Demand research answers one question: if you list this product tomorrow, will enough people actually search for and buy it? Everything below is about triangulating an answer from imperfect signals, because no single data source gives you a reliable ground-truth number for marketplace-specific unit sales.

Signals worth checking, roughly in order of effort

  1. Marketplace search volume/keyword tools — the most direct signal for marketplace-specific demand, showing how often people actually search a given term on that platform. This matters because marketplace search behavior can diverge meaningfully from general web search behavior — a term can be a huge web search term and a mediocre marketplace search term, or vice versa.
  2. Existing competitor sales volume proxies — review count and review velocity (how fast reviews accumulate) on top listings are rough but useful proxies for sales volume, since only a fraction of buyers leave a review. If a listing has gone from 50 to 150 reviews in the last quarter, that velocity tells you more about current demand than a static total review count does.
  3. General web search trend data — useful for spotting seasonality and multi-year trend direction, less useful for marketplace-specific volume. A rising multi-year trend line combined with flat or declining marketplace listing quality is often a sign of an opportunity the incumbents haven't caught up to.
  4. Category browse rank / bestseller lists — a real-time signal of what's currently moving, though it reflects current winners more than untapped opportunity. Watch how rank moves over a few weeks, not just a single snapshot — a listing bouncing wildly in and out of the top 20 tells a different story than one holding steady.
  5. Question-and-answer sections and review text — buyers often describe exactly what they were trying to accomplish and whether the product delivered. This is slower to read through but gives you qualitative texture that a volume number can't.

Reading these signals together, not alone

No single signal is reliable on its own — a high search volume with almost no existing listings could mean an underserved opportunity, or it could mean the demand doesn't translate into marketplace purchases for structural reasons (impulse buy is used elsewhere, or fulfillment/logistics make the category unattractive for other sellers too). Cross-reference at least two signals before treating a category as validated.

A worked example

Say you're evaluating a niche kitchen gadget. Search volume tools show a steadily rising query trend over 18 months. You check the top 10 listings and find: review counts in the 200-800 range (moderate, not saturated), review velocity that's picked up in the last two quarters (consistent with the search trend), and a cluster of 1-3 star reviews complaining about a specific durability issue with the current top sellers' materials. That combination — rising demand, moderate competition, and a specific buildable improvement — is a much stronger case than any one signal alone. Compare that to a product with high search volume but where the top 10 listings all have 5,000+ reviews and 4.6+ average ratings: same demand signal, very different competitive reality (see Competitive Research for how to weigh that half of the picture).

A word on demand estimation tools

Third-party sales-estimation tools that convert rank data into unit-sales estimates can be directionally useful but are estimates, not ground truth — treat their output as one more signal to combine with the others above, not a precise forecast to build a budget around. These tools tend to be least accurate at the extremes (very high-rank bestsellers and very low-rank listings) and most useful for mid-pack comparisons between similar listings.

Sizing demand for a brand-new-to-market product

Genuinely new products (nothing similar currently sells on the marketplace) are the hardest case, because there's no existing listing to read signals from. In that situation:

  • Look for demand proxies in adjacent or substitute categories — what does a buyer currently do to solve this problem, and how big is that adjacent market?
  • Check general web search and social platforms for evidence people are already discussing the problem the product solves.
  • Plan a small test order rather than a full inventory commitment — real marketplace sales data from a limited launch is worth more than any pre-launch estimate.

Mistakes to avoid

  • Treating a single week's browse-rank snapshot as a stable demand signal — check it more than once.
  • Confusing high search volume for a broad category term with demand for your specific product variant within it.
  • Ignoring seasonality — a demand snapshot taken during a product's peak season will overstate its baseline, and one taken in its trough will understate it (see Seasonal Product Opportunities).

FAQ

Is review count a reliable way to estimate unit sales? Only as a rough proxy, and the conversion ratio (reviews per unit sold) varies a lot by category, price point, and how aggressively a seller solicits reviews. Use it for relative comparison between similar listings, not as an absolute sales estimate.

How do I estimate demand for a product with literally no marketplace listings yet? Look at adjacent/substitute product demand, general web and social search signals, and plan a small test order instead of relying on a pre-launch number.

How often should I re-check demand for a product I already sell? At least quarterly for anything exposed to trend risk or seasonality, since demand signals can shift meaningfully within a few months.