How Amazon Ranks Products: What Is Documented and What Is Myth

Few topics in Amazon selling attract as much folklore as search ranking. Amazon publishes the principles but not the formula, so the rest is inference, and inference gets repeated until it sounds like fact. This article separates what is reasonably well supported from what is assumed, then gives you a working method: split search sales into impressions, clicks and orders, fix the weakest of the three, and judge the result by the profit that remains after advertising rather than by a rank number.

What Amazon has said about its own search

Amazon describes the aim of its search in shopper terms: show the products that match what someone searched for and that they are likely to want to buy. Names such as A9 or A10 circulate in the seller community as labels for the system, but they are community shorthand, not documentation you can check.

From Amazon's public guidance and from the way the results behave, the signals fall into two families.

• Relevance: how well your listing text and product data match the query. Title, bullets, product attributes, category placement and backend search terms all feed this.
• Performance: how shoppers behave once your product is shown. Clicks, purchases and conversion are the obvious inputs. Price relative to similar products and whether you can actually deliver (in stock, with a competitive delivery promise) belong here as well.

Relevance gets you into the results for a query. Performance helps decide where you sit once you are there. Amazon does not publish the weights, and they likely vary by query and category, so anyone quoting a precise recipe is guessing.

Five myths worth dropping

The comparison below pairs common beliefs with what the evidence supports. Where Amazon has not documented something, treat the effect as plausible, not guaranteed.

The pattern is the same in every row. A tactic gets promoted because it appears to correlate with results, while the real mechanism is a shopper doing something sensible, such as clicking an image that looks right or buying a product whose page answered their questions.

Relevance is about being understood

Relevance starts with Amazon knowing what your product is. That depends on structured data as much as prose: the product type, the category placement, and attributes such as size, material, color, compatibility and count. A listing placed in the wrong category, or missing attributes that shoppers filter on, can fail to appear for queries it deserves however good the title is.

Then comes language. Shoppers search with their own words, not your internal product names. Collect phrasing from your search term reports, from customer questions and reviews, and from the way competing listings describe the same use. Put the most important terms where both shoppers and Amazon look first, such as the start of the title, and write the rest for a human reader. A title that reads badly hurts clicks, and clicks sit in the performance family. Amazon also publishes style rules by category, so check yours before rewriting.

Search sales are three numbers multiplied together

Rather than chasing a rank position, think about what has to happen for a search to become an order. Your product has to be shown (impressions), the shopper has to click (click-through rate), and the page has to convince them (conversion rate). Orders from a query equal impressions × click-through rate × conversion rate.

Suppose a query shows your product 100,000 times. At a 1% click-through rate that is 1,000 clicks. At a 10% conversion rate that is 100 orders. Every number is a separate problem with a separate fix.

Because the three multiply, a modest lift in any one is worth more than it looks. Raising click-through from 1% to 1.5% gives 150 orders from the same impressions, a 50% increase. Raising conversion from 10% to 11% gives 110 orders. A better rank, if you earn one, raises impressions and lifts everything downstream.

Find the weakest link

• Low impressions: look at relevance (are the right terms in the title, bullets and attributes?), availability, and whether you are bidding on the query at all.
• Impressions fine, clicks low: the problem is in the search result itself. The main image, the title's first words, the price, the rating shown and the delivery promise are all that the shopper sees.
• Clicks fine, orders low: the problem is on the detail page. Look at images beyond the first, bullets, A+ content, reviews, price against nearby alternatives and stock.

The data is in places you probably already have. Brand Registry owners can use the Search Query Performance report in Brand Analytics, which shows impressions, clicks and purchases for the queries that matter to your products. Business Reports shows sessions and the unit session percentage by product. Your Sponsored Products reports show the same three steps for the keywords you pay for.

Availability and price shape the ceiling

Neither can be optimized by writing. An out-of-stock product cannot be bought, and Amazon has little reason to show it. A price well above comparable products depresses clicks and conversion, which feeds the performance family above. Undercutting in the hope of ranking has its own cost: you give away margin to buy a signal whose weight nobody outside Amazon knows.

The practical rule is to stay in stock, keep the delivery promise competitive, and price against the field at a level where you still make money. A sale you lose money on is not progress.

Judge the result by profit after ads

Best Sellers Rank reflects sales, in a category. It says nothing about what those sales cost you to win. A product can sit at a flattering rank while earning almost nothing, because the sales were bought with advertising.

Consider two hypothetical products. Product A sells 600 units a month and earns $5 per unit before ads, which is $3,000. It spends $2,400 on advertising, leaving $600. Product B sells 200 units a month and earns $12 per unit before ads, which is $2,400. It spends $600 on advertising, leaving $1,800. Product A moves three times as many units and would likely show the stronger rank, but Product B earns three times the profit.

None of this means rank is irrelevant. Visibility brings sales. It means rank is a means, and the end is profit. Compare ranks only within the same category, because a position in one category tells you nothing about a position in another.

The short version

• Match the query honestly in the title, bullets and attributes. Fill in every attribute that applies.
• Use backend search terms for synonyms and variants you could not fit in visible text. Do not repeat what is already there.
• Break each important query into impressions, click-through and conversion, and fix the weakest.
• Stay in stock and keep the delivery promise competitive.
• Price against the field, but never below a level where the sale pays you.
• Review profit after ads per product each month, and treat Best Sellers Rank as a side effect.