<?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: ef 32d</title>
    <description>The latest articles on DEV Community by ef 32d (@gp_sourcing).</description>
    <link>https://dev.to/gp_sourcing</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%2F4253071%2F4c8f2446-9699-44f5-ac7d-d8b605f688ae.png</url>
      <title>DEV Community: ef 32d</title>
      <link>https://dev.to/gp_sourcing</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://dev.to/feed/gp_sourcing"/>
    <language>en</language>
    <item>
      <title>Your Shopify pixel is underreporting purchases: the server-side fixes that actually matter</title>
      <dc:creator>ef 32d</dc:creator>
      <pubDate>Thu, 19 Nov 2026 06:08:02 +0000</pubDate>
      <link>https://dev.to/gp_sourcing/your-shopify-pixel-is-underreporting-purchases-the-server-side-fixes-that-actually-matter-2ndk</link>
      <guid>https://dev.to/gp_sourcing/your-shopify-pixel-is-underreporting-purchases-the-server-side-fixes-that-actually-matter-2ndk</guid>
      <description>&lt;p&gt;If you run paid traffic to a Shopify store in 2026, the browser pixel is lying to you. Not maliciously — it just never sees the conversion. Safari strips the query string after 7 days, ATT means half your iOS traffic never fires the pixel at all, and roughly a third of desktop users run an ad blocker. The platform then reports fewer conversions than you actually had, and you cut budget on campaigns that were working.&lt;/p&gt;

&lt;p&gt;Server-side tracking fixes the observation problem. Here is the part most setup guides skip.&lt;/p&gt;

&lt;h2&gt;
  
  
  What actually happens on a purchase
&lt;/h2&gt;

&lt;p&gt;Browser-side, the flow is: page loads → pixel fires → request goes from the &lt;em&gt;customer's&lt;/em&gt; browser to Meta. That request is a third-party request from &lt;code&gt;connect.facebook.net&lt;/code&gt;, which is exactly what blockers and ITP target.&lt;/p&gt;

&lt;p&gt;Server-side, the flow is: order created on Shopify → your server assembles the payload → your server calls the Conversions API directly. The request originates from your IP, not the customer's, and carries no third-party cookie dependency. Match quality goes up because you control what identifiers you send.&lt;/p&gt;

&lt;h2&gt;
  
  
  The identifier ladder
&lt;/h2&gt;

&lt;p&gt;Meta's matching is greedy in a specific order. Send as much as you legally can:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight plaintext"&gt;&lt;code&gt;em  → SHA-256 of lowercased, trimmed email
ph  → SHA-256 of E.164 phone, digits only, no +
fn  → SHA-256 of lowercased first name
ln  → SHA-256 of lowercased last name
ct  → SHA-256 of lowercased city, no spaces
st  → SHA-256 of 2-letter state code
zp  → SHA-256 of 5-digit zip
country → 2-letter ISO, NOT hashed
external_id → your customer id, NOT hashed
fbp / fbc → raw cookie values, NOT hashed
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Two mistakes I see constantly:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;Hashing &lt;code&gt;country&lt;/code&gt; or &lt;code&gt;external_id&lt;/code&gt;. Meta expects those in plaintext and silently drops the hashed version.&lt;/li&gt;
&lt;li&gt;Hashing the plus sign in a phone number. &lt;code&gt;+1 415 555 0100&lt;/code&gt; must become &lt;code&gt;14155550100&lt;/code&gt; before hashing, not &lt;code&gt;+14155550100&lt;/code&gt;.&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  Deduplication is the whole game
&lt;/h2&gt;

&lt;p&gt;If both the browser pixel and the server fire the Purchase event without a shared key, Meta counts two purchases. Your ROAS looks great and your actual revenue does not change.&lt;/p&gt;

&lt;p&gt;The fix is a shared &lt;code&gt;event_id&lt;/code&gt;:&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="c1"&gt;// browser&lt;/span&gt;
&lt;span class="nf"&gt;fbq&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="s1"&gt;'track'&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="s1"&gt;'Purchase'&lt;/span&gt;&lt;span class="p"&gt;,&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="mf"&gt;89.00&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;currency&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="s1"&gt;'USD'&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;eventID&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="s1"&gt;'order_10231'&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="c1"&gt;// server, same order&lt;/span&gt;
&lt;span class="p"&gt;{&lt;/span&gt;
  &lt;span class="s2"&gt;"event_name"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="s2"&gt;"Purchase"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
  &lt;span class="s2"&gt;"event_id"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="s2"&gt;"order_10231"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
  &lt;span class="s2"&gt;"event_time"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="mi"&gt;1763500000&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
  &lt;span class="s2"&gt;"action_source"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="s2"&gt;"website"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
  &lt;span class="s2"&gt;"custom_data"&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="mf"&gt;89.00&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="s2"&gt;"currency"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="s2"&gt;"USD"&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;Meta dedupes on &lt;code&gt;event_name&lt;/code&gt; + &lt;code&gt;event_id&lt;/code&gt; within a 48-hour window and keeps the first one it receives. Use the Shopify order id — it is stable, unique, and available on both sides.&lt;/p&gt;

&lt;h2&gt;
  
  
  First-party pixel, and why it matters more than people think
&lt;/h2&gt;

&lt;p&gt;Serving &lt;code&gt;https://yourdomain.com/…&lt;/code&gt; instead of &lt;code&gt;https://connect.facebook.net/…&lt;/code&gt; puts the request in first-party context. Blocklists key on third-party hostnames, so a script served from your own domain slips past most of them. The tradeoff: you now proxy the request, and you must keep the proxy path stable or you break in-flight attribution.&lt;/p&gt;

&lt;h2&gt;
  
  
  The thing that surprised me
&lt;/h2&gt;

&lt;p&gt;After moving to server-side, the reported conversion count went &lt;em&gt;up&lt;/em&gt;, but so did reported CPA — because Meta was finally seeing the ~30% of purchases it had been blind to, including a long tail of low-value orders from prospecting audiences. The campaigns had not gotten worse. The measurement had gotten honest, and the honest number was less flattering than the pixel-only number.&lt;/p&gt;

&lt;p&gt;That is the real reason to do this: not to make reports look better, but to stop optimizing against a number that is systematically wrong in one direction.&lt;/p&gt;

&lt;p&gt;If you want the prebuilt version for Shopify, I maintain &lt;a href="https://www.gottracked.io" rel="noopener noreferrer"&gt;GotTracked&lt;/a&gt; — server-side events to Meta CAPI, TikTok Events API and Google Ads, first-party pixel on your own domain, UTM capture, and ad spend reconciled against order-level revenue.&lt;/p&gt;

</description>
      <category>analytics</category>
      <category>marketing</category>
      <category>privacy</category>
    </item>
    <item>
      <title>Reading 1688 product pages without Chinese: what actually changes your landed cost</title>
      <dc:creator>ef 32d</dc:creator>
      <pubDate>Thu, 19 Nov 2026 06:06:58 +0000</pubDate>
      <link>https://dev.to/gp_sourcing/reading-1688-product-pages-without-chinese-what-actually-changes-your-landed-cost-152d</link>
      <guid>https://dev.to/gp_sourcing/reading-1688-product-pages-without-chinese-what-actually-changes-your-landed-cost-152d</guid>
      <description>&lt;p&gt;If you buy from 1688 and don't read Chinese, every page is a wall of text. Product titles,规格, MOQ, and shipping terms are all packed into one long block, and Chrome's built-in translate turns it into something roughly readable but useless for decisions.&lt;/p&gt;

&lt;p&gt;Here is how I approach it now.&lt;/p&gt;

&lt;h2&gt;
  
  
  The three things that actually matter on a 1688 product page
&lt;/h2&gt;

&lt;p&gt;Out of everything on that page, only three pieces change the landed cost:&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;
&lt;strong&gt;MOQ and price tiers&lt;/strong&gt; — the headline price is usually for 1-2 pieces. At 500 pieces the unit price can be 40% lower. The tier table is a grid of Chinese numbers and it is easy to misread which column applies to which quantity.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Whether the listing is the factory or a trading company&lt;/strong&gt; — 1688 is full of 贸易公司 reselling the same photo. Factory listings usually have workshop photos, a 厂房 area listed, and a 实力 section with production equipment. Traders have none of that.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;The shipping basis&lt;/strong&gt; — EXW / FOB / DDP. Most listings quote EXW, which means the price you see excludes everything until the goods reach your door.&lt;/li&gt;
&lt;/ol&gt;

&lt;h2&gt;
  
  
  Why machine translation fails on product titles
&lt;/h2&gt;

&lt;p&gt;A 1688 title is a keyword stack, not a sentence:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight plaintext"&gt;&lt;code&gt;跨境专供 新款不锈钢保温杯 便携双层真空 定制logo 礼品杯
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;A translator renders that as "cross-border special supply new stainless steel thermos portable double-layer vacuum custom logo gift cup" — technically correct, semantically flat. What it hides is that 跨境专供 means this seller already handles export documentation, which is the single biggest time saver if you're shipping to the US or EU.&lt;/p&gt;

&lt;p&gt;So I stopped reading translations as prose and started reading them as field-value pairs. 专供, 定制, 现货, 起订量 each carry a specific commercial meaning, and once you map maybe 30 of them the pages become readable.&lt;/p&gt;

&lt;h2&gt;
  
  
  Computing DDP before you order
&lt;/h2&gt;

&lt;p&gt;The number people get wrong is landed cost. A quick model:&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="n"&gt;landed_unit&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;fob_unit&lt;/span&gt; &lt;span class="o"&gt;*&lt;/span&gt; &lt;span class="n"&gt;qty&lt;/span&gt; &lt;span class="o"&gt;+&lt;/span&gt; &lt;span class="n"&gt;freight&lt;/span&gt; &lt;span class="o"&gt;+&lt;/span&gt; &lt;span class="n"&gt;duty&lt;/span&gt; &lt;span class="o"&gt;+&lt;/span&gt; &lt;span class="n"&gt;last_mile&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="o"&gt;/&lt;/span&gt; &lt;span class="n"&gt;qty&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Where freight for small parcels is usually volumetric, not actual weight:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight plaintext"&gt;&lt;code&gt;volumetric_kg = L(cm) * W(cm) * H(cm) / 6000
billable_kg   = max(actual_kg, volumetric_kg)
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;A 30x20x15cm box at 1.2kg actual weighs 1.5kg billable. On a 500-piece order that difference compounds into real money, and it is the reason quotes from two forwarders for the "same" shipment differ by 30%.&lt;/p&gt;

&lt;h2&gt;
  
  
  What I automated
&lt;/h2&gt;

&lt;p&gt;I got tired of doing this in three tabs, so I built a Chrome extension that does the mechanical part: translates the product page in place, runs an image search against the same photo to surface factory listings with matching tooling, and computes DDP landed cost plus margin at whatever quantity you type in. It is free and does not require an account.&lt;/p&gt;

&lt;p&gt;If you want to try it: &lt;a href="https://gpsourcing.gp-fulfillment.com" rel="noopener noreferrer"&gt;GP Sourcing for 1688&lt;/a&gt; — Chrome Web Store listing is on the site.&lt;/p&gt;

&lt;p&gt;The part I have not automated, and probably never will, is judging whether a supplier is reliable. That still takes a video call and a sample order.&lt;/p&gt;

</description>
      <category>product</category>
      <category>productivity</category>
      <category>startup</category>
    </item>
  </channel>
</rss>
