Amazon Product Research That Finds Real Demand

Every other decision in an Amazon business can be fixed. A bad listing can be rewritten. A bad ad campaign can be paused. A bad supplier can be replaced.

A bad product cannot be fixed. You find out it was wrong roughly ninety days after the money left your account, and the only exits are a fire sale or a removal fee.

So product research deserves more rigour than it usually gets. What follows is a filter designed to reject, not to find. Most candidates should fail it, and that is the point.

Start With the Rejections

Before evaluating anything, throw out entire categories. This takes minutes and removes most of the ways to lose money.

Gated categories. Some categories need approval, documentation, or invoices from authorised distributors. Not impossible, but not a first product.

Anything requiring certification. Electricals need BIS. Cosmetics and food need FSSAI or CDSCO involvement. Toys need safety marks. Each is a real process with real timelines.

Brand-dominated categories. If one brand holds 80 percent of the top listings, you are not entering a category, you are attacking a moat.

Fragile, perishable, or oversized. Breakage, expiry, and fulfilment fees respectively. Any one of them can consume the margin.

Anything with a patent or trademark risk. Selling something that infringes ends in a takedown and possibly an account issue. If a design looks distinctive, assume somebody owns it.

Seasonal-only products. Eleven months of storage fees to sell in one is a bad shape for a first product.

What is left is smaller than you expect, and much safer.

The Four Filters

Run every surviving candidate through all four, in order. Fail any one and stop.

Filter 1: Is there demand?

Not "would people want this". Are people buying it right now.

What to look for:

• The top 10 listings in the category collectively move 300 or more units a month
• Demand is spread across several sellers, not concentrated in one
• Search volume is stable across 12 months in Google Trends

Why spread matters: a category where one listing does 900 units and the next nine do 20 each is not a category with demand. It is one product with a brand behind it.

Where to look: Helium 10 and Jungle Scout both estimate this. Amazon's own Best Sellers, Movers and Shakers, and New Releases lists are free and directionally useful. Category rank on a listing translates roughly to volume once you learn the mapping for your category.

Filter 2: Can you compete?

Demand with unbeatable competition is worse than no demand, because it looks promising.

Signals you can compete:

• Top listings have fewer than 500 reviews
• Several have ratings below 4.3, which means unmet complaints
• Listings have weak images, thin bullets, no A+ Content
• Reviews contain specific, fixable complaints

That last one is the most valuable signal in product research and almost nobody uses it properly.

Read the three-star reviews. Not the one-stars, which are often delivery complaints, and not the five-stars, which tell you nothing. Three-star reviews are from people who wanted to like the product and were let down by something concrete. "Good bottle but the lid leaks in a bag." "Works well but the cable is too short."

Each of those is a product specification for a better version, written by a paying customer. A product that fixes a complaint appearing across four competitors is a genuinely differentiated product, and you can say so in your bullets.

Filter 3: Does the margin survive?

Most candidates die here, and they should die here rather than after the purchase order.

Full working in our breakdown of Amazon FBA fees in India, but the short version:

Selling price
  minus referral fee (2 to 17 percent by category)
  minus fulfilment fee (by size tier, at the NATIONAL rate)
  minus closing fee
  minus storage
  minus returns provision
  minus cost of goods
  minus inbound shipping
  minus advertising (budget 12 percent)
  = net margin

Require 25 percent or better. Below 15 percent there is no room for a competitor undercutting you, a weak ad month, or a returns spike, and you will get all three.

Two mistakes to avoid. Do not model at the local fulfilment rate if you intend to sell nationally. And do not leave advertising out, which is how a 22 percent product appears to be a 34 percent product.

Filter 4: What is the sourcing risk?

A product that passes the first three can still fail here.

• Minimum order quantity. A supplier requiring 1,000 units for an unproven product is asking you to gamble, not to test.
• Lead time. Ninety days from order to fulfilment centre means you commit to a reorder before you know how the first batch sells.
• Sample quality. Order samples from three suppliers, always. Pay for them yourself.
• Import exposure. Duties on many categories run 20 to 40 percent, plus GST on import. Domestic sourcing through IndiaMART or direct from manufacturers avoids this entirely and is the lower-risk starting point.
• Replaceability. If exactly one supplier in the world makes this, one bad quarter for them is a bad quarter for you.

A Worked Example

Candidate: insulated steel water bottle, 1 litre.

The leaking lid is the whole opportunity. Source a version with a genuinely better seal, photograph it sealed and inverted in a bag, and lead the first bullet with it. You are no longer a commodity listing competing on price.

Tools, Honestly

Helium 10 is the most complete and the most expensive. Worth it once you are trading.

Jungle Scout is cheaper and adequate for demand estimation.

Google Trends costs nothing and is the fastest way to catch seasonality.

Amazon's own Best Sellers, Movers and Shakers, and New Releases are free, current, and underused.

Reading competitor reviews costs nothing, takes an afternoon, and produces better insight than any paid tool. Do this even if you buy nothing else.

Estimation tools are estimates. They are directionally useful and precisely wrong. Never let a tool's number be the only evidence behind a purchase order.

How Long This Should Take

If a product passes all four filters in an hour, you have not looked hard enough.

A serious pass is two to three weeks: a fortnight building and narrowing a candidate list, then samples from three suppliers, then a decision. Against a first order that is often several lakh rupees, three weeks of research is cheap.

Have at least ten candidates before you commit to one. With three candidates you talk yourself into the best of three. With thirty you can afford to be honest about the flaws.

Once You Have Picked

The listing is what converts the demand you found. Images, bullets, and backend keywords are covered in our Amazon FBA beginner guide, and A+ Content handles the rest once you have Brand Registry.

Want a Second Opinion Before You Order?

Our ecommerce team has run this process across dozens of categories on Amazon.in. The most useful thing we do is tell people not to buy something.

Send us your shortlist.