A paid media audit is a sequence rather than a list (the order carries most of the value). If conversion tracking is broken then your cost per acquisition is fiction, your bidding has spent months optimizing toward that fiction, and every judgment about structure, audiences or creative is a judgment about fiction. Checking ad copy in an account with duplicated conversions is a pleasant way to lose a day.
1. Conversion tracking integrity decides whether the rest means anything
This step is the whole audit, because the other eight depend on it. The question isn’t whether a tag fires. It’s whether the number in the platform corresponds to something a human in the business would recognize as an event.
- Does each action correspond to a real event? Walk the funnel yourself. Submit the form. Watch what fires. I’ve opened accounts where the primary conversion was a thank-you page view that also fired on newsletter signup and on login.
- Is it deduplicated? Check the counting setting against what the action means. Google’s documentation is direct about this: use “every conversion” for sales because every sale likely adds value, and “one conversion” for leads because usually only one unique lead per click does. A demo request counted as “every” is one impatient user entering a phone number twice, recorded as two pipeline events.
- Does anything come back from the CRM? If the platform only knows about form fills, it will optimize for form fills, and the rubbish that produces is the most expensive problem in B2B paid search. Google supports qualified lead and converted lead actions for exactly this, tracking leads that have been qualified offline in your CRM, and importing them changes what the algorithm chases.
- Do the totals reconcile? Platform against analytics against CRM. They’ll never match exactly, and they should be explicably different. My own trigger for discomfort is platform conversions exceeding CRM-recorded inquiries by more than about a third across a full month, though the defensible gap depends entirely on your sales lag.
- Are there orphan or stale actions? Actions from a previous website, imported goals nobody recognizes, a duplicate tag left behind by a migration (that last one is almost a rite of passage). Each one still feeds the bidding algorithm every day.
If this step fails, stop. Fix tracking, allow a clean period, re-baseline. Everything you conclude in the next eight steps stays provisional until you have.
2. Account and campaign structure is where money hides
The question is whether the architecture reflects how the business allocates budget, by product line, market and margin, or the order in which campaigns got built over four years and two agencies.
Brand should be separated from non-brand into its own campaign with its own budget, because blending them destroys your ability to read either. No two campaigns should compete for the same query, and that gets verified in the search terms report rather than assumed. No campaign should be so small that automated bidding can’t learn inside it, and my own nervousness starts below roughly thirty conversions per campaign per month on a target CPA strategy (a rule of thumb from engagements rather than a published threshold). A naming convention should survive a new starter reading it cold. And somebody should honestly count how many campaigns run because a person once built them rather than because they earn their budget.
3. Bidding strategy and its portfolio effects
Bidding sits third because it’s the most consequential setting in the account and the one most commonly left on whatever it was when somebody switched it on.
Per campaign: which strategy, and why. When the target last changed, which change history answers in a minute. Whether the target matches a cost per acquisition the business still accepts, or one from a pricing model since abandoned. Whether a portfolio strategy spans campaigns with genuinely different economics, which is the quiet way a profitable campaign subsidizes an unprofitable one inside a single averaged target. And whether anything runs on maximize conversions with no target at all, which is less a bidding strategy than an instruction to spend the budget by close of day.
4. Search terms and negative keyword hygiene
Now that you trust the conversion numbers, the search terms report becomes readable. Export a long window, because Google omits search terms without enough query activity to keep to its privacy standards, and a ninety-day pull surfaces terms a seven-day pull won’t.
What you want out of it: the proportion of spend that went to queries nobody would have chosen. Whether negative lists exist, where they’re applied, and whether anyone has looked since. Whether they use the right match type. Whether brand terms are negated out of non-brand campaigns. And whether the list now blocks a term you’d want back, which happens in every account maintained by more than one person over more than a year.
5. Audience settings and exclusion lists fail quietly
Are remarketing lists populating and large enough to use? Is customer match data stale? And is each audience attached as observation or as targeting? Google’s documentation is clear on the difference: with observation the reach of the campaign isn’t affected and you’re only measuring, while targeting restricts who can see the ads. A campaign accidentally set to targeting against a small list is the most common explanation for impressions falling off a cliff on a date nobody can explain.
Then exclusions, which in B2B matter more than the audiences do. Existing customers out of acquisition campaigns. Open opportunities out of anywhere sales would rather own the conversation. Placement exclusions on anything running across display, video or Performance Max inventory.
6. Budget pacing and wasted spend
Three questions. Which campaigns are limited by budget while others underspend, since reallocating is usually the fastest uncontroversial win in the account. Where spend is concentrated, because a long tail that looks healthy in aggregate is often five terms and a rounding error. And what has spent across a meaningful window with nothing at all to show for it.
Be disciplined about “meaningful”. I wouldn’t pause a keyword on fewer than about a hundred clicks with no conversion in a mid-funnel B2B account, and on a long sales cycle even that is thin evidence. Pausing on small numbers is how accounts get pruned down to nothing by people being careful.
7. Geographic and ad schedule settings hold the cheapest findings
These get configured once and never revisited, which is exactly why they pay.
Location targeting set to presence or interest rather than presence alone, which puts budget in front of people in another country reading about yours. Regions you can’t practically serve, still included because somebody ticked the country at the top level. An account time zone different from the business time zone, which shifts every day-parting decision and every day-of-week report by a few hours. And an ad schedule written for a business that no longer operates that way, usually one switching off overnight in a market now worth a third of revenue.
8. Creative and landing page alignment belongs eighth, not first
Everybody does this step first, because it’s visible and you can have an opinion about it without exporting anything. It belongs here.
The useful question isn’t whether the ad is good. It’s whether the query, the ad and the page make one argument. Does the page answer the thing the person typed, in the first screen, in their words? Is the form asking eleven questions at the top of the funnel? Does a click from a comparison query land on a homepage? Then check pinning on responsive search ads, because heavy pinning is often correct for regulated or brand reasons and is worth knowing about before you read any asset-level report.
9. Reporting honesty decides whether anything gets acted on
Last, and the step that determines whether the other eight ever change anything.
The usual culprit is brand and non-brand reported as one number. Brand search converts at a rate non-brand can’t approach, costs almost nothing, and would largely have happened anyway. Blend them and the account looks fine in every board pack while the acquisition half quietly does nothing. Separating the two is the most uncomfortable slide in an audit and reliably the most useful one.
After that: which attribution model produced the numbers, and whether it changed mid-period. And whether the report leads with the metric the business cares about or the metric that moved.
What a free automated PPC audit tool can and can’t see
Let me be fair to these, because they’re useful and I run one before I start. A grader parses the account in minutes, costs nothing, and reliably finds the mechanical layer: disapproved ads, broken final URLs, missing assets, duplicate keywords, conflicting negatives, budget-limited campaigns, unattached audience lists, and spend against keywords with no conversions. Finding that by hand is an hour of clicking nobody needs to do.
What no automated tool sees is everything in steps one, two and nine. It can’t know whether the conversion action corresponds to a real event, because it can’t know what your business considers a lead. It can’t read your CRM and tell you the leads were unqualified, or that your target CPA was set against a pricing model you abandoned two years ago. And it can’t see what’s absent: the market you aren’t bidding in, the campaign that should exist and doesn’t.
Most free graders are also lead-generation assets for the company that built them. That doesn’t make the findings wrong, and it does shape which ones get rated critical. A severity rating is an opinion with a colour applied.
An automated audit tells you the account isn’t doing what it was told. It can’t tell you whether it was told the right thing.
How to run the PPC audit checklist without losing the week
Work top to bottom and resist jumping to the interesting parts. On a single-platform account of moderate complexity this is a few days of work, most of it in steps one, two and four, and that’s an indication rather than a quote.
The output that matters is the ordered list of what to change first, with the findings that invalidate other findings at the top, and an honest note against anything you couldn’t verify. That’s what a PPC audit buys, and the sequence is the same whether the account sits in Google Ads or spans the four platforms a B2B paid media program covers. Either way, fix the measurement before you argue about the creative.