Revenue is down 20% on the month. Sessions look normal. Nothing in Account Health is red. And the instinct, almost universally, is to push PPC bids up until the number comes back.
That instinct is usually wrong, and it’s expensive. If the cause is a conversion problem or lost organic position, higher bids buy you more traffic into a listing that has already stopped converting, or paper over a rank slide that keeps widening underneath. You spend more to stand still, your TACoS climbs, and the underlying problem gets a month older.
Diagnosis first. It takes about twenty minutes and it’s the same sequence every time.
Step one: is this a traffic problem or a conversion problem?
Two numbers answer it, and Amazon states the logic itself in Seller Central: if sessions are down but conversion rate is stable, you have a traffic problem. If sessions are steady but conversion is dropping, you have a listing or pricing issue.
Open Business Reports, go to Detail Page Sales and Traffic by Child ASIN, and pull sessions and unit session percentage over a rolling 30 to 60 days. Compare against the same period last month and the same period last year, so a seasonal dip doesn’t get diagnosed as a crisis.
Do this at child ASIN level, not account level. Account-level averages hide the thing you’re looking for — one collapsing SKU inside a stable catalogue disappears into the mean.
What does each pattern mean?
| Sessions | Unit session % | What it is | Where to look |
|---|---|---|---|
| Flat | Falling | Conversion problem | Price, images, reviews, Buy Box, a competitor’s new offer |
| Falling | Flat | Visibility problem | Organic rank, indexing, ad budget, impressions |
| Falling | Falling | Something structural | Suppression, variation split, category change, account issue |
| Flat | Flat, revenue down | Mix or price problem | Discounting, unit price, which SKU is selling |
That last row catches people out. If both metrics hold and revenue still falls, you haven’t lost customers. You’ve changed what they’re buying or what they’re paying: a coupon running longer than intended, a promotion on the wrong variant, or the £34 SKU quietly outselling the £52 one.
Is it a cliff or a bleed?
The shape of the decline tells you which family of causes to investigate, and it’s the single most useful thing to establish early.
A cliff is a drop of 30% or more inside a day or two, with a clear before and after. Something broke. Technical causes almost always.
A bleed is 3 to 5% a week for six weeks, no obvious start date, and by the time anyone raises it in a meeting the account is down a third. This is nearly always rank decay, and it’s the one that does the real damage, because nothing alerts you to it.
Seller Central will not tell you when a product slips from position three to position fourteen. There’s no notification, no flag, no email. The traffic simply stops arriving, and unless someone is tracking rank deliberately, the first signal is a revenue number that has already moved.
Cliff-edge causes: check these in order
Buy Box loss. Check Featured Offer percentage in Business Reports. It’s the fastest possible explanation and the most commonly missed, because sellers assume the Buy Box only moves on price. It doesn’t. Late shipment rate, inventory availability, fulfilment performance and response metrics all affect rotation, and a brief fulfilment wobble can reduce your share without removing eligibility outright. If Featured Offer percentage has fallen, stop reading and fix that.
Suppression. Manage All Inventory, filter by suppressed listings. Missing required attributes, image compliance, a policy flag. The listing stays live enough to look fine at a glance while ranking nowhere.
Stockout, including a brief one. A stockout costs more than the days you were out. It costs the organic position you spend the following weeks rebuilding. And low inventory hurts before you run out. Amazon deprioritises listings with weak inventory signals, so days-of-cover matters even while units remain available.
Variation split. If a parent-child relationship breaks, the reviews and sales history that were carrying the child disappear from view. Check the parent still holds all its children.
Price move. Yours or a competitor’s. Also check whether a coupon or deal ended on exactly the day the graph turns.
Each takes about two minutes to rule out. Do all five before considering anything more sophisticated.
The six 2026 causes most checklists still miss
The standard advice above hasn’t changed in years. These have, and they’re where the unexplained cases are ending up right now.
1. Amazon rewrote your title
On 27 July 2026, Amazon cut product titles to 75 characters and began rewriting non-compliant ones with AI, gradually, listing by listing. If you weren’t in Brand Registry you got no review window. If you were and let the 14 days lapse, Amazon’s version went live.
Those rewrites are generated from your existing listing content, and they can drop a keyword that was holding organic position. The traffic loss shows up weeks later as a slow bleed with no apparent cause.
Check your titles against what they said in June. If they’ve changed and nobody on your team changed them, you’ve found it. The full breakdown of the 75-character rule and how to rebuild a title without losing rank is here.
2. Amazon changed other listing content too
Title rewrites are the visible case, but Amazon routinely tests and modifies listing content with automated tools — bullet points removed, categories reassigned, images suppressed. Any of these can damage keyword indexing or conversion before you notice.
Worth building into a weekly habit: pull your live listing content and diff it against your master copy. It takes minutes and catches changes nobody told you about.
3. Discovery moved to Alexa for Shopping
Amazon retired the standalone Rufus assistant on 13 May 2026 and replaced it with Alexa for Shopping. The difference matters more than a rename suggests. Rufus sat in its own chat window that shoppers had to open. Alexa for Shopping is built into the main search bar, shows AI overviews above search results, and puts side-by-side product comparisons inside the results page. It’s the default layer for signed-in US customers, with no Prime membership or Echo device required.
That is a structural change in how people find products, not a feature release, and expect 30 to 60 days of volatility in PPC metrics while discovery patterns settle. If your decline started in the weeks after mid-May and none of the technical checks explain it, this belongs on the list of candidates.
4. Impressions fell before sessions did
This one is genuinely diagnostic. If impressions decline ahead of sessions, you’re looking at partial suppression or lost keyword indexing rather than a demand problem. Demand shifts move both together. Indexing problems move impressions first.
Most sellers only watch sessions, which is why this gets caught late.
5. One bad ASIN is dragging the others
Late shipments and complaints tied to a single SKU can pull performance down across a catalogue. If the decline is broad and shallow rather than concentrated, look for the ASIN with the worst fulfilment history rather than assuming a market-wide cause.
6. Account Health is green and still hiding something
A clean dashboard doesn’t rule out soft suppressions or minor policy risks that never generate a formal alert but still affect ranking and Buy Box eligibility. Green means no open violations. It doesn’t mean no friction.
How do I confirm it’s rank decay?
If sessions are falling, conversion is holding, and none of the technical checks explain it, you’re almost certainly looking at lost organic position. Now prove it, because the fix depends on knowing which terms went.
Establish the baseline. Which keywords were actually driving your organic sales before the decline? Search Query Performance in Brand Analytics gives you impressions, clicks and purchases by query for your ASIN. That’s the list that matters, not the one in your listing.
Check current position on those terms. If you have rank history, compare it. If you don’t, start recording today. You can’t diagnose a slide you never measured, and this is usually the moment that lesson gets learned.
Find out who took the position. Run a reverse ASIN lookup on the products now sitting where you used to. It tells you which terms they’re winning and gives you a reasonable guess at how — a new variant, an aggressive exact-match campaign, external traffic, or a price move that lifted their conversion rate above yours.
That last part is what turns a diagnosis into a plan. “We lost rank” isn’t actionable. “We lost positions three through seven on our two highest-converting terms to a competitor who launched a 3-pack at £4 under us” is.
What not to do while you’re diagnosing
Don’t raise bids to hide it. You’ll spend more per order and lose the signal that told you something was wrong.
Don’t change five things at once. If revenue recovers you’ll have no idea which change did it, and you’ll carry all five into the next launch as folklore.
Don’t rewrite the listing on instinct. If conversion is holding, the listing isn’t the problem, and rewriting it risks dropping a term that’s still working.
Don’t accept “it’s just the market” without checking. Sometimes it genuinely is: a category-wide seasonal shift, or a demand change. But that’s a conclusion you reach after ruling out the alternatives, not the first explanation you accept because it requires no work.
The short version
Split traffic from conversion. Establish whether it’s a cliff or a bleed. Rule out the five technical causes in ten minutes. Then work the 2026 list. A meaningful share of currently unexplained declines trace back to an Amazon-generated title change or the discovery shift to Alexa for Shopping, and neither appears on a checklist written before this year.
Most importantly: start tracking rank now if you aren’t. Every other metric in this article is available retrospectively. Organic position is the one you can only measure going forward, and it’s the one that explains the declines nobody can account for.
FAQ
Why are my Amazon sessions the same but sales down? Steady sessions with falling revenue points to a conversion or price problem rather than a visibility one. Check unit session percentage first — if it fell, look at price, images, reviews and Featured Offer share. If it held, check whether your product mix or unit price changed, since a coupon or a shift toward a cheaper variant lowers revenue without moving either traffic metric.
What is a good unit session percentage on Amazon? It varies far too much by category for a single benchmark to be useful. What matters is your own trend: a 4-point fall against your 60-day average is a signal regardless of whether the absolute number is 8% or 25%.
How do I know if my Amazon listing is suppressed? Manage All Inventory, filter by suppressed listings. Also compare impressions against sessions — impressions falling ahead of sessions suggests partial suppression or lost keyword indexing rather than reduced demand.
Does losing the Buy Box stop my ads working? Largely, yes. Sponsored Products ads generally won’t serve for an offer that doesn’t hold the Featured Offer, so a Buy Box loss shows up as an abrupt ad performance collapse alongside the organic one. Check Featured Offer percentage before investigating campaigns.
How long does it take to recover organic rank after a stockout? There’s no published figure, and anyone quoting a specific number of weeks is estimating. What’s consistent is that recovery takes longer than the outage did, and that a short stockout on a high-velocity SKU costs more rank than a long one on a slow mover.
Sami Sultan runs Ecom Enable, an Amazon growth partner for DTC brands. Policy and platform details verified 27 August 2026 — Amazon has changed title rules, fee structures and its shopping assistant inside the last twelve months, so check Seller Central before acting on any specific figure here.
Sources: Amazon Seller Central Business Reports guidance and forum posts on diagnosing sales declines; Canopy Management on the Rufus to Alexa for Shopping transition; SentryKit on non-obvious causes of sales decline.