Key takeaways:
- Amazon search ranking has two halves. Relevance decides whether your ASIN is eligible to show for a query. Performance decides where it sits among the eligible set.
- Relevance comes from text: title, bullets, description, A+ content, backend search terms, category and attributes. If a phrase appears nowhere, you usually will not rank for it.
- Performance is mostly sales per impression for that specific query, plus click-through, conversion rate, price position, availability and delivery speed.
- You control relevance directly and in a day. You influence performance indirectly, over weeks, by making the page convert better and by buying early traffic.
- Rank is per keyword, not per product. You can be page 1 for one phrase and page 6 for a near-synonym on the same ASIN.
The short answer
Amazon's search engine answers two questions in order. First: which ASINs are even candidates for this shopper's query? That is relevance, and it is driven by the words on and behind your listing. Second: of those candidates, which ones are most likely to produce a sale right now? That is performance, and it is driven by how your listing has behaved for that query in the recent past.
Google ranks documents on whether they answer a question. Amazon ranks products on whether they are likely to be bought. That single difference explains almost every ranking factor below.
Why the two halves matter separately
Sellers lose months by treating ranking as one thing. It is not. The fix for "we do not appear at all" is completely different from the fix for "we appear on page 4".
If you are not indexed for a phrase, no amount of advertising will fix it. Ads can buy a placement above the organic results, but they do not create organic eligibility for a term your listing never mentions.
If you are indexed but ranking low, rewriting the title again will not help much either. The phrase is already there. What is missing is evidence that shoppers who search that phrase click your result and buy.
So the first diagnostic question is always: indexed or not? Everything after that is a performance problem.
Read this flowchart as text
- Shopper types a query
- Amazon builds the candidate set from indexed ASINs
- Is your ASIN indexed for this phrase?
- Not eligible Fix the text first
- Eligible Now ranked against peers
- Performance signals read CTR, conversion, sales velocity, price, availability, delivery
- Does your page convert better than peers on this query?
- Rank rises over following weeks
- Rank holds or slips Fix conversion or traffic quality
Half one: relevance, and the fields that carry it
Relevance is text matching plus structured data. Amazon reads several places, and they are not equal.
Title. The heaviest text field and the one shoppers read. Lead with the phrase a buyer would actually type, then the qualifiers that separate you from near-identical products. Most categories cap titles around 200 characters, but several categories enforce shorter limits and style rules of their own. Check the style guide for your category in Seller Central before you write, because a title that violates category rules can get the listing suppressed.
Bullets. Carry secondary phrases and the attributes that matter in your category: size, material, compatibility, quantity, use case. Bullets also do the conversion job, so they cannot be a keyword list. Write them for a person and the keywords land naturally.
Backend search terms. The hidden field, limited to 250 bytes per listing in most marketplaces. It exists for spellings, synonyms and phrasings you do not want in visible copy. Do not repeat words already in your title, do not add competitor brand names, do not add subjective claims. Our full treatment is here:
Category, browse node and attributes. A large share of searches are refined with the left-hand filters. If your size, color, material or compatibility attributes are blank, you drop out of those refined results even when your text is perfect. This is the most common silent relevance leak we find in audits.
A+ Content and description. A+ Content replaces the product description when it is live. The A+ text is not weighted like title and bullets, and historically image alt text in A+ has been the indexable part rather than body copy. Treat A+ as a conversion asset first. Any indexing you get is a bonus.
Rufus and COSMO. Amazon's shopping assistant, Rufus, and the COSMO layer behind it read the page differently from classic keyword matching. They look for context: who the product is for, what problem it solves, what it fits, what it does not do. Listings written as flat keyword strings have less to work with. Listings that answer real questions in plain sentences give both the AI layer and the human shopper something to use.
Half two: performance, and what it actually counts
Once you are in the candidate set, Amazon is estimating one thing: if we show this ASIN for this query, how likely is a purchase?
The signals that feed that estimate, roughly in order of weight:
Sales velocity for that specific query. Not total sales. Sales attributed to that search term, in the recent window. This is why a best-seller can still rank badly on a phrase it has never converted on.
Conversion rate from that query. Sessions that turn into orders. You can see the ASIN-level version as unit session percentage in the Business Reports. For query-level data you need Search Query Performance in Brand Analytics, which shows impressions, clicks and purchases by search term for your ASINs against the category total.
Click-through rate. How often shoppers pick your result over the others on the page. Main image, price, rating, review count and the first 60 characters of your title do nearly all of this work.
Price position. Amazon does not publish a price factor, but in practice a product priced well above comparable options converts less, and lower conversion drags rank down. Price acts on rank through conversion.
Availability and delivery speed. Out of stock means no rank. Prime-eligible, fast delivery converts better than a long shipping window, so it feeds the same loop.
Reviews and rating. They move click-through and conversion rather than acting as a separate ranking lever. Build them only through Amazon's own routes: the Request a Review button, Vine for eligible enrolments, and genuine follow-up inside Amazon's communication rules. Incentivised or solicited positive reviews break Amazon's policy and risk the account.
Recency. The windows Amazon reads are short. Sales from last week count far more than sales from last quarter. That is why rank falls fast after a stockout and why it has to be rebuilt rather than reclaimed.
How paid and organic feed each other
Advertising does not directly raise organic rank. A sale from a Sponsored Products click raises organic rank exactly like any other sale on that query, because it is a sale on that query.
That is the whole mechanism. You pay to manufacture early sales velocity on a term where you have no history, the term starts to rank organically, and then you can reduce the ad support. Launch ACoS is high on purpose for this reason, and it should fall as organic takes over.
The loop only works if the page converts. Buying clicks to a page with a 4 percent conversion rate burns budget and teaches Amazon that your ASIN is a poor answer for that query. Fix the page first, then buy traffic.
Track the shift with TACoS rather than ACoS alone. Falling TACoS at steady revenue is the clearest sign organic rank is carrying more of the load.
What you control, and how fast
| Lever | Half it affects | Who controls it | Time to show |
|---|---|---|---|
| Title, bullets, backend terms | Relevance | You, fully | Index in 24–72 hours |
| Category and attributes | Relevance | You, fully | Days |
| Main image and price | Performance (CTR) | You, fully | 1–2 weeks of data |
| A+ and gallery images | Performance (CVR) | You, fully | 2–4 weeks of data |
| Ad spend on the target term | Performance (velocity) | You, fully | 2–6 weeks to move rank |
| In-stock and delivery speed | Performance | You, mostly | Immediate downside |
| Reviews and rating | Performance | You, within policy | Months |
| Competitor pricing and entries | Performance | Not yours | Ongoing |
| Seasonality and demand shifts | Both | Not yours | Ongoing |
The honest summary: you control relevance completely and quickly. You control the inputs to performance, but Amazon controls the output, and it reads the market as well as your listing.
How to check whether you are indexed
Do not guess. Check.
- Take the exact phrase. Search it on the marketplace with your ASIN appended, for example
blue yoga mat B0XXXXXXXX. If your product returns, you are indexed for that phrase. If nothing returns, you are not. This is a rough check and it can be noisy, so repeat it. - Cross-check in Search Query Performance. If the term shows impressions for your ASIN, you are indexed and being served.
- If the term appears in your copy but returns nothing after 72 hours, look for the usual blockers: a restricted or policy-flagged word, a suppressed listing, a variation where the parent holds the copy, or bytes over the 250-byte backend limit causing the field to truncate.
Re-verify 48 hours after every copy change. Indexing is not permanent, and a bulk upload that drops a field will quietly remove terms you had.
Rank is per keyword, and per marketplace
Two things sellers forget.
First, there is no single "rank" for a product. You hold a position for each query, and those positions move independently. A rank tracker that reports one number is hiding the picture. Pick 15 to 30 target phrases, split them into head terms, mid-tail and long-tail, and track each weekly.
Second, relevance does not travel. A listing translated badly into German indexes on the wrong words, and the performance history from the US does not carry over. Each marketplace builds its own relevance and its own velocity history from zero.
Common mistakes
Chasing the biggest keyword first. The head term in your category is defended by ASINs with years of sales history on it. You will spend a lot and move slowly. Start on attainable mid-tail phrases where the top results have modest review counts, win those, then climb.
Repeating keywords everywhere. Amazon only needs a term once to index it. Repeating it in title, bullets, description and backend wastes space you could use on a different phrase, and it reads badly to shoppers, which costs conversion.
Rewriting the listing every three weeks. Each rewrite resets the performance data Amazon has on that page. Change one variable, wait for enough sessions, then decide. Manage Your Experiments is the clean way to do this on eligible brands.
Treating a stockout as recoverable. Rank drops within days of going out of stock and does not come back on its own. Plan the gap: lower the ad spend rather than killing it, keep at least one child variation sellable if you can, and budget for a rank rebuild on restock.
Stuffing competitor brand names in the backend. It does not index reliably and it breaks Amazon's terms. It is one of the fastest ways to get a listing pulled.
Ignoring attributes. Half the relevance loss we find is not in the copy at all. It is in blank structured fields that keep the ASIN out of filtered results.
How WeSpark does it
We treat relevance and performance as two separate work streams with two separate owners, running in parallel.
In practice: the keyword work in B3 builds a tiered list from your own Search Query Performance data, competitor gaps and category language, then verifies indexing 48 hours after upload and tracks it weekly. C6 picks the short list of phrases where page 1 is actually attainable for your review count and price point, and runs the paid-to-organic push on those. D19 is the playbook we run when a term you already held starts slipping, which is usually stock, a price change, a new competitor or a rating drop rather than anything you did to the copy.
What we will not do is promise a position. Rank depends on what competitors do, and nobody controls that. What we commit to is the process and the weekly visibility into where each tracked term sits.
If you want the current state of your relevance and performance before committing to anything, the audit covers both.
Frequently asked questions
How long does it take to rank on a new keyword?
For a term with reasonable volume and a page that converts, expect 4 to 8 weeks of consistent sales on that query before organic position stabilises. Indexing happens in 24 to 72 hours. The gap between those two numbers is the performance half doing its work, and it cannot be rushed much beyond what traffic and conversion allow.
Do Sponsored Products ads improve organic rank?
Indirectly. The ad itself is not a ranking signal, but the sale it produces on that search term is, and Amazon counts it the same as an organic sale. That is why ads are the standard tool for building velocity on a new keyword. It only works if the page converts; otherwise you are paying for clicks that teach Amazon the wrong thing.
Why does my product rank for one phrase and not a near-identical one?
Because rank is calculated per query. A near-synonym is a different query with its own candidate set and its own sales history. You may be indexed for both but have conversion data for only one. Add the missing phrasing to your copy or backend field, then build velocity on it the same way you built the first.
Does the product description still matter for ranking?
It matters less than title, bullets and backend terms, and on most brand-registered listings A+ Content replaces it visually anyway. Write the description for the shoppers who see it, usually on mobile or where A+ has not loaded. Do not rely on it to carry keywords that are not elsewhere.
How many keywords should I target per ASIN?
Usually 15 to 30 tracked phrases per ASIN, split into a handful of head terms you are building toward, 10 to 15 mid-tail terms you are actively fighting for, and a long tail you expect to pick up with no direct effort. More than that and you cannot act on the data. Fewer and you miss where demand is shifting.
What breaks ranking fastest?
Going out of stock. Rank falls within days and recovery after restock typically takes two to three weeks of rebuilt velocity. After that: a rating drop below roughly 4.0, a suppressed listing, losing the Buy Box to a hijacker, or a price rise that pushes you out of the comparable set. All four show up as a conversion fall before they show up as a rank fall, which is why we watch conversion weekly.