sellershortsHow Amazon Search Works: Matching and Ranking Explained
Amazon search runs in two steps. First it finds products that fit the search, then it puts them in order. Knowing which step you are failing tells you what to fix.
On this page
Quick Answer
Amazon search works in two steps. Step one is matching: Amazon gathers every product that could answer the search, using a keyword index plus a system that understands related words. Step two is ranking: Amazon puts that group in order using relevance and shopper behavior, mainly clicks and purchases. If your product never appears for a keyword, you failed step one. If it appears but sits on page four, you passed step one and need to work on step two.
- Matching decides if you are in the running at all
- Ranking decides what position you get
- Amazon understands related words, so you do not need to repeat phrases
- Fix matching first, then work on conversion to climb
Most advice about Amazon search jumps straight to tips. That skips the part that decides everything, which is how the system is built. Amazon has published its own research papers on this, and they describe a two step process rather than one big scoring formula.
Once you see the two steps, a lot of confusing problems get simple. A product that shows up nowhere has a different problem than a product that shows up on page five.
Reviewed by the SellerShorts editorial bench. SellerShorts runs an AI tool marketplace for Amazon teams.
Our Amazon Listing Optimizer takes an ASIN and returns a full optimized listing, covering title, bullets, description, and backend keywords, in one run. Push live to Seller Central in one click.
The two steps behind every Amazon search
When a shopper types something into the Amazon search bar, two different jobs happen one after the other.
| Step | What it does | What it decides |
|---|---|---|
| 1. Matching | Gathers products that could answer the search | Whether you appear at all |
| 2. Ranking | Puts those products in order | What position you get |
This split is the single most useful thing to understand about Amazon search, because the fix for each step is completely different.
Step 1: matching, or how Amazon builds the pool
Matching is the step where Amazon decides which products are even candidates. Amazon does this two ways at the same time.
- A keyword index. This is the traditional method. Amazon has a pre-built list of which products contain which words, so it can pull candidates fast.
- A meaning based system. Amazon's research team published a paper describing a second system trained on real shopper behavior. It catches products the keyword list would miss.
Amazon built the second system because plain keyword matching has real gaps. Their paper names three: it does not understand broader or narrower terms, it trips over word forms like woman and women, and it breaks on spelling mistakes.
The practical takeaway is simple. You still need your important phrases present in your listing, but you do not need to list every single variation of every word.
Step 2: ranking, or how Amazon sorts the pool
Once Amazon has the pool, it puts those products in order. This is where the signals most sellers talk about come in.
Amazon's own seller guidance points to a set of factors that affect where you land.
- Relevance. How well your listing fields match what was typed.
- Sales performance. Whether shoppers actually buy after seeing you.
- Competitive pricing. Whether your price is in line with the alternatives.
- Listing quality. Complete, clear fields and good images.
- Account health. Amazon's guidance notes this can contribute to search rankings.
Amazon's research adds one more piece. The model behind the results is trained on shopper behavior, and it learns from the difference between products people bought, products people saw but skipped, and products that were not relevant at all.
One honest warning. Amazon has never published how much each factor is worth. Any article giving you an exact percentage per ranking factor is guessing, not quoting.
Why the order matters more than the tips
Because matching happens before ranking, the two steps fail in different ways and need different fixes.
- You do not appear for a keyword at all. That is a matching problem. Price, ads, and reviews cannot help, because you are not in the pool being sorted.
- You appear, but far down. That is a ranking problem. You are in the pool. Now you need shoppers to click and buy.
Sellers waste a lot of money by treating the first problem with the second problem's tools. Running ads on a keyword you are not indexed for does not fix the underlying gap in your listing.
What Amazon actually says, in its own words
It is worth separating what Amazon has published from what the seller community repeats.
- Amazon publishes research papers describing a keyword index working together with a meaning based system, and a ranking model trained on shopper behavior.
- Amazon's search speed budget is tight. The same paper describes working inside roughly 20 milliseconds, which explains the reliance on a pre-built index.
- Amazon's seller guidance names factors, not weights. It lists relevance, sales, price, listing quality, and account health, with no numbers attached.
- Amazon does not use the names A9 or A10 in its current seller guidance. More on that below.
How to fix a matching problem
If your product does not show up for a search at all, work through these in order.
- Check your product type and category. If Amazon has your item filed as the wrong kind of product, correct matching becomes very hard.
- Fill your attribute fields. Size, color, material, count, and compatibility are structured data that helps Amazon understand what your product is.
- Make sure the phrase exists somewhere. It needs to appear once across your title, bullets, description, or backend search terms.
- Use the backend search terms field properly. Lowercase, separated by spaces, no commas, and no repeats of words already in your title.
- Do not repeat to try to force it. Amazon's guidance says once is generally enough, so use the space on a phrase you do not cover yet.
How to fix a ranking problem
If you appear but sit too low, you are being sorted, not blocked. Ranking work is mostly conversion work.
- Improve the main image. It decides whether anyone clicks you out of the results at all.
- Make the first bullet answer the buying question. Shoppers scan, they do not read.
- Check your price against the visible alternatives. Price is a factor Amazon names directly.
- Build reviews through legitimate routes. Reviews lift conversion, and conversion is what the ranking model reads.
- Stay in stock. Running out interrupts the sales signal that got you there.
Common mistakes this model explains
A few familiar frustrations make more sense once you separate the two steps.
- Spending on ads for a keyword you do not rank for organically. Ads can buy you a placement, but they do not fix a listing that fails matching.
- Repeating the same phrase in every field. This buys almost nothing and costs you coverage of other phrases.
- Leaving attribute fields blank. These are structured signals, and empty fields make your product harder to match correctly.
- Chasing a single ranking trick. Ranking is a sort of a pool you already qualified for, not a switch you flip.
What this means for a brand new listing
A brand new product has no sales history, so the behavior signals ranking depends on do not exist yet.
Amazon published research on exactly this problem. Rather than leaving new products invisible, Amazon predicts starting values for those behavior signals using the product's attributes, things like brand, product type, and color. In an online test described in that research, this approach increased impressions and shopper engagement for new products.
The practical lesson for a launch is direct. At the start, your attributes are doing the work that sales history will do later, so filling them accurately matters more in week one than it will in month six.
Myths worth ignoring
Three claims come up constantly and none of them hold up.
- The A10 algorithm. Amazon has never announced an algorithm by that name. It is a term the seller community invented. Amazon's current seller guidance does not use A9 or A10 either. Optimize for how search works, not for a nickname.
- A guaranteed honeymoon boost for new listings. There is no Amazon source for a promised window of free visibility. What does exist is the cold start approach above, where Amazon predicts signals from your attributes. That is a head start, not a gift.
- Exact ranking factor percentages. Amazon has never published weights. Any specific split is someone's estimate presented as fact.
Conclusion
Amazon search is two jobs, not one. Matching decides whether you are in the running, and it rewards correct attributes plus coverage of the phrases you want to win. Ranking decides your position, and it rewards the things that make shoppers click and buy.
Diagnose which step is failing before you spend money. If you are invisible for a keyword, fix the listing fields and attributes. If you are visible but low, fix images, price, reviews, and copy. For the keyword coverage side of that work, our Amazon Listing Optimizer handles the field by field build in one run. Next reads: how does Amazon's search algorithm work, how many keywords Amazon allows per listing, and Amazon SEO ranking signals.
References
Frequently asked questions
How does Amazon search actually work?
Amazon search runs in two steps. Step one is matching, where Amazon pulls a pool of products that could answer the search. Step two is ranking, where Amazon puts those products in order. Matching uses a keyword index plus a second system that understands meaning, so a search for 'kids water bottle' can still find a listing that says 'children's flask'. Ranking then sorts the pool using relevance and shopper behavior signals like clicks and purchases.
What is the difference between matching and ranking on Amazon?
Matching decides if your product is even in the running. Ranking decides what position you get. If your listing never matches the search, no amount of price cutting or ad spend will help, because you are not in the pool. This is why keyword coverage and correct attributes matter first, and why conversion work matters second.
Why does my product not show up for a keyword at all?
It usually means you failed the matching step. The words a shopper typed do not appear in your indexed fields, or your product type and attributes tell Amazon your item is something else. Check that the phrase appears once somewhere in your title, bullets, description, or backend search terms, and that your category and attributes are correct.
Does Amazon search only look for exact keywords?
No. Amazon's own research describes a keyword index working alongside a system trained on shopper behavior that understands related words and meaning. That system was built because plain keyword matching misses synonyms, plural and singular forms, and spelling mistakes. You still need the important phrases present, but you do not need to repeat every variation.
What signals does Amazon use to rank products?
Amazon's seller guidance points to relevance, sales performance, competitive pricing, listing quality, images, and account health. Amazon's published research adds that shopper behavior, mainly clicks and purchases, trains the ranking model. Amazon has never published the weights, so treat any blog claiming an exact percentage per factor as guesswork.
How fast does Amazon search reply to a shopper?
Very fast. Amazon's own paper on its search system describes working inside a budget of about 20 milliseconds. That speed limit is one reason Amazon leans on a pre-built index and a compact model rather than reading your full listing from scratch at search time.
Does this mean keyword stuffing still works?
No. Amazon's own guidance says that if you use a key phrase once, you probably do not need to use it again. Because a system that understands meaning sits next to the keyword index, repeating the same phrase adds little and wastes space you could use on a phrase you do not yet cover.
What should I fix first if I want better search results?
Fix matching first, then ranking. Confirm the product type and attributes are right, then confirm every phrase you want to win appears once across your fields. After that, work on the things that lift conversion, such as images, price, reviews, and clear bullets, because that is what moves you up inside the pool you already qualified for.
AI Tools You Can Try
Cover the right keywords in every listing field.
Drop your ASIN. Get a title, bullets, description, and backend keywords built to match more searches and convert more of them.
Try the Amazon Listing Optimizer →