Reporting & Analytics

Measure what you manage

Black belt is knowing which numbers drive decisions and ignoring the rest. A handful of metrics explain most of what happens in an account, and a great many dashboards exist to make you feel informed rather than to make you act.

The test for any metric is simple and brutal: if it moved, would you do something different? If not, it is decoration.

The five numbers that carry the signal

Everything else is a slice, a ratio or a rollup of these:

• CTR — click-through rate. Whether the search page picks you. A listing and relevance signal.
• CVR — conversion rate. Whether the detail page closes. The master lever from the White Belt.
• RPC — revenue per click. Conversion rate times order value. The basis of every bid.
• ACoS — ad spend over ad sales. The efficiency of the advertising alone.
• TACoS — ad spend over total sales. The efficiency of advertising against the business.

Your recent data is incomplete, not bad

This is the reporting fact that causes the most bad decisions, and almost nobody is told it explicitly. Amazon attributes a sale back to the click that caused it, within an attribution window — 7 days for Sponsored Products and 14 for Sponsored Brands.

That means yesterday has barely any of its conversions attached yet. Today’s ACoS is not your ACoS; it is your spend against the fraction of sales that have landed so far, and it will keep improving for days.

So the panic cycle writes itself: check yesterday, see a terrible ACoS, cut bids, and then never see the sales that would have arrived — because you changed the thing you were measuring.

ACoS and TACoS tell different stories

ACoS measures your ads. TACoS measures your business. The gap between them is where the interesting decisions live.

Optimising for ACoS alone systematically makes brands under-invest, because a campaign with an uncomfortable ACoS can still be buying the velocity that lifts organic rank — which shows up as total sales rising while ad spend holds flat. That is TACoS falling, and it is the signal that advertising is compounding rather than renting.

Averages hide almost everything

An account-level ACoS is an average of averages, and averages are where losses go to hide. An account at a respectable 28% can easily be two campaigns at 12% subsidising four at 60%.

The habit worth building is to distrust any single figure and immediately ask for its distribution. Not “what is my ACoS”, but “what does the spread look like, and what is at the bad end of it”.

Which report answers which question

Amazon gives you several reports and they are not interchangeable. Picking the wrong one is why people conclude their data is missing when it is simply in a different table.

Targeting is not the same as search terms

The most common report mix-up deserves its own note. The targeting report shows the keywords *you* bid on. The search-term report shows what shoppers actually typed to get there.

For an exact-match keyword those two are nearly the same. For broad, auto or product targeting they are wildly different — and every keyword-research and waste-mining decision from the Green Belt depends on the second one, not the first.

Halo sales, and what your ads really did

When someone clicks an ad for one product and buys a different one of yours, that sale is still yours. Amazon reports this explicitly: total conversions equal promoted conversions plus brand halo conversions — promoted being the products in the campaign, halo being everything else in the brand. Reading only the promoted column understates what your advertising actually earned.

This matters most for brands with ranges. If your ads are the top of the funnel for a whole catalogue, judging each campaign only on the ASIN it advertised will make you turn off the campaign that introduces people to your brand.

New-to-brand: buying customers or renting sales?

ACoS treats every sale identically. A repeat buyer who would have found you anyway and a first-time customer who has never heard of you count the same, which is obviously wrong if you are trying to build something.

New-to-brand metrics separate them, reporting purchases from shoppers who are buying from your brand for the first time. Two campaigns at the same ACoS are doing completely different jobs if one is 70% new customers and the other is 10%.

What a customer is worth after the click

The attribution window closes after a week or two. Customers do not.

Amazon’s long-term sales metrics estimate the additional sales a brand can expect over the following year from shoppers acquired now, based on how far campaigns move new-to-brand shoppers down the funnel — counting engagements like detail page views, branded searches and add-to-carts, not only immediate purchases.

You do not need Amazon’s version to use the idea. If you know roughly what a customer repurchases over a year, a campaign at 45% ACoS acquiring first-time buyers can be a better investment than one at 20% selling to people who already buy from you monthly.

Would this sale have happened anyway?

This is the hardest question in the discipline and the one most reporting quietly avoids. Attribution tells you a click preceded a sale. It cannot tell you the sale needed the click.

The clearest case is your own brand term. Someone searching your exact brand name was already looking for you. Some of those clicks are genuinely incremental — a competitor was about to intercept them — and some you are paying for traffic that was yours for free. Both are true at once, and the split varies by category and by how aggressive your competitors are.

You cannot settle it from a report. You settle it by testing: pause branded advertising for a defined period, watch total sales rather than ad sales, and see what actually happens. That is uncomfortable and it is the only honest answer.

Judge the trend, not the day

Daily numbers on a normal account are mostly noise: weekday and weekend behave differently, attribution is still landing, and a single large order can move a small campaign’s ACoS several points.

Compare like with like — a full week against the previous full week rather than Tuesday against Sunday — and expect any change you make to take a couple of weeks to read cleanly. That is the same discipline as the Green Belt click threshold, applied to time instead of clicks.

From metric to move

Every number should point at an action. If you cannot name the move, you are collecting rather than analysing:

• Low CTR → relevance or the search-page tile. Image, price, review count, or the keyword is wrong for you.
• Good CTR, low CVR → traffic arrives, page does not close. A listing, price or stock problem, not a bid one.
• Rising ACoS, falling TACoS → advertising is compounding into organic. Usually lean in, not back.
• Falling ACoS, rising TACoS → the ads look efficient because organic is shrinking underneath. Investigate.
• Healthy everything, flat sales → you are budget-capped or out of demand. Check which.

A reporting cadence

Speed to insight beats more dashboards. The point of a cadence is to stop you reacting to noise while making sure nothing silent runs for a month.

• Daily — anomalies only. A spend spike, a stockout, a campaign gone dark. Two minutes.
• Weekly — the optimisation loop: search terms, bids, budgets, week against week.
• Monthly — trends. ACoS against TACoS, organic share, harvests, what changed and what it did.
• Quarterly — the honest question: is the account structured for what the business is now?

Build a report you will actually read

Most reporting fails from being ignored rather than from being wrong. A dashboard with forty tiles gets glanced at; a dashboard with six gets used.

The discipline is to earn every element. If a number would not change what you do this week, it belongs in the archive, not on the screen — and the comparison should already be made for you, because a figure without a reference point is not information.

• Every number carries its comparison: this week against last, not a bare total.
• Pair ACoS with TACoS. Neither is trustworthy alone.
• Show a distribution somewhere, not only account averages.
• Flag the two silent killers directly: capped winners, and stock about to run out.
• Anything you have not acted on in three months comes off the screen.

Do this in SellerMateWhite-Label Reports →Role-based dashboards and scheduled PDFs under your own brand and domain — for an agency, the difference between a report a client reads and one they file unopened.

Reporting pitfalls to avoid

Put the numbers where you will actually see them

The reports exist; the friction is getting to them. Most sellers know exactly which numbers matter and still look at them monthly, because the answer lives three exports and a pivot table away.

Put the handful of decision-driving numbers where you will actually see them, with the comparison already made — week against week, ACoS against TACoS, spend against organic share. Everything else can stay in the archive until a question needs it.

Master Sifu — A number you have to go and fetch is a number you will check monthly, whatever you intended. Make the comparison you actually act on the one that greets you, and let everything else wait in the archive until a question needs it.