="n"&gt;c&lt;/span&gt; &lt;span class="k"&gt;for&lt;/span&gt; &lt;span class="n"&gt;c&lt;/span&gt; &lt;span class="k"&gt;in&lt;/span&gt; &lt;span class="n"&gt;cells&lt;/span&gt;&lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="mi"&gt;1&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt; &lt;span class="k"&gt;if&lt;/span&gt; &lt;span class="ow"&gt;not&lt;/span&gt; &lt;span class="no"&gt;NUM&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="k"&gt;match&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;c&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
                    &lt;span class="s2"&gt;"amount"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="nf"&gt;to_amount&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;amounts&lt;/span&gt;&lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="o"&gt;-&lt;/span&gt;&lt;span class="mi"&gt;1&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
                &lt;span class="p"&gt;}&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;
    &lt;span class="k"&gt;with&lt;/span&gt; &lt;span class="nb"&gt;open&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;out_csv&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="s2"&gt;"w"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;newline&lt;/span&gt;&lt;span class="o"&gt;=&lt;/span&gt;&lt;span class="s2"&gt;""&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;encoding&lt;/span&gt;&lt;span class="o"&gt;=&lt;/span&gt;&lt;span class="s2"&gt;"utf-8"&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="k"&gt;as&lt;/span&gt; &lt;span class="n"&gt;fh&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt;
        &lt;span class="n"&gt;w&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="n"&gt;csv&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nc"&gt;DictWriter&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;fh&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;fieldnames&lt;/span&gt;&lt;span class="o"&gt;=&lt;/span&gt;&lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="s2"&gt;"date"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="s2"&gt;"description"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="s2"&gt;"amount"&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;
        &lt;span class="n"&gt;w&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;writeheader&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;
        &lt;span class="n"&gt;w&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;writerows&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;rows&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;
    &lt;span class="k"&gt;print&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;f&lt;/span&gt;&lt;span class="s2"&gt;"{len(rows)} rows -&amp;gt; {out_csv}"&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;It is about forty lines, it runs locally, and it handles the common case: a text-layer PDF with&lt;br&gt;
one amount column. It will not handle every bank, and it does not attempt the balance check —&lt;br&gt;
add that before you trust it.&lt;/p&gt;

&lt;h2&gt;
  
  
  7. What a finished tool adds
&lt;/h2&gt;

&lt;p&gt;If you would rather not maintain the parser, the two things worth paying for are the ones you&lt;br&gt;
cannot see from a demo: handling the layouts that break, and the balance check that tells you&lt;br&gt;
when something was read wrong. I built a small converter in that shape —&lt;br&gt;
&lt;a href="https://8299181964131.gumroad.com/l/bank-converter-free" rel="noopener noreferrer"&gt;the free Lite edition&lt;/a&gt; converts one&lt;br&gt;
PDF of up to 20 rows so you can test it against a statement you already know the total of, and&lt;br&gt;
&lt;a href="https://8299181964131.gumroad.com/l/bank-statement-to-excel" rel="noopener noreferrer"&gt;the full version&lt;/a&gt; does batches and&lt;br&gt;
keeps the running balance column. Both run on your machine; neither uploads the file.&lt;/p&gt;

&lt;p&gt;Whichever route you take — forty lines of Python or a tool — the rule is the same: &lt;strong&gt;do not&lt;br&gt;
trust the spreadsheet until the closing balance matches.&lt;/strong&gt; Everything else is formatting.&lt;/p&gt;




&lt;p&gt;&lt;em&gt;Related, from the same small store: if the converted statement is going into a seller's books&lt;br&gt;
rather than an accountant's inbox, &lt;a href="https://8299181964131.gumroad.com/l/seller-profit-tax-calculator" rel="noopener noreferrer"&gt;Seller Profit &amp;amp; Tax Calculator&lt;br&gt;
2026&lt;/a&gt; is the next step — one&lt;br&gt;
row per order, fees and refunds taken out, then profit per sale and the quarterly tax reserve&lt;br&gt;
(&lt;a href="https://8299181964131.gumroad.com/l/seller-profit-calculator-lite" rel="noopener noreferrer"&gt;free Lite edition&lt;/a&gt;).&lt;/em&gt;&lt;/p&gt;

</description>
      <category>python</category>
      <category>excel</category>
      <category>finance</category>
      <category>productivity</category>
    </item>
  </channel>
</rss>
