Answer in brief
Impressions, clicks, orders, buyouts — four numbers read from left to right. What a drop in each one actually means, which platform report answers which question, and the weekly rhythm that catches a problem while it is still cheap.
The short answer: four numbers, read in one direction
Watch four numbers, and always in the same order: impressions, clicks, orders, buyouts. Each one is a subset of the one before it, so a fall starts at exactly one link in the chain and then echoes down every link below it.
That gives you the rule which saves weeks of work. Look for the cause only in the first number that moved. If impressions dropped, the smaller order count underneath is not a second problem — it is the same problem, seen later.
Reports in a seller account carry different names on every platform, but the set is the same everywhere: a listing funnel, search queries, advertising, returns and uncollected orders, and stock. Below, each one is matched to the question it answers.
Then the rhythm. Twenty minutes a week on four numbers for the SKUs that actually move, plus an hour a month on segments, is enough to catch a problem while it still costs an edit to a listing rather than a repack of a shipment.
Impressions: what the first number actually counts
An impression records that your listing reached somebody's screen — in search results, inside a category, on a recommendation shelf, or within an ad block. It is not demand for the product and it is not interest in it. It is access to attention, and nothing more.
So a fall in impressions almost never means the market went off your product. Far more often the results page itself moved: the set of queries you ranked against changed, or competitors took the positions you had been standing in.
The second common source is that the listing stopped being available at all. Stock ran to zero, an edit sent the card back through moderation, or the category and required attributes are filled in such a way that filters no longer return it.
Reading impressions without the season is risky too. A whole category can sag for a week, and then the correct conclusion is to leave the listing alone and compare your own curve against neighbouring listings over exactly the same days.
Clicks and CTR: a percentage without its denominator lies
A click is somebody opening the listing. Clicks divided by impressions is usually called CTR, and it is the number most often misread, because the percentage moves whenever either half of it moves, not only the half on top.
A worked example: your listing starts ranking on a broad query, impressions triple, clicks rise in absolute terms, and CTR falls. That is not decay, it is dilution. A less specific audience arrived, and that is an ordinary thing to happen.
The reverse case is the dangerous one. Impressions sink, CTR climbs, and there are fewer clicks than last week. The percentage looks like a win and the revenue does not. Keep absolute impressions and clicks in the weekly sheet beside the rate.
Few things decide the click itself: the first image, the title, your price next to the neighbours in the results grid, and the delivery label. All of it is visible before the listing opens, so change one element at a time and write down the date.
Orders: conversion on the listing page, and what pulls it down
The order is the first moment a buyer gives something up — time, a slot in the basket, sometimes money up front. This is where the contents of the page start working rather than its cover: description, attributes, sizing, delivery windows.
When clicks arrive and orders do not, the question is narrow: what could the buyer not find out on the page. Usually it is dimensions, materials, what is in the box, compatibility, or return terms — the things nobody is willing to guess about.
The other cause is a gap between the promise and the page. The first photo said one thing, the gallery said another, and the reviews say a third. People leave not because it is expensive but because they stopped knowing what they were buying.
Cutting the price here is the most expensive way to test a hypothesis. It will lift orders, and it will also hide the original cause and take margin that you will struggle to win back without losing your position in the results.
Buyouts: the number that turns an order into money
The fourth number is the share of orders the buyer actually collected and kept. An order that travels to a pickup point and comes back has paid for the logistics in both directions and has paid for nothing else at all.
That is why buyouts cannot be left out of the picture. A listing with strong order conversion and weak buyouts looks healthy in the orders report and loss-making in the money, and the second view is the one closer to the truth.
A fall in buyouts reads differently from a fall in conversion. It nearly always points to a gap between the promise and the object in the hand: size, shade, the weight of a fabric, what was in the box, the state of the packaging on arrival.
The reports for this are returns with stated reasons, plus the text of reviews from the same days. When the same reason repeats across different buyers, it is a defect in the description or in the product, not one unlucky delivery.
Which report answers which question
The listing funnel answers the question of where the break is. It holds impressions, clicks and orders in one place and lets you see within a minute which link the movement started at.
The search-query report answers why the impressions moved. It lists the wordings your listing surfaced for, where it stood on each of them, and what changed in the set of queries itself.
The advertising report answers how much of this was actually bought. Without it, paid and unpaid traffic sit in the same row, and every judgement you make about a listing is built on a blend of two different things.
The returns report and the stock report answer why the money never arrived and whether the listing was available at all. Platforms name them differently and export them differently; the questions behind them do not change.
Comparison windows: week over week, and the same-weekday rule
Compare equal stretches of time. Seven days against the previous seven is the working minimum: it smooths out weekdays and weekends, which behave very differently in most categories.
Month-to-date is a trap. An incomplete period always loses against a complete one, so the report shows a fall where nothing has happened. The same applies to holding one week up against a whole month.
Attribution rules differ from platform to platform. An order may land on the date of the click or the date it was placed, and a return can reduce an earlier period after the fact. Yesterday's figures are therefore always provisional.
The practical consequence is simple. Decide on a closed week, not on the last two days. A sharp number for yesterday is a reason to look more carefully; it is not a reason to rewrite a listing.
Split paid from organic before you draw a conclusion
Paid traffic changes the shape of the funnel. It lifts impressions and often lowers both CTR and conversion, because it brings in people who were looking for something adjacent — and in the combined row the listing looks worse than it really is.
So the first segment in any analysis is by source. The organic funnel tells you how the listing lives on its own; the paid funnel tells you what an extra order costs and whether that cost comes back to you.
The second segment is by SKU rather than by the whole brand. A store average hides your best position and your worst one alike, and decisions are made on individual listings that each carry their own margin and their own price.
The third segment is the one people forget: by variant. Colours and sizes inside a single listing can have very different buyout rates, and the headline number for the listing is then an average of two unlike stories.
The events that explain most sudden drops
Before you look for a complicated cause, check the simple chronology. Sharp changes usually coincide with an event, and that event is almost always already written down somewhere in the seller's own paperwork.
The list is short: an edit to the listing and the re-moderation that follows it, a price change or entry into a promotion, stock hitting zero even for a day, a change of fulfilment scheme, a change of category or required attributes.
A change log with dates is therefore worth more than any dashboard. Without one, the link between what you did and what you got has to be reconstructed from memory, and memory is reliably and systematically wrong about exactly this.
The one-edit-at-a-time rule is boring, and it is the only thing that turns observation into knowledge. Two simultaneous edits produce a result from which no conclusion of any kind can be drawn.
The search-query report: why the impressions moved
This report shows the wordings your listing was surfaced for and the place it held on each of them. It is the only one that explains the movement of the first number rather than simply recording that it happened.
Look for three things in it: queries that appeared, queries that used to carry volume and have now gone, and positions that slid downwards on the wordings that stayed in place.
It is also where the language of your title and attributes should come from — not invented, but the words people actually use for the product. That is cheaper than any guess, and it is testable in the same report a week later.
Do not let it turn into stuffing. A title glued together out of queries hurts the click, because the person in the results grid reads it with their eyes rather than a parser, and picks whatever is clear at first glance.
The second door: your own site, Product markup and readability
A marketplace reports only on what happened inside the marketplace. Demand that reached you through general search, through social platforms, or through a direct visit to your own site does not exist in its reports at all.
That is a reason to keep a product page of your own with schema.org Product markup: price, availability, identifiers and reviews described in a machine-readable form, so that search engines read the page as a product rather than as prose.
Your own site gives you what a seller account never will: the full path a visitor took, your own traffic sources, and a way to measure external demand separately from internal. It is a second set of numbers, independent of the platform.
Then readability. A quick pass against WCAG 2.2 — text contrast, alternatives for images, keyboard operation — removes a slice of the loss between click and order that no report will ever name for you.
The weekly rhythm, and where an outside team pays for itself
The rhythm looks like this. Once a week, pull the four numbers for your moving SKUs over a closed week, find the first link that shifted, open the report that owns it, and make one single change.
Once a month, run the segments: paid against organic, variants inside a listing, reasons for returns, search queries. It is an hour of work, and it is usually the hour that finds what the weekly sheet was only showing you as noise.
If you would rather not learn all of this from scratch, the Demand Audit is $50 and takes 1-2 working days, with one round of revisions on the findings document. It does not include access to your seller account.
Beyond that there is the Marketplace + SEO System at $110 — 3-5 working days, two rounds on listings and copy, photography and paid placement not included — and Managed Demand Growth at $200/mo on a monthly cycle with 30 days notice to stop. Details at /services/marketplaces.
Practical checklist
- Export impressions, clicks, orders and buyouts for every moving SKU over one closed week.
- Compare against the previous seven days rather than against month-to-date.
- Separate paid traffic from organic before judging any conversion rate.
- Open the search-query report whenever it is impressions that moved.
- Match a fall in buyouts against return reasons and reviews from the same days.
- Log every listing edit with its date and change one element at a time.
Questions and answers
Which numbers should I start with if I only have twenty minutes?
Four of them, in one order: impressions, clicks, orders, buyouts. Find the first one that moved against last week and work only on that. Everything below it in the chain is a consequence, not a separate problem.
How often should I open the reports?
Weekly, on a closed week — twenty minutes on four numbers for the SKUs that move. Then an hour a month on segments: paid against organic, variants inside a listing, reasons for returns, and the search-query report.
Impressions fell but orders held. Is that a problem?
It is a reason to open the search-query report, not to rewrite the listing. Often it is untargeted volume leaving: fewer impressions, a tighter audience. Worry if you also lost position on the wordings that used to bring orders.
Buyouts are falling while orders stay flat. Where do I look?
At the gap between the promise and the object in the hand: size, shade, the weight of a fabric, what was in the box, the packaging on arrival. Open the returns report with stated reasons and the reviews from the same days; the phrase that repeats is your answer.
What does it cost to have someone else read these numbers?
The Demand Audit is $50 over 1-2 working days, with one round of revisions on the findings document; access to your seller account is not included. For continuing work, Managed Demand Growth is $200/mo on a monthly cycle with 30 days notice to stop, and advertising spend on the marketplace is not included.