PPC Audit & Optimization

The audit mindset

Optimisation is not a project you finish, it is a rhythm you keep. A good audit is a loop you run on a schedule that asks the same questions in the same order — so nothing profitable stays ignored and nothing wasteful runs for long.

The order matters more than most people expect, and it is the main thing this guide adds. An audit done in the wrong sequence will have you carefully tuning bids on a campaign you were about to restructure, or cutting a term that was only losing because the page was out of stock.

Diagnose before you fix

Almost every bad optimisation starts with treating a symptom at the wrong level. Sales are impressions × click-through × conversion × price, so before touching anything, find out which of those four actually moved.

Amazon’s own troubleshooting follows the same ladder: too few impressions is a bid and budget question, a weak click-through rate is a search-page question, and clicks that do not convert are a detail-page question rather than an advertising one.

Work from cheapest fix to most nuanced

Run the steps in this order every time. Each one changes the data the next step sees, which is exactly why sequence matters — cut the waste first and every efficiency number you look at afterwards is telling the truth.

• Cut obvious waste. Cheapest, fastest, highest certainty.
• Fix placements. Often a large CPC change for no structural work.
• Correct bids by bucket, now that the numbers are clean.
• Free the capped winners, funded by the waste you just cut.
• Check structural drift, which is slow work and rarely urgent.

Size the prize before you act

An audit produces a list of things you could fix. Without a value attached to each, people work the list top to bottom, or worst-first, and spend an hour on a campaign carrying $40 of spend.

Put a rough number on each finding before you touch anything. It takes minutes and it reorders the whole list: a term wasting $310 a month outranks six terms wasting $12, and a capped campaign leaving an estimated $900 of sales unserved outranks both.

Step 1 — cut the obvious waste

Open the search-term report and find terms that have spent real money and returned little or nothing. These are the clearest wins on the account because they require no forecast — the money has already been spent and the answer is already in.

Judge on clicks rather than days, using the Green Belt threshold: roughly one divided by your conversion rate is the clicks needed to expect a single sale, and a few multiples of that before a zero means anything. Then negate by pattern where you can, rather than one term at a time.

Master Sifu — This exact decision is the White Belt trial in the PPC Dojo: a term burned $312 with zero sales — what do you do? (Answer: negate it.)

Step 2 — check the placements

The same keyword performs very differently at top of search, in the rest of search, and on product pages, and a placement modifier multiplies your bid rather than adding to it.

Pull the placement report and look for modifiers pushing spend into placements running well past your target, and for converting placements you are under-serving. A modifier applied months ago to a campaign that has since changed shape is one of the most common quiet leaks in an account.

Step 3 — correct bids by bucket

Now that the waste is gone and placements are sane, the efficiency numbers are trustworthy. Sort keywords into buckets and treat each differently rather than nudging everything by the same percentage:

Thresholds worth borrowing

Audits stall on “is this enough data yet?”, so it helps to have defaults you can argue with rather than deciding fresh every time. Amazon publishes a few, and they are reasonable starting points:

• A search term with five or more orders and healthy returns over the last 14 days is worth isolating into its own campaign.
• A term generating clicks and zero orders is a negative exact match, not a bid reduction.
• Before a known event, raise budgets incrementally — on the order of 20–30% — three to five days ahead rather than on the day.
• Check average time in budget, and the estimated impressions, clicks and sales missed to running out. Those numbers turn “it caps sometimes” into a figure.

Step 4 — free the capped winners

Now spend the money you just freed. Profitable campaigns hitting their daily cap are the easiest return in the account, because you are not making a bet — you are removing a restriction from something that already works.

A perfectly flat daily spend line is the tell. Real demand is never that smooth, so a campaign landing on the same figure every day is showing you a ceiling rather than steady performance.

Step 5 — check for structural drift

Structure decays quietly, and inefficient structure is on Amazon’s own list of common mistakes. Products get added to whichever ad group was open, a harvested term never gets negated in its discovery campaign, a test from March is still running in November because nobody switched it off.

This is slow work and rarely urgent, which is exactly why it needs a scheduled slot rather than good intentions:

• Terms live in two campaigns at once, bidding against each other.
• Ad groups that have accumulated products across different price points.
• Campaigns with spend and no clear job you can state in a sentence.
• Budgets that no longer match current priority.

Auditing an account you did not build

Everything above assumes an account you know. Taking over someone else’s — a new client, an acquisition, a colleague who left — is a different job, and running the routine loop on day one is how people break things that were working.

The first pass is archaeology, not optimisation. Nothing in a strange account is obviously wrong, because you do not yet know what it was for. That campaign with a 70% ACoS might be defending a brand term deliberately.

What is not an advertising problem

A large share of what looks like poor ad performance is not fixable from the ad console at all, and this is where audits waste the most effort. If clicks arrive and do not convert, the advertising is working — something after the click is not.

What a pass actually looks like

Concretely, on a small account spending around $3,000 a month, a weekly pass is perhaps twenty minutes and produces a handful of changes:

• Search terms: three terms over 30 clicks with no orders. Two share a root word, so one phrase negative covers both. Recovered: roughly $190 a month.
• Placements: a top-of-search modifier of 150% on a discovery campaign nobody meant to leave there. Reduced to 0%. Recovered: roughly $240 a month.
• Bids: four terms comfortably below target ACoS with a hundred-plus clicks each, raised 15%. One term above target with heavy spend, cut 15%.
• Budgets: one campaign capped by 3pm every day at 22% ACoS. Cap raised, funded by the two fixes above.
• Structure: one term found live in both the auto and exact campaign. Negated in auto.

Change less than you want to

The instinct during an audit is to fix everything visible. Resist it. Every simultaneous change costs you the ability to attribute the result, and a week later you will have a different account and no idea which edit produced it.

Cap each pass at a manageable number of deliberate changes, write down what you changed and why, and let it run long enough to read. An audit that produces three well-understood changes beats one that produces forty you cannot untangle.

Run it on a cadence

Different questions deserve different frequencies. Checking structure weekly is a waste of a morning; checking for anomalies quarterly is negligence.

Do this in SellerMateAutomation →The repetitive half of an audit — zero-sale negations, budget-capped winners, bid corrections by bucket — is exactly what a rule does well. Automate that half and spend your own attention on the judgement half.

What a healthy account looks like

It helps to know what you are aiming at, because an audit with no target becomes fiddling. A healthy account is not one with a perfect ACoS — it is one where every number has a reason:

• Almost no spend on terms with zero sales over a meaningful click volume.
• Profitable campaigns are not sitting at their budget cap.
• Every live campaign has a job you can state in one sentence.
• Harvested terms are negated in the campaigns that discovered them.
• Organic share is flat or growing — TACoS is not quietly climbing.

Audit pitfalls to avoid

Automate the repetitive half

Most of this audit is mechanical: the same queries, the same thresholds, the same obvious negations, every week. That half does not need judgement — it needs consistency, which is the part humans lose first.

Automate the mechanical passes and keep your attention for the parts that genuinely need it: whether a campaign still has a job, whether a placement bet is working, and whether the account is drifting away from the structure you designed.

Master Sifu — The test for whether a step should be automated is not how hard it is, it is whether you would make the same decision every time. Same decision every time — automate it. Anything you would want to think about first, keep.