A brand rewrites its listing every quarter. New title, new bullets, a fresh export from a keyword tool pasted into the backend fields. Six months later the organic position on the one term that actually matters has not moved, and nobody can say which of the thirty changes was supposed to move it.
That is the normal outcome, and it is not because the advice was wrong. It is because the work treated every field as equally weighted and every change as equally testable. Neither is true. Below is the sequence we actually work through, ordered so each check only runs once the one before it has passed.
Indexation and ranking are two different failures
Before you ask why a listing ranks badly for a term, establish whether it ranks for that term at all. Being absent from the result set and being buried inside it look identical from the outside and have nothing in common underneath.
Indexation means Amazon’s search engine associates your ASIN with the term. Ranking means where you sit among the ASINs it already associates. If you are not indexed, no bid, no price change and no new hero image will help — the listing is not in the set being sorted, so nothing you do to it gets sorted.
The quick field check is to search the term together with your ASIN on the marketplace you are actually checking — amazon.co.uk, not amazon.com, because the two indexes are separate — and see whether your listing comes back. Treat that as a heuristic rather than a guarantee: Amazon has changed how it handles combined queries more than once, and a result can be returned for reasons other than the index. Where you have Brand Registry, the Search Query Performance report in Brand Analytics answers the question properly. A query you are receiving impressions on is a query you are indexed on, and the report says so with numbers instead of inference.
Where the index actually reads
Not every field carries weight, and at least one carries far less than the effort usually spent on it.
- Title — the strongest single field and the most constrained. Most of the ranking movement you will ever get from copy comes from here.
- Bullet points — indexed, and the field most people overfill. Five bullets doing one job each beat five bullets doing four jobs each.
- Backend search terms — indexed and invisible to the shopper, which makes this the right home for synonyms, regional vocabulary and alternative spellings that would read badly on the page.
- Structured attributes and the browse node — these decide which refinements and filtered result pages you are eligible to appear on at all. An incomplete attribute set quietly removes you from result sets you would have ranked well on, and no amount of copy work brings you back.
- Product description — still in the record even when A+ content covers it on the page, so it is worth writing properly rather than leaving to whatever the last flat file carried.
- A+ content — Amazon’s guidance on whether A+ text is indexed has not been consistent across its own help pages and has changed over time. Treat A+ as a conversion surface first. If it also helps you index, that is upside; it is not a plan.
Take your terms from your own data, not from a volume estimate
Third-party tools model search volume. Search Query Performance reports it. If you are brand registered, that report gives you, per query: impressions, clicks, add-to-carts and purchases for the whole query, plus your own share of each. It is the difference between an estimate of the market and a measurement of your place in it.
Two comparisons in that report are worth more than the volume column. Your click share set against your impression share tells you whether the main image and the price are doing their job. Your purchase share set against your click share tells you whether the listing is doing its job after the click. A term where you take a large impression share and a small purchase share is not a ranking problem at all, and every hour spent treating it as one is wasted.
Here is the arithmetic, using illustrative figures rather than anyone’s account. Say a report shows a listing at 20,000 query impressions in a month, 300 clicks and 24 orders. That is a 1.5% click-through rate and an 8% conversion rate. Lifting the click-through to 3% — a better main image, a price that stops looking wrong next to the listing beside it — adds roughly another 24 orders a month at the same conversion rate, on traffic you already had. Climbing the rankings to achieve the same thing would mean holding several positions higher for a full month. One of those is a fortnight of work. The other is a quarter.
The title: highest leverage, tightest rules
Amazon’s title policy has tightened. A general cap of 200 characters has been rolled out, alongside restrictions on repeated words and on special characters, and individual category style guides set their own shorter limits — some considerably shorter. Check the style guide for your category rather than assuming the general figure applies to you, and check it again the next time you touch the title, because this is one of the rules Amazon has revised repeatedly.
Within whatever room you have:
- Brand, then the term the shopper actually types, then the two or three qualifiers that decide the purchase — size, count, material, compatibility. Not five qualifiers. The ones that decide it.
- Do not repeat a word to rank harder for it. The index does not count occurrences that way, and repetition is among the things the current title rules restrict.
- Read the title truncated. Search results and mobile both cut it. If the qualifier that decides the sale sits at character 150, the shopper who decided at character 60 never saw it.
Bullets written for a four-second skim
Nobody reads five bullets. Shoppers scan the first few words of each and stop as soon as they have what they came for. So lead every bullet with the noun phrase that answers a question and put the evidence behind it. “Fits a standard 60cm oven cavity” is a bullet. “PREMIUM QUALITY CONSTRUCTION” is a shout.
Two checks people skip. First, open your own listing on a phone: on mobile the bullets are frequently collapsed behind a tap and pushed below the A+ modules, which changes what the first screen has to accomplish. Second, spend twenty minutes in your three-star reviews and your returns reasons. That is where the specific objection lives, and one of your five bullets should exist purely to answer it. A bullet that removes a reason not to buy is worth more than a bullet listing a feature nobody was disputing.
Backend search terms: what belongs there
Amazon documents a limit of 250 bytes on the generic keyword field, per marketplace. Bytes, not characters — an accented character can cost more than one. Both the limit and how strictly it is enforced have varied by category and have changed before, so confirm the current figure in your category’s help documentation before you fill the field to the brim.
- Do not repeat words already in the title or bullets. Repeating a term across fields does not compound; it spends bytes you could have used on a term you are not indexed for at all.
- No competitor brand names and no competitor ASINs. Amazon’s terms prohibit it, and it is exactly the sort of thing a rival reports.
- No commas, no repeating your own brand, no subjective claims such as “best” or “number one”. Space-separated single words, and let the engine assemble the phrases.
- Do use it for the words you would never put on the page: regional vocabulary, informal names for the product, and the use case a shopper describes rather than names.
The conversion side is the ranking side
Amazon’s ranking is driven substantially by what its engine expects a shopper to buy after typing a given query. That makes conversion rate a ranking input rather than a separate discipline downstream of one, which is why a listing can be beautifully written and still slide.
- Main image. Pure white background, product filling most of the frame, no added text, badges or borders — these are documented style requirements. It is the highest-leverage single asset you own, because it is the only thing of yours competing in the result grid.
- Buy Box. If you are not holding it, the listing has no add-to-cart button for most shoppers and your ads have nowhere to send them. Fix that before you touch copy.
- Reviews and rating. Not a dial you turn, and not something to treat as a direct ranking lever. They matter because of what they do to conversion rate on the way past.
- Availability. Going out of stock costs position, and regaining it takes longer than losing it did. Inventory planning is an SEO activity whether or not the business files it that way.
What does not move rank
- Stuffing the same term into every available field.
- Hidden keyword blocks at the bottom of the description.
- Rewriting the title every month. Each rewrite destroys the read on the previous one, so you never learn anything and the listing never settles.
- Backend terms in languages the shoppers on that marketplace do not search in.
- Chasing the biggest head term in the category when your listing converts badly on it. You win the impressions and then give the position back.
What to do first
Take your five most commercially important terms — not your fifty. For each one, confirm indexation, pull the Search Query Performance row, then classify the failure as one of four things: not indexed, indexed but few impressions, impressions but few clicks, clicks but few purchases. Those four have four different fixes, and only the first two are SEO in the sense most people mean.
Then change one field at a time and leave it alone for a fortnight. That is slower than a full rewrite, and it is the only version of this work that tells you which change did anything. In most catalogues we look at there is more money sitting in the purchase column than in the impression column — which means the fastest Amazon SEO win is usually not an SEO task at all.