Somewhere in a notes app there is a product you keep coming back to. You can picture the listing, maybe even the ad. The only thing between you and a purchase order is a question that feels impossible to answer from the outside: will this actually sell?
Framed that way, it is impossible. "Will it sell?" asks for a prediction, and no public data source hands out predictions. But the question decomposes. Hidden inside it are four smaller questions you can answer today, with free tools, from a laptop: is anyone paying to sell this thing, is anyone looking for it, is anyone already buying it, and is anyone talking about it.
Each of those leaves a public evidence trail. This post walks through all four signals — what each one measures, how it lies, and what to do when they disagree — and ends with a one-hour manual pass you can run on any idea before you commit money to it.
The question behind the question
The four signals matter because they are independent. Competitor ads live in the Meta ad library. Search demand lives in Google's query logs. Retail proof lives in Amazon's ranks and reviews. Community chatter lives on Reddit, TikTok and X. Different platforms, different populations, different incentives.
Any single one of them can mislead you. An ad can be a founder burning savings on a bad bet. A search spike can be a news story. A best-seller rank can be juiced for a week. A viral thread can be entertainment rather than intent.
But they rarely lie in the same direction at the same time. When ads, search, retail and chatter all point the same way, you are looking at the cheapest reliable evidence a seller can get without spending on inventory or traffic. Agreement across independent witnesses is the whole method.
Signal 1: competitor ads are skin in the game
Of the four, paid advertising is the only public signal where the publisher loses money by being wrong. Reviews can be farmed, threads can be astroturfed, searches cost nothing. An ad account bleeds real cash every day a losing campaign stays live — which is why sustained spend is such a strong tell.
The Meta ad library makes this visible for free. Search your product term and read three things. First, advertiser count: how many distinct stores are paying to sell this right now. Second, ad age: an ad that has run for thirty days or more is, in all likelihood, paying for itself — nobody funds a loser for a month. Third, variation clusters: an advertiser running six versions of the same product ad is scaling something that works, not testing something that might.
One old ad from one store is a hint. Five advertisers with month-old ads and creative variations is a market. The full reading method is in our deep dive on the Meta ad library.
Signal 2: search demand is unprompted intent
Nobody makes a person type a query. Search volume is demand in its rawest form — unprompted, unpaid, expressed in the buyer's own words. Google Trends shows you its shape for free.
The shape matters more than the size. A term that has climbed steadily for eighteen months tells you a market is forming; a single tall spike tells you something happened once. Direction and stability beat raw volume, because you can enter a small growing market and grow with it, while a spike has usually collapsed by the time your stock clears customs.
Two traps catch almost everyone. Trends reports a relative index from 0 to 100 — the 100 is the term's own peak, not a real quantity, so a niche term sitting at 90 can still be tiny. And seasonality masquerades as growth: compare against the same month last year, not against last month. Both traps, and how to sidestep them, are covered in our Google Trends field guide.
Signal 3: retail proof is the honest trailing indicator
Amazon's best-seller ranks and review counts are downstream of one thing: completed purchases. That makes retail the most honest of the four signals — and the most backward-looking. It tells you what sold last month, not what will sell next quarter.
The honesty comes from scale. A single listing can be gamed with bought reviews or a burst of manipulated sales. An entire category — a dozen listings all carrying ranks, all accumulating fresh reviews — cannot be faked cheaply. So read the category, not the hero listing.
Review velocity is the working metric: open the leading listings and look at the dates on the newest reviews. A steady drip from the past two weeks means orders are flowing now. Ten thousand reviews that mostly date from two years ago mean you have found a monument, not a market. The full method is in our piece on Amazon rank and review velocity.
Signal 4: community chatter is the leading indicator
People complain about problems before they search for solutions, and they search for solutions before they buy them. That ordering makes Reddit threads, TikTok comments and X complaints the earliest signal in the chain — pain shows up in community language months before it registers in search volume.
Chatter carries a second gift: vocabulary. The exact phrases buyers use to describe their problem are the phrases that will later make your ads and listings convert. No keyword tool hands you "my dog hates the cone and I haven't slept in three days."
The discipline is separating curiosity from purchase intent. "This is so satisfying to watch" is entertainment. "Where do I get one," "does it work on thick hair," "mine broke after a week, what's better" — that is a buyer talking. Weigh the second kind, discount the first, and treat any sudden burst of attention with suspicion until the other three signals confirm it.
When the signals disagree
Perfect agreement is rare. The useful skill is reading the mismatches, because the mismatch is the diagnosis.
- Ads without search. Someone is spending, but nobody is looking. This is manufactured demand — the product only moves while advertising pushes it. It can work, but you must buy every single customer, and the market evaporates the day the ad economics stop working.
- Search without ads. People are looking and nobody is paying to be found. Either you have found a genuine opening, or the unit economics have already broken everyone who tried. Check margins, shipping weight and return rates before celebrating.
- Chatter without retail. A community is loudly describing a problem no product on Amazon solves well. This is the classic unserved market — the earliest and riskiest of the good signs, and the one worth the most when it is real.
- Retail without chatter or ads. Steady purchases, no conversation, no paid pressure. Usually a mature staple with commodity margins: real demand, little room.
- Everything strong at once. A validated market that everyone else has also validated. The question shifts from demand to crowding — see how to tell if a product is saturated.
The one-hour manual pass
Here is the whole method as a checklist. One idea, one hour, one page of notes.
- Ads — 15 minutes. Search the Meta ad library for your product term and two variants. Record: the number of distinct advertisers, the start date of the oldest still-active ad, and whether any advertiser runs three or more creative variations of the same product.
- Search — 10 minutes. Open Google Trends, set the window to five years, and check the product term plus its category term. Record: direction over the past eighteen months, the seasonal shape, and whether the current level sits above the same month last year.
- Retail — 15 minutes. Search Amazon for the product. Record: the best-seller ranks of the top three listings, their review counts, and the dates of the ten most recent reviews on the leader.
- Community — 20 minutes. Search Reddit for the problem the product solves, skim TikTok comments on related videos, and run the term through X. Record: three recurring complaints in the buyer's exact words, and a rough ratio of purchase language to curiosity.
At the end you hold four verdicts on one page. Three or four pointing the same way is a decision. A split is a diagnosis you now know how to read.
Where a radar compresses this
The honest cost of the manual pass is about an hour per idea — and most sellers are weighing five or ten ideas at a time, which is where the discipline usually breaks down. Kestrel exists to run exactly these four checks on demand: it reads live competitor ads from the Meta ad library, Google search demand, Amazon retail proof and community chatter, then returns a 0–100 score — Hot, Promising or Weak — with the evidence broken out per signal, so you can audit its reasoning the way you would audit your own notes. A real report is public at /specimen if you want to see the output before trusting it, and 20 market scans are free with no card. The manual method above is still worth knowing; the score summarizes it, it does not replace understanding it.
What validation cannot tell you
All four signals measure the market. None of them measure you. A hot market says nothing about whether your supplier ships consistent quality, whether your margins survive returns and ad costs, or whether your creative stops a thumb mid-scroll. Execution — pricing power, sourcing, creative, cash discipline — still decides the outcome inside whatever market you pick.
So use validation for what it is actually good at: killing bad ideas cheaply. Most product ideas fail one of the four checks in under an hour, and every idea that dies on paper is inventory you never bought and ad budget you never burned. The occasional winner you pass on is the price of never boarding the clear losers.
Whether you spend the hour by hand or spend a free scan instead, the sequencing is the point: read the market before the purchase order, not after.
