Most Google Ads audit templates open with keywords and bids. That is the last stage, not the first — and starting there is why so many audits produce a long list of changes that make the account busier without making it better.
This is ordered by dependency. Each stage assumes the one before it is sound. Work down; the first stage that fails is the finding, and everything below it is speculation until that is fixed.
Stage 1 — Tracking
Nothing below this matters. Smart Bidding optimises toward whatever conversions you give it, so wrong data produces confident, well-funded mistakes.
- Conversion actions reflect genuine business value — an enquiry, a qualified call, a sale. Not a page view, not a scroll
- Only one conversion action is set as primary per meaningful outcome
- No double counting. The same action recorded by two tags, or by both a thank-you page and a form-submit event, inflates everything downstream
- Counting setting matches the business. "Every" suits ecommerce; "One" suits lead generation, where one person enquiring three times is one lead
- Conversion window is deliberate and reflects a realistic decision period
- Attribution model chosen consciously, not left at whatever the account defaulted to years ago
- Offline conversions imported where the real outcome happens after the form — otherwise you are optimising for enquiries rather than customers
- Tracking verified end to end with a real submission on a real device, not by the tag assistant reporting that a tag exists
- Consent handling correct for your markets, and understood well enough that you know whether the reported numbers are complete
The offline-conversion item is the one that separates accounts. Optimising toward form fills teaches the system to find people who fill forms. Optimising toward closed revenue teaches it to find customers. Those are different audiences.
Stage 2 — Account structure
Structure determines whether the data you now trust is legible.
- Campaign types separated. Search, Shopping, Display and Performance Max compete for the same budget and report differently
- Brand and non-brand separated. Mixed together, brand traffic's high conversion rate masks non-brand performance entirely
- Budgets allocated by intent, not evenly
- No two campaigns bidding on the same terms without a deliberate reason
- Geographic targeting set to "presence" rather than presence-or-interest, unless you genuinely want people researching your location from elsewhere
- Ad schedule reviewed — running at hours nobody answers is real waste for any business whose conversion is a call
- Networks reviewed. Search Partners and Display expansion are enabled by default and behave very differently from Search
- Campaigns without enough conversion volume are not on Smart Bidding, or are consolidated until they are
Rebuilding an account on top of tracking that was never verified is the most common reason a rebuild disappoints, which is why
the Google Ads work that starts with tracking rather than bids.Stage 3 — Targeting and search terms
Now that structure is legible, look at what you actually paid for.
- Search terms report reviewed — not the keyword list. This is the only view of what people genuinely typed
- Irrelevant terms excluded as negatives, at the right level
- A negative keyword list exists and is maintained, not created once
- Match types deliberate. Broad match with weak conversion data spends fastest in the least relevant places
- Job seekers, students and competitors excluded where they are a meaningful share — "jobs", "salary", "free", "course", "tutorial"
- Audience exclusions applied — existing customers, recent converters, and anyone the business cannot serve
- Performance Max exclusions configured where the campaign type allows, and its limits understood rather than assumed
Stage 4 — Ads and landing pages
- Ad copy matches the search intent, not just the keyword
- Enough assets for the format, without stuffing keywords into headlines
- Assets pinned only where genuinely necessary — over-pinning removes the system's ability to test
- Extensions and assets in use — sitelinks, callouts, structured snippets, call and location where relevant
- Landing page matches the ad promise. An ad for a specific service landing on a general homepage wastes the click you already paid for
- Landing page loads quickly on mobile, which is where most paid clicks arrive from
- The conversion action is obvious above the fold, without scrolling
- Form length justified. Every field is a decision to lose some proportion of the people who started
Stage 5 — Bidding and budget
Last, because bidding is the layer that acts on everything above it.
- Bid strategy matches the data available. Target CPA and Target ROAS need consistent conversion volume; without it they amplify noise
- Targets set from actual business economics, not from what the platform suggests
- Budget-limited campaigns identified — a capped campaign converting well is the cheapest available improvement
- Recent bid strategy changes noted, with a learning period accounted for before judging the result
- Spend distribution reviewed. One campaign consuming most of the budget needs a reason
- Wasted spend quantified — cost against search terms that could never convert. This is the number that justifies the audit
Prioritising what you find
| Priority | Description |
|---|---|
| Critical | Any Stage 1 failure. Everything downstream is unreliable until fixed |
| High | Structural problems affecting whole campaigns — mixed brand/non-brand, wrong networks, wrong geo setting |
| Medium | Targeting waste, weak landing page alignment |
| Low | Ad copy refinement, extension coverage, incremental bid adjustments |
A single tracking fault outranks fifty keyword optimisations, because it is producing wrong decisions continuously and at scale. Category-ordered audit templates bury it at position 30 with the same weight as a missing callout.
Two things to be careful about
Do not judge a metric without its intent. A low click-through rate on a broad exploratory term can mean the ad is correctly repelling people who would not buy. The same rate on your own brand name is a problem. Benchmarks without context produce changes that feel productive and cost money.
Do not act during a learning period. Bid strategy changes need time before the resulting data means anything. Auditing an account mid-change and "fixing" what you see restarts the process and guarantees you never observe a stable result.
Working through this list?
Most teams get through the audit and stall on the remediation. That is the part we usually take on.