Three sales pages went up in eight days: Harbor Counsel on the 9th, Northmark on the 12th, Summit Legal on the 16th. All three are Next.js homepages for small professional firms, all three cost $29, and all three end the same way — "Email us to order. Write 'T3' in the message." No cart, no account, no Stripe keys, no webhook handler.
That last part is the decision people push back on, so here is the trade rather than the preference.
A cart would have bought me a week of work and no information
The obvious path was Stripe Checkout plus a signed download link. Realistically that is a week: tax handling for buyers in countries I cannot name, refunds, delivery for a 40MB ZIP, and an email that still has to be sent by hand. Every one of those pieces is a failure mode I would be debugging at 2am instead of looking at whether the page persuades anyone.
What I get instead is an invoice and a reply. The reply tells me what the buyer is building and whether they wanted the site built for them. That is the product I actually sell; the $29 file is the sample that proves it is real.
The cost is real and I am paying it: anyone who will not write an email at 1am is gone. I decided that was acceptable while the pages are days old and unproven. It stops being acceptable the moment the email step is the reason a sale dies, and I will not know that from a dashboard.
$29 prices a sample, not a website
Pricing a template at the cost of two coffees is not generosity, it is positioning. The money in this business is in building and maintaining the site afterwards, so the template has one job: convince a stranger in ninety seconds that the code behind it is finished. Low price plus high finish is the cheapest credibility I can buy.
The risk is equally plain: a low price attracts people who want a $29 custom site. The pages filter them by stating what the ZIP is not — no CMS, no database, no subscription, content edited in one JSON file.
What I deleted so three pages would not be one page
All three share five sections and one data file, and differ in the hero: a word-block with a highlighted phrase, an editorial full-bleed photograph, a stat band over a skyline with a scrolling outcomes strip. Everything else was subtraction — no logo carousel, no badge wall, no twenty-item services menu. Four lanes in one, three practice cards in another, three numbered process steps in each.
I also paused the template station that was supposed to produce the next batch, on the same day I shipped the third page. Eight pages sharing one skeleton is a colour scheme, not a catalogue.
The parts that could have been lies, and are labelled instead
The strongest page opens with counters — attorneys, years, intake hours — and a ticker of verdicts and settlements. Those figures are sample copy, and the page says so in the eyebrow, because advertising outcomes is regulated in most jurisdictions a buyer would deploy into. The ticker also stops for visitors who ask for reduced motion.
That costs me the most persuasive screenshot in the set. The alternative was a page that looks better and puts a buyer in front of a regulator.
The only step I did not hand off
Before anything ships I run the page through the detector I wrote at prismicaudit.com, mostly to catch the boring failures — a missing canonical, a description that does not match the page. It is my own tool, so I treat its output as a checklist, not a verdict.
The bet, stated so it can be wrong: a $29 ZIP that runs on npm install && npm run dev with no CMS account is the cheapest proof that the build work behind it is worth paying for. Testing it cost three pages and eight days. The catalogue is at slicelab.prismicaudit.com, where each page still ends with an email address instead of a button.
Top comments (0)