Work backwards from failure. Most products lose money for one of six reasons, and five of the six are visible before you order. Screen candidates by the mistake each one is most likely to make rather than by revenue estimates, and a winner is whatever survives every screen.
The short version
- Winners are found by elimination. Nobody spots a winner. They eliminate the losers and something is left standing.
- Five of the six failure modes are visible in advance. Only one of them requires selling to discover.
- Rank your screens by what the mistake costs. Cheap mistakes go last, expensive ones go first.
- A high revenue estimate is not evidence. It tells you the category earns, not that you will.
- The product that survives will look boring. Boring is what a solved screen looks like from the outside.
What buying brands taught me about picking them
Before Flapen I ran data and technology at BRANDED and at Moonshot Brands, two large Amazon aggregators. My job was to look at hundreds of seller businesses that were already trading and work out which ones were worth buying. That is an unusual angle on this question, because you are reading the ending rather than the pitch.
The pattern that stuck with me is how rarely a failed product failed for an exotic reason. It was nearly always one of a handful of things, decided long before launch, and usually visible in the listing and the reviews if you knew where to look. The businesses that were worth acquiring had not found a magic category. They had avoided the six mistakes below.
Six ways an FBA product loses money, ranked by cost
| Rank | Failure mode | About what it costs | Visible before you order |
|---|---|---|---|
| 1 | Legal or gating problem discovered after stock lands | The entire run, plus freight | Yes, with a patent and category check |
| 2 | Category too small to pay back acquisition cost | The run and the year | Yes, by sizing the market first |
| 3 | Unit economics that never worked | Every unit sold makes it worse | Yes, with a full cost stack |
| 4 | No real difference from the incumbents | Permanent price competition | Yes, from the reviews and the rating gap |
| 5 | Product quality fails in the buyer's hands | Rating collapse and returns | Partly, through samples and supplier vetting |
| 6 | Demand is real but the buyer does not choose you | A slow, confusing underperformance | No, this one needs live sales |
Read that ranking as your screening order. Most people research in the exact opposite direction: they get excited about a difference, then check economics, then think about legal at the end if at all. Run it top down instead and you throw candidates away at the cheapest possible moment.
Failure one, in more detail
A patent complaint or a gating requirement discovered after the container arrives is the worst available outcome, because the money is spent and the listing cannot trade. The check is not expensive. Search the specific mechanism your improved version relies on, read the category requirements, and confirm no single brand controls the main search term through registry enforcement. Twenty minutes of this saves a five figure mistake often enough that I treat it as gate one everywhere.
Failure two, in more detail
Small categories fail politely. Nothing goes wrong, the reviews are fine, the listing converts, and there is not enough revenue in the room to pay back what it costs to acquire a customer. The tell is a category where the tenth ranked listing sells almost nothing. Depth matters more than the top seller's number, because you will not be the top seller in year one.
Failure four, in more detail
Sameness is the most common quiet killer. If your product differs only in color and packaging, you have entered a price war with people who bought more units than you did. Your difference has to come from somewhere real, which in practice means the repeated complaint in competitor negative reviews and the rating gap it produces. Invention is not required and is usually a trap, because inventing demand costs more than capturing it.
The screen that catches five of the six
- Legal room. Patent search on the feature, category requirements read, no brand controlling the term.
- Market depth. Category revenue deep enough that a realistic share pays back acquisition cost.
- Cost stack. Landed cost, referral fee, fulfillment, storage, returns, promotions, advertising. Written out, not estimated.
- A difference you can point at. Name the review complaint and the manufacturing change that answers it.
- Supplier proof. Samples from more than one factory, and a supplier who asks questions rather than agreeing to everything.
- A small live test. The only screen that requires buying inventory, which is why it goes last.
Run in that order, most candidates die on steps one to three, where the cost of being wrong is a few hours.
What most agencies will not tell you
Product research is the least profitable part of this business to do properly and the easiest part to appear to do. A shortlist can be generated from a subscription tool in an afternoon, presented in a deck, and charged for. Doing it properly means reading reviews by hand, running searches that mostly return nothing, quoting suppliers on a change that may not be manufacturable, and rejecting almost everything. The second version takes days per candidate and looks like far less work from the outside.
The second thing, and this is from the buy side rather than the sell side: nobody acquiring brands cares which tool found the product. They care about rating stability, margin after all fees, and whether growth came from something repeatable. If you are choosing products with an eventual sale in mind, screen for those three now, because that is what the room will ask you about later.
Related answers
- Product criteria checklist for Amazon private label
- Recommend a product research framework for Amazon
- Niche ideas for first-time Amazon sellers
- How to use reviews to discover product gaps
- Amazon seller roadmaps and capital: the complete guide
Send us a candidate and we will tell you which of the six it is most likely to fail on, free, at Flapen.
