Before fixing a zero-view marketplace shop, measure the base rate: an AI agent pulled 410 competing listings through the public API

A new digital product shop gets almost no views, and the founder is tempted to keep tweaking titles and tags or to build more products without knowing whether the silence is unusual.

1. Check the obvious setup errors first: the agent found the shop was registered in English while the listings were written in French, so the owner added French as a shop language and the agent translated the listings. 2. Write down a prediction with a confidence level for each fix (for example at least 30 lifetime views by a set date, 40 percent confidence) and check it later. 3. Measure the base rate: using the marketplace's public API, the agent pulled 410 listings from 224 shops across six search phrases and looked at listings 8 to 60 days old (median 3 lifetime views), then checked 60 shops (37 had never sold, and sales were concentrated in three established shops). 4. Put shop rules in the shop's own playbook: the agent broke its own focus-on-visibility rule because that rule lived in a different file. Publishing new listings became the owner's decision; the agent builds a draft through the API and the owner publishes. 5. In a parallel freelance channel, skip jobs that forbid AI tools and watch the job feed every 15 minutes. 6. Have a separate reviewer agent check every deliverable before publication; it found real problems in all eight demo apps, including a payment webhook that could be replayed.

After two weeks: 37.18 euro spent of 200, 0 euro revenue, 5 listings live with 12 lifetime views and 0 sales, 10 freelance proposals sent with none viewed. The base-rate study suggested near-zero views is normal for new listings in that niche. A prediction that title and tag fixes would bring 10 views in a week failed (1 view). The shop-language prediction was still pending when the post was written.

Written by the AI agent itself (Claude in Claude Code) with a human owner, two weeks in, with no sales yet. Numbers are self-reported and the API sample and scripts are not published. The marketplace (Etsy is named in the numbers) and the niche (French templates for micro-entrepreneurs) are specific. The agent notes that its AI-assistance disclosure was lost on the English listings and was silently removed by the blog editor twice. The business sells the templates described.

A marketplace with a public API, a way to run scripts, and a product listing to compare.

• DEV Community (Clair Et Net) ↗ Firsthand write-up, published October 31, 2026

Source published: October 31, 2026. Last reviewed here: November 1, 2026. Spot a mistake? Tell us.