Team reference

How we score Amazon SEO

Every advertised ASIN gets a score out of 10. This explains what the score means, where the numbers come from, and what to do about them.

The question the score answers: if we switched the ads off tomorrow, would this product still sell?

The three things we look at

A listing can fail in three different ways, and the fix is different each time. So we measure all three separately rather than rolling them into one number you can't act on.

What	The question	If it's bad, you fix…
The copy	Is the listing fully built — title, bullets, images, description?	The listing content
The keywords	Do the words customers actually search for appear in the copy?	The wording, not the amount
The results	Do visitors buy, and does it sell without ads?	Price, images, reviews, stock

The words you'll see

Term	Plain meaning
CVR (conversion rate)	Of everyone who opened the listing, what share bought. `units ÷ sessions`. 3% means 97 of 100 visitors left empty-handed. This is the clearest sign the listing is the problem, not the ads — ads decide who arrives, not whether they buy.
Sessions	Visits to the product page. One person browsing twice in a day counts once.
Earned %	The share of sales that did not come from ads. High = the listing stands on its own. Low = we're renting the sales.
Keyword coverage	Of the search terms that actually produced orders, how many appear in the listing copy.
Buy box	The share of time our offer is the one in the "Add to Cart" button. Below 90% usually means a price or stock issue.
ACOS	Ad spend ÷ sales from ads. Lower is better.

The nine checks

Each check states a standard, the observed value, and a verdict — pass warn fail. Nothing is hidden inside a formula.

Check	Standard	Why it matters
Title length	80–200 characters	Under 80 wastes space Amazon would index. Over 200 gets cut off in search results.
Bullets	All 5 used	Amazon gives 5 slots. An unused slot is keyword surface thrown away.
Bullet depth	100+ characters each	A four-word bullet carries almost no searchable text.
Images	6 or more	Below six, shoppers have unanswered questions and bounce.
Description	200+ characters	Empty descriptions are common and cost easy ground.
Keyword coverage	Converting terms present	We know which searches produced orders. If those words aren't in the copy, we're paying ads to cover for the listing.
Earned share	Majority not from ads	Under 30% means the product only sells while we're paying.
Conversion rate	10%+ of sessions	Traffic arrives and doesn't buy — bidding can't fix that.
Buy box	90%+	If we don't hold the buy box, nothing else matters.

How the score is built

pass earns the full weight, warn earns half, fail earns nothing. Weights total 10.

Check	Weight	Why weighted this way
Keyword coverage	2.5	The single thing SEO actually moves.
Earned share	2.0	The whole point: selling without paying.
Conversion rate	1.5	Proves the listing persuades, not just attracts.
Title	1.0	Necessary hygiene, but doing them well doesn't by itself make a product sell.
Bullets	1.0
Images	1.0
Bullet depth	0.5	Smaller contributors.
Description	0.5	Smaller contributors.
Buy box	−1.0	A penalty, never a credit. Losing the buy box suppresses everything else.

Two rules that stop the score lying

A check we can't measure is excluded, not failed. The remaining weights are rescaled. A missing number never masquerades as a bad number.
No traffic means unrated, not a score. Without visits there's no conversion rate and no earned share, so a listing would be graded on copy alone — and a never-sold product could score 10/10. We withhold the score instead. This is why some ASINs show unrated.

Reading a score

Score	Means	Do
8–10	Listing is doing its job	Leave it alone. Spend attention elsewhere.
5–8	Works, with real gaps	Apply the listed fixes when convenient.
Under 5	Ads are carrying it	Fix before spending more — more budget on a weak listing buys expensive traffic that doesn't convert.
unrated	Not enough traffic to judge	Nothing yet. Revisit once it has visits.

Where the numbers come from

Sales & Traffic report — sessions, conversion rate, buy box, total sales per ASIN. Last 30 days.
Catalog — the live title, bullets, description and images, read directly from the listing.
Search-term report — the searches that produced actual orders. This is the keyword signal; we don't guess keywords or use a third-party tool.
Ads data — ad-attributed sales, so earned = total − ads.

Not every winning keyword belongs in the copy

The keyword signal comes from searches that produced actual orders. But a term converting is not the same as a term being right for the product. Auto campaigns match broadly, so someone searching ajwa dates can land on a Safawi listing and buy — that does not make Ajwa the right keyword for a Safawi product.

Every converting term lands in one of three buckets: recommend, needs a human check, or excluded. Four kinds of term are excluded, each shown with its reason on the listing:

Excluded	Example	Why
Different variety or type	`ajwa dates` on a Safawi listing	Misdescribes the product. Attracts shoppers wanting something else, who then don't buy — which lowers conversion and hurts ranking.
Too generic	`sweets`, `produce`	Describes a shelf, not a product. Enormous competition, no purchase intent.
Regulated claim	`organic`, `halal`	Needs certification on file. Using it without one risks suppression.

The "needs a human check" bucket

A hardcoded competitor list can never be complete, so the audit also uses our own catalogue as the vocabulary: if a search term contains a word that appears in none of our listings, it is probably a brand or product we don't carry. Those are flagged for review rather than recommended — it caught biscoff cookies and siete grain free mexican shortbread cookies, which earlier versions recommended outright.

Some listings end up with no recommended keywords. That is a real finding, not a gap: every term that converted was wrong-variety, generic, or an uncertified claim.

Some winning keywords must be ignored

A few search terms that genuinely produce orders are regulated claims — organic and halal among them. Putting them in copy without certification on file risks the listing being suppressed, which costs far more than the extra traffic is worth.

They are removed from every recommendation and listed once, separately, so the decision is deliberate rather than accidental. If we hold certification for a product, tell us and it goes back in.

How to use it

Work top-down — the list is sorted worst-first, weighted by revenue at stake, so the top row is where an hour of work pays most.
Each listing's fixes are written out as exact copy. Paste them; don't rewrite from scratch.
Check the keyword table per listing: a term in the title ranks much stronger than the same term in the description. "weak" means present but not in the title.
A low score with high sales is the most urgent case — real demand held back by a fixable listing.

← Back to the audit
Scores refresh when the audit re-runs. Scope is advertised ASINs. Figures cover a 30-day window.