Why did my sales on Amazon drop? Finding the actual reason

Most advice on this question is a list. Check the Buy Box. Check your price. Check inventory. Check whether a promotion ended. Check whether ads paused. Check for a listing suppression. Check seasonality. Ten causes, and you go and look at each one.

The list is not wrong. It is just the slow way round. A sales drop has one of a small number of shapes, and each shape leaves a signature in numbers you already have. Read the signature first and the list collapses to one or two candidates before you open a single report.

This article walks that order: decompose the change, find where it lives, check the connected evidence, then say only what the evidence supports. It is the order MixShift Intelligence runs automatically, and it is also a perfectly good order to run by hand.

Step 1: Traffic, conversion, price, or mix

Ordered sales is a product: sessions, times the share of sessions that buy, times the average price of what they buy, summed across everything you sell. So a change in ordered sales is always some combination of four movements:

Traffic: fewer or more sessions reached your listings.
Conversion: a different share of those sessions bought.
Price: what they paid per unit changed.
Mix: the composition of what sold changed, so the total moved even where no single item's own rate did.

A bridge splits the headline change into exactly those four legs, and the legs sum to the headline with nothing left over. That footing matters more than it sounds: it means you cannot end up with an explanation that accounts for a drop of one size while the drop was actually another. See Reading a bridge for how to read the legs and the caveats.

The two-question version, if you are doing this by hand from Business Reports:

Sessions fell, conversion held. Traffic problem. Something stopped people arriving: advertising, organic visibility, a suppressed listing, or demand itself.
Sessions held, conversion fell. Listing problem. People arrived and did not buy: price, Buy Box, availability, content, or reviews.

Price and mix are the two that a hand read usually misses. A price leg tells you whether the average paid per unit moved, which is different from whether you changed your price. A mix leg tells you whether the drop is real or an artifact of which products sold, which is the difference between a trend and a composition shift.

Step 2: Whole account, or one Item Group

Once you know which leg carried the change, the next question is where. An account-level number is a sum, and a sum can fall because everything slipped a little or because one thing fell off a cliff. Those have different causes and different fixes.

Walk the hierarchy: account, then Item Groups (the sets of products a team manages together, a line or a pack-size family), then individual ASINs. Two patterns are worth recognizing on sight:

Proportional across Item Groups. Everything moved by roughly the same share. That is consistent with something external and account-wide: seasonality, a marketplace-wide demand shift, a change to advertising at the account level, or a data-window mismatch.
Concentrated in one Item Group or one ASIN. The account number is being dragged by a single place. Now the diagnosis is about that place, and the rest of the account is a distraction.

Intelligence groups by ASIN, Item Group, or sub-brand and reports the contribution of each, so the concentration is visible in the result rather than something you infer from scrolling a grid.

Step 3: Check the connected evidence

You now know which leg moved and where. The remaining question is what else moved with it. This is the part the checklist was trying to get at, done in the right order: instead of checking everything, you check the evidence that would confirm or rule out the one or two causes consistent with the shape you found.

What you found	What that shape is consistent with	What confirms or rules it out
Conversion leg down, sessions held, one or a few ASINs	Buy Box loss, or a competitor undercut	Buy Box percentage over the period; your price against the winning offer
Price leg down, conversion up, same ASINs	A price cut or a promotion running	ASP series; a promotion caveat on the result
Price leg up, conversion down at a date boundary	A promotion ended, or a price increase landed	ASP series at the boundary; promotion caveat
Units and conversion down on specific ASINs	Stock-out or thin availability	Sellable inventory and weeks of cover; the Lost Sales estimate for the window
Traffic leg down, concentrated in advertised ASINs	Advertising paused, budgets cut, or bids lowered	Spend, impressions and clicks in the Advertising Bridge; attributed share in the Combined Read
Traffic collapses on one ASIN while inventory is fine and rising	A suppressed or de-indexed listing	Sessions near zero; inventory growing because inbound receipts exceed outbound shipments
Traffic down, advertising unchanged	Organic visibility loss	Not something the bridge measures; a traffic decline with ads held steady is consistent with it, and you confirm it outside the result
Proportional decline across every Item Group	Seasonality or marketplace demand	The year-over-year leg of the Monthly Read; same shape a year ago
Mix leg carries the change, rates flat	Composition shift, not a performance change	The hierarchy walk shows which items gained or lost share
Headline looks dramatic, legs do not add up to a story	A dark period, matched window, or restatement	The caveats on the result; read those before anything else

Two of those rows deserve a closer look because they are the ones teams most often get wrong.

The suppressed listing. The signature is strange enough to be diagnostic on its own: traffic on a single ASIN falls to almost nothing while inventory is healthy and actually growing, because units keep arriving and nothing is shipping out. Ad spend and clicks fall with the traffic, because there is no listing to send them to. Nothing in the sales data explains it, which is the point. The data narrows the question to that one ASIN, and the answer comes from business context: the listing had been flagged. One team found exactly this pattern in a monthly review; the report surfaced the outlier, the Workspace brought the inventory, traffic and advertising evidence together, and the team recognized a listing-suppression issue they were already working. The bridge did not name the cause. It made the cause the only thing left to check.

The caveat that is not a cause. A nine-day gap in the data, two periods of different lengths compared as if they were the same, a prior month restated after the fact: any of these will produce a headline change that looks like a business event and is not one. The engine raises these as caveats on the result. Read them first. A dramatic drop that turns into a modest, like-for-like move once the dark period is accounted for was never a diagnosis problem.

Step 4: Say only what the evidence supports

This is the step that separates a defensible explanation from a plausible one.

The bridge tells you what moved and where, exactly. The evidence tells you what else moved alongside it. Neither of those, on its own, proves a cause. A conversion drop that coincides with losing the Buy Box is consistent with the Buy Box loss having driven it; it is not proof, because other things may have changed in the same window. Most of the time the association is strong enough to act on. It is still an association, and a report that writes "caused by" where the evidence supports "consistent with" has quietly promoted a real number into a conclusion it cannot carry.

Intelligence enforces this boundary in its own output. A driver is only named as a driver when it clears a dominance threshold in the decomposition, and the language stays at "consistent with" rather than "caused by." Where the data cannot settle the question, the result says what to check next instead of guessing. If you are doing this by hand, hold yourself to the same standard: it is the difference between a monthly review that ends in a decision and one that ends in an argument about whose number is right.

Doing this by hand, or letting the service do it

Everything above can be done from Business Reports, an inventory report, and the advertising console, and for a single ASIN on a single occasion it is worth knowing how. The cost is time, and the risk is the step you skip because you already have a theory.

The Operational Bridge runs steps one and two for the whole account in one request: the four legs footing to the total, the contribution of every Item Group and ASIN, and the caveats. The Advertising Bridge and Combined Read supply the advertising side of step three. Add evidence to the request and the result also carries the measured facts and candidate explanations behind it, each attached to its own evidence. In the Intelligence Workspace you drill from the account result to the Item Group to the ASIN and inspect the connected evidence at each level, which is where step three becomes a five-minute task instead of an afternoon. Monthly Report Max does the discovery for you across the whole period and hands you the outliers worth investigating.

What none of them do is skip step four. The result states what the evidence supports. The judgment about what to do next is still yours, and it is a better judgment for starting from a number that reconciles.