Testing Demand with a Small Batch: How to Avoid Dead Stock

Spreadsheet math does not sell — the shelf does. A small-batch test is the last filter before a serious purchase: it turns the hypothesis "this product will sell" into hard numbers for conversion and buyout rate. Here is how to test so the result is meaningful and the cost of a mistake stays minimal.

A minimal purchase (usually a few dozen units), a full-fledged product card, and limited promotion for 3–6 weeks. The goal is not to earn but to gather data: main image CTR, add-to-cart and order conversion, buyout rate, first reviews. The test pays for itself in information even if the batch sells at break-even.

• Enough for a signal. A batch that is too small (5–10 units) goes out of stock before the algorithm has time to show the card; aim for several weeks of forecast sales.
• One SKU — one test. If you are testing variations (colors/sizes), plan depth for the popular ones and remember: the spread across variations is a test result too.
• Do not cut corners on content. A test with weak photos tests the weak photos, not the product. The card must be battle-ready: main image, attributes, description.

1. 1Create the card in advance, let it pass moderation, and build the semantics: keyword set for the card.
2. 2Give it launch traffic. A modest amount of ads on target keywords — a card with no impressions does not count as a test; on the newcomer window, see advertising a new product.
3. 3Collect the first reviews through legitimate mechanics: how to get the first reviews.
4. 4Log metrics weekly: impressions, CTR, order conversion, buyout rate, ad spend share (DRR).

• Scale: conversion and buyout rate at or above the category level, ads pay off in the unit economics, reviews show no systematic negatives.
• Rework: traffic but no conversion — fix the card or the price and remeasure; conversion but no buyouts — check the reasons behind failed buyouts.
• Close: no clicks despite impressions (main image/price/demand), negative feedback on product quality, or the economics do not work even without ads. Sell off the remainder and exit.

Treat the test as a cost of data: a small-volume purchase costs more per unit, ads on a cold card cost more than average — a loss on the test itself is normal. The question is not "did the test make money" but "did it confirm the economics at scale": the break-even point is calculated on the main purchase volume, where both the unit cost and the per-unit advertising are lower.

"How much product do you need for a demand test?" A few dozen units — enough for several weeks of forecast sales. A batch that is too small goes out of stock before the card gathers statistics, and the test has to start over.

"How long does the test take?" Usually 3–6 weeks with weekly metric tracking. No decision is made on the first days — early sales are distorted by the newcomer effect.

"The test failed — what to do with the leftover stock?" Diagnosis first: if the problem is the card or the price, reworking and remeasuring is cheaper than starting a new niche. If it is the product itself, sell off through promotions or markdowns and exit.

"Can you test without advertising?" Almost never: a new card with no traffic gets no impressions, and you end up testing your patience rather than demand. A small ad budget is a mandatory part of the test.

"When should you scale the purchase?" When conversion and buyout rate stay stable at the category level for several weeks in a row, ads pay off in the unit economics, and reviews have revealed no quality problems. Then volume grows while the unit cost and per-unit DRR fall.