Seller Central: The Reports Worth Reading

Seller Central has more screens than any one person needs. Most of them exist for a case you do not have, and the front page is a summary designed to reassure rather than to be acted on.

The working set is small. These are the ones that change what somebody does next, and roughly how often each is worth opening.

Account health. Not for the score, for whether anything new has appeared. A policy notice caught the day it lands is a reply; caught three weeks later it is an appeal. This is the whole argument for the daily look, and it is covered properly in what account health is measuring.

Anything unfulfillable or stranded. Stock that exists and cannot be sold, usually for a listing reason rather than a physical one. It sits there silently and it is almost always a five minute fix.

The business report, by ASIN. Sessions, page views, unit session percentage and buy box percentage per product. This is the single most useful screen in the account, because it separates the three ways a product can fail: nobody arrived, they arrived and did not buy, or they would have bought and you did not have the button.

The search terms report. What people typed to find you, and what you paid for it. Read alongside the guide to reading it, because the columns invite a conclusion that is often the opposite of what they say.

Inventory age. What is approaching the long term storage bands, which is money leaking rather than a metric.

The business report is most useful at the child ASIN level, the individual size or colour, because a parent listing averages a strong variation and a weak one into a number that describes neither. Read this week against the previous four, not today against yesterday, and read each product against itself rather than against the catalogue.

Each of the three failures has a column. Sessions falling with conversion steady means fewer people are arriving: rank has slipped, a campaign has stopped, or a season has turned. Sessions steady with unit session percentage falling means people arrive and leave: a price change, a new competitor on the page, a critical review now sitting at the top, a main image that changed. Unit session percentage steady with the featured offer percentage falling means shoppers would buy and your offer is not the one being shown: another seller's price, your stock, or your delivery promise.

That last column was called the buy box percentage for years and now carries the featured offer name, which is worth knowing before searching an old guide for it. For a brand that is the only seller of its own products, anything short of the top of that column is a problem with the offer, not with demand.

The business report says how a product is doing. The advertising reports say what the spend is doing about it, and three of them answer questions the business report cannot.

The placement report. Where the ads appeared: the top of the search results, the rest of the search results, or other products' pages. The same keyword can be profitable at the top of search and wasteful on product pages, and the placement report is where that shows. It is the report to read before changing a placement adjustment.

The advertised product report. Spend and sales by the product advertised, rather than by keyword. It shows which products the budget actually goes to, which is often not the split anybody planned, and it is the quickest way to find a product eating budget without the sales to match.

The purchased product report. Sales of products other than the one advertised, after a shopper clicked an ad. A brand whose ads sell its other products more than the advertised one learns something about which product shoppers really want.

Read the advertising reports in the same weekly session as the business report, so a change in one can be checked against the other while both are fresh. How the advertising figures relate to the account's own is set out in ACoS, TACoS and ROAS, and the weekly habit for the campaigns themselves in a weekly PPC routine.

The payments report. The one place the fees are itemised rather than netted. A category that looks profitable in a spreadsheet and thin here is a spreadsheet with the wrong fee assumptions in it, which is why the fee breakdown is worth a read once a year.

Returns. Not the rate, the reasons. A rising return rate on one product with the same reason attached is a listing problem or a product problem, and both are fixable once you know which.

Brand analytics, if you are enrolled. Search term share and repeat purchase behaviour: the only view of the wider market Amazon gives a seller.

Put together, the weekly reading takes about half an hour once it is a habit, and it works best in the same order every time. Account health first, in two minutes: anything new, anything with a deadline. Then the business report by child ASIN, looking only for products that moved more than their normal week to week swing, in sessions, in conversion or in the featured offer.

Then the search terms report for the campaigns carrying most of the spend: what to add and what to stop paying for, as negative keywords, the cheapest work in Amazon PPC sets out. Then inventory: what needs reordering before it runs out, and what is getting old enough to start costing storage.

The last five minutes are the part most routines skip: writing down one action for each thing flagged, and who will do it. A report looked at without an action written next to it has been looked at, not read. Last week's list of actions is also the first thing to check the following week, because whether they were done says more about the account than any single number in it.

Sooner or later two reports give different figures for what looks like the same thing, and somebody decides one of them is wrong. Usually neither is. They are counting different things, and knowing which is which saves a week of chasing.

Sales in the business report against the payments report. The business report counts orders when they are placed, at the price ordered, including orders later cancelled or refunded. The payments report counts money when it moves, after refunds and fees. A month with heavy returns looks better in the first than in the second, and the second is the one the bank agrees with.

Advertising sales against the business report. The advertising reports credit a sale to an ad when it falls within the attribution window named in the column heading, counted from the click, and they include sales of other products from the same brand. The business report counts every sale, by the day it was ordered. Advertising sales can never be read as a share of the business report's figure for one product without allowing for both.

Recent days against the same days a week later. Reports finalise on a delay, so the last two days of any window are provisional and usually rise as late data arrives. A comparison that ends yesterday will always look like a dip.

The rule that settles most arguments is to decide which report answers which question, write it down, and use that one every time. A team that quotes a different report each week will never agree on whether anything changed.

Orders today against orders yesterday, which is a number that moves for reasons nobody controls. It is fine as a heartbeat and it is not a report.

The trap is treating a bad morning as a signal and changing a price or a bid because of it. Daily variance on most catalogues is wide enough to swallow any change you would make in response.

Seller Central will not tell you what your listing looks like against the competition, what a shopper saw before they chose somebody else, or whether your photography is the reason. Those are answered by looking at the page rather than the report.

It also has no view of anything off Amazon. If you sell elsewhere, the comparison lives in your own numbers rather than in any one platform's console, which is the argument for planning across channels rather than per channel.

A number that moves is only useful if somebody can say why. Most of the time the reason is something the business did itself, and most of the time nobody wrote it down, so the reports show the effect and the cause is lost.

One line for every change. Keep a simple log, one line per change, with the date, the product, what changed and who changed it: a price, a main image, a title, a coupon, a campaign paused, a bid raised, a shipment that arrived late. It takes a minute per change and turns every later question into a lookup.

Read the reports against it. When conversion falls on a product, the log shows whether a price rose that week or a new image went live. When sessions jump, it shows whether a campaign started or a deal ran. Without it, the same movement gets three explanations in three meetings.

Keep your own copy of the numbers. Pull the weekly figures for each child ASIN into the same file as the log: sessions, unit session percentage, the featured offer percentage, advertising spend and sales. Amazon's own history does not run back far enough in every report for a year on year comparison, and a copy kept weekly does.

Change one thing at a time. A week in which the price, the image and the bids all changed is a week nobody can learn from. Spacing changes out, product by product, is what lets the reports say which change worked. A proper test of one change is in testing a main image with Manage Your Experiments.

In a business with more than one person on the account, the reports are better shared out than read by everybody. Whoever runs the advertising owns the search terms report and the advertising side of the business report. Whoever runs stock owns inventory age, stranded and unfulfillable items, and the restock view. Whoever owns the profit owns the payments report and the reasons for returns, because those are where margin disappears without anybody deciding it should.

Seller Central lets the account owner give each user access to the areas they need, with view rights and edit rights set separately for each. A person who only reads reports does not need to be able to change a price, and an account where everybody can change everything is an account where nobody is sure who did. Account health is the exception: everyone who works on the account should be able to see it, because a notice seen late by the one person who could see it is still a notice seen late.

The weekly business report and the search terms report are the backbone of how we run an Amazon account, because between them they say which of the three failures a product is having and what the advertising is doing about it.

The same discipline applies wherever you sell: Walmart and eBay each have their own version of the same three questions, with different names and different reports.

If nobody is reading these weekly, that is usually the cheapest thing to change.