How to Research Products to Sell on Amazon Before Buying Inventory

Finding a product to sell on Amazon starts with a customer need, not a screenshot of a bestseller list. A popular listing tells you that something is being bought. It does not tell you whether a new seller can source a legitimate version, get approved to list it, offer something worthwhile, or earn enough to cover the risk.

A useful product research process ends with a decision record. You should know what the idea is, which evidence supports it, which assumptions remain, and what you can afford to test. You do not need to choose a final product or buy inventory just to finish researching one.

Define one specific customer problem

Start with a situation that leads someone to search for a product. For example, “people need a small organizer that stays put on a narrow desk” is more useful than “desk accessories are trending.” The first description suggests requirements you can investigate: footprint, stability, materials, installation, and the difference between a temporary and permanent setup.

• Who is the buyer, and what are they trying to accomplish?
• What existing products do they already use or consider?
• What constraints make an alternative more useful: size, durability, instructions, storage, safety, or price?
• What evidence would change your mind about pursuing this product?

Avoid deciding that a niche is “underserved” just because you personally dislike current listings. Look for repeated, specific buyer friction and confirm that the friction matters enough to affect a purchase.

Look at demand and alternatives without confusing signals

Browse the relevant search results and product details. Note the range of prices, common features, product quality claims, delivery expectations, and how differentiated the offers really are. Read both positive and critical reviews to understand what customers value, but do not treat any individual review as proof of a whole market.

For sellers with appropriate Seller Central access, Amazon offers Product Opportunity Explorer to examine customer search behavior, trends, competition, and review insights. Its data can inform research, but it does not guarantee demand for your proposed product or that you can reach those buyers profitably.

1. Search for the customer problem using several ordinary buyer phrases, not only your preferred product name.
2. Record comparable listings, price ranges, core features, and visible evidence of buyer concerns.
3. Separate recurring observations from isolated opinions; write down the number of listings and reviews you actually inspected.
4. Check whether demand appears seasonal, price-sensitive, or concentrated in a few established brands.
5. Write one possible point of difference that could be delivered and demonstrated honestly.

Do not mistake a demand signal for permission to sell

A strong search pattern does not solve sourcing, brand rights, safety rules, or category approval. A seller might identify demand for a branded or regulated product yet lack authorization or the documentation required to list it. Those are possible stop conditions, not details to handle after the first inventory order.

Check listing eligibility before sourcing

Amazon explains in its seller FAQ that some products or categories need approval, some carry restrictions, and some cannot be offered by third-party sellers. Requirements also vary by marketplace and product. Check the actual category, brand, condition, and intended product before purchasing.

• Category, brand, and product-level approval requirements.
• Authentic, traceable sourcing and any permitted resale or distribution rights.
• Safety, labeling, testing, regulatory, and product claims requirements where relevant.
• FBA-specific product eligibility and handling restrictions if you intend to use FBA.
• Intellectual-property concerns, including protected branding, imagery, and designs.

If you need permission or documents you do not have, mark the candidate “not currently eligible” rather than assuming you can obtain them later. A product listing you can see as a shopper is not confirmation that you personally may sell it.

Make a supplier evidence record

Write down what you actually know about the proposed source. A supplier’s catalog picture is not proof that the delivered units will match the listing, meet quality requirements, or arrive on time. If you do not have a source, “source not identified” is a legitimate research result.

• Supplier identity, production or resale route, geographic location, and the exact SKU or specification.
• Price and minimum order quantity, clearly labeled as quoted or estimated.
• Samples or test reports available and what they do not prove.
• Lead time, packaging, shipping, inspections, replacements, and documented requirements.
• Whether the source can provide appropriate invoices, authorization, or compliance information where necessary.

For a broader method of comparing customer alternatives, use our competition-research guide. This Amazon-specific research adds the permissions and marketplace operations that a general business idea study cannot resolve.

Test the unit economics using conservative inputs

Now estimate the product as one completed sale. Use the expected customer price, landed unit cost, referral charge, fulfillment method, inbound and inventory allocation, customer acquisition, and a reasonable treatment of returns. Separate confirmed numbers from rough scenarios.

One deliberately hypothetical product

Suppose a candidate sells for $24. For a planning exercise, estimate $7.50 landed cost, a $3.60 referral fee, $4.70 fulfillment, $1.10 inbound and storage allocation, and $3 for advertising and return allowance. That leaves $4.10 per completed sale before fixed costs, owner pay, tax, and unusual expenses: $24 − $7.50 − $3.60 − $4.70 − $1.10 − $3 = $4.10.

If competing sellers force the price down to $21 while all those assumed dollar costs stay constant, the contribution becomes $1.10. In practice, some percentage-based fees may also change with selling price, so a real estimate needs to be recalculated. Either way, the scenario reveals how little room the original plan may have for mistakes.

Check actual costs with Amazon’s Revenue Calculator and record the date. For a deeper checklist of fee categories, see our FBA introduction. Do not turn a hypothetical fee rate into a claimed current platform rate.

Evaluate inventory risk, not just a sale

The most dangerous assumption might be the minimum order quantity. A supplier asking for 500 units changes the risk of a product that has no verified buyers. Slow sales can tie up working capital and create storage or removal decisions. A good per-unit estimate cannot tell you whether you should commit to that number of units.

• How much cash must leave before a single sale is possible?
• How many units would remain if sales were half the optimistic case?
• What could go wrong with the first batch, and who pays for fixing it?
• What is the exit route for unsold, damaged, or ineligible inventory?
• Can a smaller sample or another evidence-producing step reduce risk before the order?

Make a go, revise, hold, or stop decision

1. Go to the next bounded test only if listing feasibility, customer evidence, source credibility, and downside economics are sufficiently supported.
2. Revise if a particular requirement or cost can be meaningfully changed without weakening the actual customer value.
3. Hold if material information, authorization, quality evidence, or capital protection is missing.
4. Stop if the model depends on a permission you cannot obtain, claims you cannot support, or a loss you cannot responsibly absorb.

When your largest unknown is demand rather than marketplace operations, use how to validate a business idea before spending money to design the next test. You do not need to turn every research idea into an inventory order.

Leave with a product-research brief

For each candidate, save a one-page brief with the customer and problem, two or three competing alternatives, price evidence, product requirements, listing eligibility, supplier route, fee estimate, required cash, downside case, and next decision. Date every changeable fact. Keeping this record prevents a promising-looking listing from silently becoming an approved purchase.

Research cannot eliminate uncertainty. It can reveal whether the remaining uncertainty is small enough to justify the next responsible experiment.