Amazon PPC audit: the twelve checks we run, in the order we run them.

A real audit answers one question: where is money leaving the account, in what order of size, and what would have to change for it to stop. Most "free PPC audits" are a lead form with a PDF at the end. The PDF says your ACoS is high and your negative keyword list is short. You knew that.

That question is a mechanical exercise. Below is the exact list we work through, in the order we work through it, and what a bad answer looks like at each step. You can run all twelve yourself with a Search Term Report and a Bulk Operations export.

1. What fraction of spend is on search terms you never chose?

Pull 60 days of search terms. Split spend into three buckets: terms that match a keyword you deliberately added, terms that came in through broad or auto and converted, and terms that came in through broad or auto and did not.

That third bucket is the number that matters. Under 10% is healthy. Over 30% and the account is not being managed, it is being funded.

This check goes first because it is usually the largest single line and the cheapest to fix. You do not need a new bidding strategy to stop paying for a term that has spent $180 across four months and converted twice. The waste calculator prices the gap in a minute if you want the number before you start.

2. Is your cheapest traffic your worst traffic?

Sort search terms by average CPC, ascending. Look at the conversion rate of the cheapest decile.

There is a reflex in PPC to treat a falling CPC as an efficiency win. Often it is the opposite: broad match has found a pool of loosely related, low-competition queries that nobody else wants, and it is buying them because they are cheap. Cheap clicks that never convert are more expensive than expensive clicks that do.

If your account CPC is well below the category norm and conversions have not moved, that is this problem.

3. How much of your spend has no conversion data behind it at all?

For every keyword, count clicks in the last 60 days. Anything under roughly 15 clicks has no statistically meaningful conversion rate — you cannot tell a 2% converter from a 6% converter at that volume.

Now sum the spend across all of those keywords. That is the share of your budget being bid on guesswork. It is usually much larger than people expect, and it is the single biggest argument for Bayesian bidding over rules: a rules engine treats "0 conversions in 8 clicks" and "0 conversions in 300 clicks" the same way. They are not the same evidence. That difference is most of what separates the tools in this category.

4. Are your bids consistent with your own target ACoS?

Take target ACoS, multiply by average order value, multiply by the keyword’s conversion rate. That is the theoretical max bid for that keyword. Compare it to what you are actually bidding.

Do this across the account and you will usually find two populations: keywords bid far below their ceiling, leaving volume on the table, and keywords bid above it, buying unprofitable clicks on purpose without meaning to. Both are common in the same account at the same time.

5. Does your ACoS target survive contact with your margin?

A surprising number of accounts run a target ACoS that was set once, by someone who has left, based on a margin that has since changed.

Recompute it: unit economics after COGS, FBA fees, referral fee, returns and storage. The break-even ACoS falls out of that. If your target sits above break-even you are buying revenue at a loss, which is a legitimate strategy for a launch and a slow bleed everywhere else.

6. What is the TACoS trend, and is ad spend actually buying organic rank?

ACoS on its own tells you about the ad. TACoS — ad spend over total revenue — tells you whether the ad is doing anything for the business.

Plot it over six months. A TACoS that falls while ad spend holds steady means paid is pulling organic up behind it. A TACoS that is flat or rising means you are renting the sales, not building toward anything.

7. Is any of this running while you are out of stock?

Cross-reference campaign activity against inventory history. Spend on an ASIN in the days before it went out of stock is spend on clicks that could not convert, and it also damages the conversion-rate history that every subsequent bid decision will be based on.

The second-order damage is worse than the first. You lose the money once, then you bid badly on the poisoned data for weeks afterwards.

8. Are you bidding at hours that never convert?

Break spend and conversions out by hour of day and day of week. Most catalogues have dead windows — often 01:00–06:00 in the marketplace’s local time — where spend continues and conversions do not.

Two cautions. First, you need enough volume for this to be signal rather than noise; under a few hundred conversions a month, hourly splits are mostly random. Second, a low conversion rate at a given hour is not automatically a reason to stop bidding if the cost per acquisition still works.

9. Is anything cannibalising anything?

Look for the same search term winning impressions across multiple campaigns. You end up competing against yourself, paying more per click than you would have, and splitting the conversion history across ad groups so that neither accumulates enough data to be bid on properly.

The fix is usually negatives at the campaign level and a clear rule about which campaign owns which term — not a new tool.

10. Does the negative list reflect what actually happened?

Not "is there a negative list", but: does it contain the terms your own search term report says have spent money without converting?

Go back through the report, sort by spend among zero-conversion terms, and check each one against your negatives. The gap between what the data says and what the list contains is the honest measure of how actively the account is being managed.

11. Is budget distributed by intent or by inertia?

Group campaigns into brand defence, category/generic, competitor conquesting and retargeting-adjacent placements. Now look at what share of budget each one takes and what share of conversions it returns.

Brand campaigns almost always look spectacular here, because they are capturing demand that already existed. That does not make them worthless — brand defence is insurance — but it does mean the account-level ACoS is flattered by them, and every decision made off that blended number is skewed. Recompute your ACoS excluding brand and see whether the picture holds.

12. For each bid currently set: can anyone say why?

The last check is the one that decides whether the first eleven stay fixed.

Pick five bids at random. For each, ask what evidence set that number, when, and what would have to change for it to move. If the answer is "the tool did it" or "we raised everything 10% in June", the account will drift back to where it was inside a quarter, because nothing about how decisions get made has changed.

This is the check we built Mirox around. Every bid it sets carries the reasoning with it — the conversion-rate estimate and how confident it is, which guardrails were evaluated, which constraint ended up binding, and what the bid would have been without it. Not a score. The actual chain.

Running this on your own account

Nine of these twelve need only a Search Term Report and a Bulk Operations export. Check 7 needs inventory history, check 5 needs your unit economics, and check 12 needs an honest conversation with whoever owns the account.

If you would rather not spend the afternoon: we will run all twelve on your account and send you the findings. No card, and you keep the report whether or not you go any further.

And if you want to see what check 12 looks like in practice before you talk to anyone — Simulation Mode runs on your live account in read-only for 30 days. It changes nothing. At the end you get every bid it would have set differently from your current setup, the reasoning behind each one, and the CSV. That file is yours either way.

The one-line version

Twelve checks, in size order: find the spend you never chose, the spend with no evidence behind it, and the bids nobody can justify — then fix the process that produced them, or you will be running the same audit again next quarter.

Common questions

What is an Amazon PPC audit?

A structured review of an advertising account that identifies where spend is being lost and why. A useful one goes beyond reporting ACoS: it quantifies spend on search terms you did not choose, spend on keywords with too little data to bid on, and bids that are inconsistent with the account’s own profitability targets.

How often should you audit an Amazon PPC account?

A full audit quarterly, plus a weekly review of the search term report. The search term report is where new waste appears first, and it compounds fast under broad match.

Can you audit an Amazon PPC account yourself?

Yes. Nine of the twelve checks need only a Search Term Report and a Bulk Operations export from Campaign Manager. The remaining three need inventory history and your own unit economics.

How much does an Amazon PPC audit cost?

Agencies typically charge between $500 and $2,500 for a one-off audit, and many offer a limited version free as a sales step. Mirox runs the twelve checks above at no cost and sends the findings regardless of whether you become a customer.

Start with the free audit, or read what PPC management actually involves week to week if you are deciding who should own the account afterwards.