How to know if a product will sell on Amazon

It comes down to three questions

Strip away the tooling and the frameworks and product research is three questions, in this order. Each one can kill the idea, and the order matters because they get more expensive to answer.

1. Do people buy this? Demand. Usually yes, and cheap to confirm.
2. Can I take a share of it? Competition. The one that decides most outcomes, and the one most often skipped.
3. Is the share worth having? Economics. Kills more good products than people expect.

A product needs all three. Most failed launches passed the first, never seriously asked the second, and were optimistic on the third.

Question 1 — Do people buy this?

The easy one, and the one people over-research. If there are twenty listings on page one with real review counts and recent reviews, demand is proven — twenty sellers are making sales right now. You do not need a precise figure to proceed.

What is worth checking:

• Are reviews recent? Recency is live demand. Totals are history.
• Does the depth run down the page, or stop at position three? If only the top few have traction, demand concentrates at the top and position eight is a much worse place than it appears.
• Is it seasonal? Rank history answers this in a minute and prevents the most expensive demand error — researching during a peak.
• Is the trend flat, rising or dying? A declining category is a bad bet regardless of its current size.

Amazon's own Product Opportunity Explorer in Seller Central gives first-party demand data for free, which is better than any modelled estimate. Use it before paying anyone for an inference of the same thing.

Question 2 — Can I take a share?

The decisive question, and the one that separates sellers who succeed from sellers who researched thoroughly and still failed.

Demand is not the scarce resource. Position is. There are twenty slots on page one and they are occupied. Whether you can have one depends entirely on whether the current occupants are beatable — by you, with your budget, within a timeframe you can fund.

The check

1. Incognito window, correct marketplace, Sponsored placements excluded.
2. Open the top 20 organic listings.
3. For each: review count, rating, seller name, price, and a strict quality judgement.
4. Look at the review distribution — flat, deep, steep or barbell — not the average.
5. Count generic sellers versus real brands. Generic holders do not defend positions.
6. Count how many listings you could produce a visibly better version of, within your actual budget.

Being strict here is the whole discipline. "I could probably do better" is not the standard. "I can produce a visibly better listing with the money I actually have" is.

Question 3 — Is the share worth having?

A winnable position with no margin in it is a worse outcome than a rejection, because you spend months and capital discovering it.

1. Take the median price across the top 20 — not the top listing's, not your hoped-for price.
2. Run it through Amazon's FBA revenue calculator for referral and fulfilment fees, using real dimensions and weight.
3. Subtract landed cost — factory price plus freight, duties and inspection.
4. Subtract a real PPC allowance. New listings are invisible without it.
5. Subtract a returns allowance where the category warrants it.

Then stress it: drop the price ten percent, double the advertising cost, halve the volume. A product that survives all three is robust. One that fails two should not receive an order.

What research genuinely cannot tell you

Worth being honest about the limits, because overconfidence in research is its own failure mode.

• Whether a new competitor arrives next month. Markets move. A page open in March can have three strong entrants by June.
• Whether your supplier will deliver the quality in the sample. Only production runs answer that.
• How your listing will actually convert. You can assess your competitors' presentation; you cannot know how yours performs until it is live.
• Whether Amazon changes something. Fee schedules, category rules and algorithm behaviour all change.
• Whether demand shifts. Especially for anything trend-adjacent, where private label lead times are long enough for the trend to move first.

This is why the goal is not certainty but cheap wrongness. Keep the first order small enough that a mistake is survivable. Validate before spending rather than after. Re-check the page before committing inventory, because weeks have passed. A seller who is right seventy percent of the time with small downside outperforms one who is right ninety percent of the time with a business-ending bet.

Getting to the answer faster

The full sequence takes about forty-five minutes per candidate done properly. That is fine once. Across a shortlist of fifteen, plus variants, plus other marketplaces, it becomes the constraint on how many ideas you check — and the ideas you skip are disproportionately the non-obvious ones.

SellRadar compresses question two: type a product, and it reads the top 20 organic listings on your chosen marketplace, flags each competitor's specific weaknesses, runs the fee stack against the page's price band, and returns LAUNCH, WATCH or SKIP with the reasoning written for your selling model.

Three verdicts a month are free with no card. It will not make you certain — nothing will — but it makes checking the uncertain idea cost a minute instead of an hour, and that changes which ideas you check.

How confident should you actually be?

Research produces a judgement, and judgements come with confidence levels that people rarely state explicitly. Being honest about yours changes how much you should commit.

Notice that the two lowest-confidence rows are at the end, and neither is addressable by more research. That is the ceiling: you can reach solid confidence about the market and only weak confidence about execution and the future.

Which is exactly why order sizing matters more than certainty. If your confidence about the market is high and your confidence about execution is low — which is the normal state for a first product — the correct response is to commit enough to test the execution and not so much that being wrong about it is terminal.

Sellers who last are not the ones who found a way to be certain. They are the ones who priced their uncertainty correctly.

Frequently asked questions