Check what is being measured before judging how it performs, because a badly configured account makes good work look bad and bad work look fine. Establish conversion tracking, attribution settings and account access first, then record the baseline in writing before changing anything you will later be judged against.
What should be checked first?
Measurement, before performance. A conversion that fires on every page view makes a hopeless campaign look excellent, and no amount of analysis on top of that is worth anything.
Work outward from the conversion definition: what event counts, where it fires, whether it deduplicates, and whether the value attached to it is real or a placeholder someone set once.
What settings quietly distort the history?
The ones that change how results are counted rather than what was done.
Any of these can explain a performance change that looks like a strategy change.
- Attribution window, and whether view-through is included.
- Whether conversions are deduplicated across sources.
- Audience exclusions that silently suppress remarketing.
- Automatic placements or partners the previous owner may not have reviewed.
Why record a baseline before touching anything?
Because after you make changes you can no longer prove what you inherited. A written baseline — spend, conversions, cost per acquisition, over a defined period on a defined window — is the only protection against a later argument about what improved.
Take it as a dated export, not a screenshot of a live dashboard that will keep moving.
What should the access review cover?
Who can change things, and who will still have access next month. Accounts handed over informally usually carry old agencies, ex-staff and a personal account somebody set up years ago.
Getting the client to own the accounts and grant you access, rather than the reverse, avoids the more awkward version of this conversation at the end of the relationship.
Frequently asked questions
- How far back should the audit look?
- Twelve months if the data exists, so seasonality is visible. Anything shorter risks reading a seasonal low as a structural problem.
- Should I fix problems during the audit?
- Record first, then fix, and log what changed on what date. Fixing before recording removes the evidence that the problem existed.
- What if tracking has been broken the whole time?
- Say so before agreeing targets. Committing to improve numbers you cannot yet measure is the fastest route to an argument at the first review.