Validation is a purchase, not a spreadsheet. Buy about 200 units, spend $5,000 to $10,000, run up to four candidates at once, and let rating, conversion rate, and acquisition cost decide. Scale only when all three hold. Everything before that order is research, not proof.
The short version
- A small real order beats a large forecast. Only paying customers report return rate and conversion rate honestly.
- Test four candidates in parallel, not one at a time. Sequential testing costs a year to learn what a quarter can teach.
- Three numbers rule the decision. Rating trend, conversion rate, acquisition cost. Two out of three is not a pass.
- Differentiation comes from competitor negative reviews. Not from invention, and never from what you personally find annoying.
- Write the kill criteria before the inventory lands. Nobody has ever written fair criteria while sitting on unsold stock.
The moment this page is for
You have shortlisted a product. The supplier quotes are in, the sample is on your desk, and the minimum order is larger than you expected. The spreadsheet says the margin works. Everyone you have asked says it looks good.
This is exactly the point where money gets lost, because everything you have so far is a projection built on public data, and public data cannot tell you what your particular version of this product does to a customer once it arrives in their kitchen. That answer costs a few thousand dollars and about eight weeks to buy. Skipping it does not save the money. It moves the same lesson to a purchase order ten times larger.
The failure modes, ranked by what they cost
| Failure mode | What it costs | The check that catches it |
|---|---|---|
| Scaling a product the market never wanted | The full purchase order plus a year of attention | A small validation buy with pass or fail written first |
| Judging by revenue instead of unit economics | Slow bleed until the cash runs out | Contribution margin after acquisition cost, per unit |
| Ignoring returns | Rating collapse, then rank collapse | Return reasons read one by one on the first 200 orders |
| Differentiating from imagination | Features nobody searched for | Negative reviews of the top ten competitors, sorted by frequency |
| Depending on a single route to the customer | Sudden fragility when that route moves | A written plan for more than two channels |
| Research that ends at review count and revenue | Entering a category where you cannot win | Depth: growth, returns, segment dynamics, rating gap |
The first one is the expensive one, and it is expensive because it is slow. A product that fails quickly costs its inventory. A product that fails slowly costs three quarters of your attention while you keep believing the next fix will do it.
The channel row deserves a specific note. There are five ways traffic reaches an Amazon listing: organic search, paid placements, promotions and deals, influencer and creator content, and off-channel traffic you send yourself. Most sellers run two, usually organic and paid, then discover their entire growth model depends on one auction. Validation should tell you whether a product can be sold through at least one channel beyond those two, because that is what makes it durable rather than merely viable.
The validation buy, step by step
- Size the category first. Growth direction, price band distribution, average rating of the top ten, and how concentrated the reviews are. If the top three own the page and nobody is complaining in their reviews, there is no wedge.
- Read the complaints before you design the product. Sort competitor negative reviews by theme. The most repeated fixable complaint becomes your primary differentiator and your hero image.
- Order about 200 units. Budget $5,000 to $10,000 including freight and samples. Run up to four candidates simultaneously so the comparison is real rather than seasonal.
- Write the pass and fail lines now. Target rating floor, conversion rate floor, and maximum acquisition cost, with the date you will judge them. Write them into a document with your own name on it.
- Sell them properly. A validation buy sold through a weak listing tests your listing, not your product. Photography, copy, and a modest advertising budget are part of the test, not optional extras.
- Read the returns individually. The first thirty return comments are the most valuable market research you will ever get, and they are free.
- Decide on the date, not the mood. Scale, fix, or kill. Then act the same week.
What most agencies will not tell you
An agency paid to run a product has no financial reason to tell you the product should not exist. That is not corruption, it is structure. The scale, fix, or kill conversation reduces their revenue, so it tends to arrive late or never, dressed up as another creative refresh.
The way to fix the structure is to ask for the criteria in advance. Ask any provider what evidence would make them tell you to stop, and by what date. Vague answers about giving it more time mean the decision will be made by your bank balance instead of by data. We publish our criteria before onboarding for that reason, and I would rather lose a product from the roster than manage one that should not be on it.
Related answers
- What products to start with for first Amazon private label
- What tools to use for Amazon product research
- How to manage inventory and avoid stockouts for private label
- Common mistakes in Amazon FBA launches and how to avoid
- Amazon launch services: the complete guide
We will size your category and draft the pass or fail lines with you in the free written audit at Flapen.
