The Amazon PPC Audit: A 90-Minute Pass That Finds Wasted Spend

An Amazon PPC audit is a fixed pass over an account that is already running: pull four downloadable reports — targeting, search term, placement and budget — sort each one by cost, and judge every expensive row against a threshold you set before you open the file. Ninety minutes is enough because an audit is a diagnosis, not a rebuild. What you produce is a short list of bids to cut, targets to pause, negatives to add and budgets to raise, plus a dated log of every change.

This guide is the diagnostic pass only. Break-even ACoS, campaign structure and match-type strategy are reasoned out in the profit-first Amazon PPC playbook; if the account is under a month old, run the beginner setup instead — there is nothing to audit yet.

Before the Clock Starts: Two Numbers and Four Reports

Write down two numbers first; everything below is a comparison against them.

1. Break-even ACoS — the ad cost ratio at which a sale contributes nothing and loses nothing. Derivation and worked example: the ACoS guide.
2. Break-even cost per order — your contribution margin on one unit, in currency. It tells you when a zero-order target has spent enough to be judged.

Then pull the files. All four are downloadable reports in the advertising console’s reporting center (budget report for Sponsored Products, checked 2026-09-02).

The lookback windows are Amazon’s, stated on each report’s help page (checked 2026-09-02). The 65-day search term window is the shortest of the four, so a quarterly cadence silently loses search term history.

Set the end of your date range 14 days before today. Every sales column here is attributed within 7 or 14 days of the click (advertised product report column definitions, checked 2026-09-02), so recent rows are still filling in, and judging last week’s spend against last week’s sales makes healthy targets look dead. The buffer is my recommendation, not an Amazon rule.

One naming trap: these four reports carry ROAS, not ACoS. ACoS is the reciprocal — cost divided by sales, so a ROAS of 4 is a 25% ACoS.

Minutes 0–20: The Targeting Report

Start here, not with search terms. The targeting report covers “targets in all campaigns that received at least 1 impression”, while the search term report “includes only the search terms that resulted in at least 1 ad click” (targeting report; search term report, both checked 2026-09-02). Targets that burn impressions and never earn a click are invisible there and visible here.

Sort by Cost, descending, and sort the top 50 rows into three buckets:

My suggested spend threshold: a zero-order target has had a fair trial once its Cost reaches twice your break-even cost per order. At one break-even cost per order, a single delayed order still makes the target neutral; at twice that it is losing money even in the best case. This is a rule of thumb derived from your margin — Amazon publishes no such standard. The click-count version of the same question is worked through in the negative keyword guide.

One column changes verdicts: the report separates Sales (promoted), sales of the advertised SKU, from Sales (halo), sales “where the purchased SKU was other than the one advertised” (same page, checked 2026-09-02). A target that looks unprofitable on promoted sales alone may pay for itself across your catalogue, so read both columns on a multi-SKU account.

Minutes 20–35: The Search Term Pass

This pass is short because the full method has its own guide. Sort by Cost, take the top 50 rows, and mark two things: queries with meaningful spend and no purchases, and queries semantically wrong for the product regardless of spend. Add them as negatives; match-type and campaign-level decisions are in how to find negative keywords.

The audit-specific detail is what you should skip. Amazon’s documentation notes that alphanumeric entries in the Customer Search Term column “correspond to ASINs and the related product detail page on which your ad displayed”, and that they appear only for automatic-targeted campaigns and product-attribute-targeted ad groups (search term report, checked 2026-09-02). Those rows are a product-targeting decision, not a keyword one. Negative keywords will not touch them.

Minutes 35–55: Placement, and the Adjustment That Multiplies

This is the pass most sellers never run, and the one that most often finds a single large leak.

Amazon groups Sponsored Products placements into three: “top of search (first page), rest of search, and product pages” (placement report, checked 2026-09-02). Top of search is defined as “Up to the first 5 positions in the top row of the first page search results on desktop, and up to the first 4 positions on mobile” (adjust Sponsored Products bids, checked 2026-09-02).

The leak compounds. A placement bid adjustment “works in addition to your selected bidding strategy” (adjust Sponsored Products bids, checked 2026-09-02), and dynamic bidding may raise bids “by up to 100%” (bidding strategies, checked 2026-09-02). Amazon’s worked example: a $1 dynamic bid with a 50% top-of-search adjustment becomes $1.50, and dynamic bidding up and down can push it to $3 (adjust bids page, checked 2026-09-02). Adjustments are capped at “a combined 900% increase (10x)” to the base bid (same page).

So: list every campaign with a top-of-search adjustment above zero, find its top-of-search row in the placement report, and compute ACoS for that row alone. Three verdicts:

• Below break-even → the adjustment is earning its keep. Leave it.
• Between break-even and roughly 1.5× break-even → halve the adjustment rather than removing it, and re-read in two weeks.
• Far above break-even → set the adjustment to zero before touching a single keyword bid.

The 1.5× line is my threshold, chosen so a placement worth defending for rank is not deleted on one noisy fortnight. Amazon publishes no guidance here.

If performance swings by time of day rather than by placement, that is a scheduling question — see Amazon PPC dayparting.

Minutes 55–70: The Budget Report Finds the Opposite Problem

Every other pass removes spend. This one usually adds it, which is why it belongs in an audit about profit rather than cuts.

The budget report “shows how your campaign performance may be impacted when your campaigns are out of budget” and carries three columns you will not find elsewhere: Average Time in Budget, the share of the date range the campaign was in budget; Estimated Missed Sales Range Min/Max; and Recommended Budget (budget report, checked 2026-09-02). Amazon’s tip on that page: find high-performing campaigns below 100% Average Time in Budget with estimated missed sales, and raise the budget.

Sort ascending by Average Time in Budget and apply one rule that Amazon does not state:

• Below break-even ACoS and below 100% time in budget → raise the budget. You are rationing a profitable campaign.
• Above break-even ACoS and below 100% time in budget → change nothing. Running out of budget is currently limiting the damage, and raising it before you fix targeting converts a small loss into a large one.

Treat the missed-sales figures as directional: Amazon labels each “an estimate, not a guarantee” (same page, checked 2026-09-02), and Recommended Budget optimises for staying in budget, not for your margin.

Minutes 70–85: One Pass on Products, Not Keywords

Open the advertised product report — one row per advertised ASIN, 90-day lookback (advertised product report, checked 2026-09-02). Sort by Cost and read the Purchase rate column, which Amazon defines as ad-attributed purchases over ad-attributed clicks within 7 days (same page).

Suggested threshold: any ASIN whose purchase rate is below half your account average goes on a listing list, not a bid list. Clicks arriving and not converting is a price, image, review-count or availability problem, and no bid work fixes it. This is the commonest way an audit wastes its own 90 minutes: the account gets re-optimised around a listing that was never going to convert.

One quirk so you do not misread the file: this report “may include PAUSED campaigns within the period with 0 impression” that were active in a different period (same page, checked 2026-09-02). Rows with zero everything are not errors.

Minutes 85–90: The Change Log

Write one line per change: date, what changed, old value, new value, why, and a review date. Because sales are attributed within 7 to 14 days of the click, set every review date at least 14 days out.

If the audit produced more than about twenty edits, apply them as a bulk file — the mechanics and the columns you must not touch are in Amazon Ads bulk operations.

Which Numbers Are Amazon’s, and Which Are Mine

Every figure above falls into one of two categories.

Four Ways an Audit Goes Wrong

1. Auditing an unsettled window. Judging the last seven days makes every target look unprofitable: the clicks have landed, the attributed sales have not.
2. Cutting on ACoS with no break-even number. A 45% ACoS is excellent on a 60% margin and ruinous on a 25% one. Without the number, an audit is just a preference.
3. Treating a conversion problem as a bid problem. High impressions with no clicks, or high clicks with no orders, are listing symptoms. Bids cannot fix either.
4. Changing forty things at once. With a 14-day attribution tail, simultaneous edits produce one unattributable result. Change the expensive things, log them, leave the rest for next month.

The Amazon PPC Audit Checklist

• Break-even ACoS and break-even cost per order written down
• Date range ends 14 days before today
• Targeting, search term, placement, budget reports downloaded
• Targeting report top 50 rows split into cut / bid-down / listing buckets
• Search term report top 50 rows screened; ASIN rows routed to product targeting
• Every campaign with a top-of-search adjustment checked against its placement row
• Budget report sorted by Average Time in Budget; rationed profitable campaigns raised
• Advertised product report scanned for low-purchase-rate ASINs
• Change log written, with review dates 14+ days out

Frequently Asked Questions

How often should I run an Amazon PPC audit?

Monthly for accounts spending enough that a bad fortnight matters, quarterly at the very least. The search term report’s 65-day lookback sets the outer limit: audit less often and you lose search term history permanently.

What date range should an Amazon PPC audit cover?

Thirty to sixty days of data, ending about 14 days before today. Sixty days gives most targets enough clicks to judge; ending two weeks back keeps unsettled attribution out of the file. Both are practical choices, not Amazon requirements.

Can I audit an account properly without third-party PPC software?

Yes — every step above uses only downloadable reports and a spreadsheet. Software pays for itself when the audit stops being the bottleneck: too many targets to read by hand, or changes needed daily rather than monthly. The buy-or-not decision is in the PPC software guide; AMZFinder, an independent tool review site, keeps a roundup of Amazon PPC management software.

Does pausing a keyword hurt my organic ranking?

Not directly. What moves organic rank is sales velocity, so pausing a target that was quietly producing orders removes those sales and the rank they supported — the argument for lowering bids first and pausing second.

What is the fastest way to reduce PPC spend without losing sales?

In order: zero out top-of-search adjustments on campaigns whose top-of-search ACoS is above break-even, cut zero-order targets past twice your break-even cost per order, then add negatives. All three touch spend that produced nothing, so the sales you keep are unchanged. Bid-down passes on converting targets come after, because they do cost volume.

The Bottom Line

An Amazon PPC audit is a fixed sequence, not an act of judgement: two numbers written down, four reports sorted by cost, one threshold per report, a dated log. The placement and budget passes are the two most sellers skip, and between them they hold most of the money an account loses unnoticed — a bid adjustment multiplying against dynamic bidding, and a profitable campaign rationed to half the day. Ninety minutes a month, logged, beats a rebuild once a year.

Data checked 2026-09-02 against Amazon Ads help documentation. Report names, column and placement definitions, lookback windows and bid-adjustment limits are Amazon’s published figures; the spend, ACoS and purchase-rate thresholds are rules of thumb derived from your own margin, labelled as such throughout.