See first, then act
The temptation on day one is strong: new creatives, budget moved around, fixing the campaign that looks obviously bad. Do not, before you know whether the numbers you would be steering by are correct.
Almost every account you take over has gaps in tracking. That is a rule of thumb from practice, not a statistic, but it holds in a small account and a large one alike. And it is your first visible win: finding and closing a gap costs no ad spend, and by day three it already proves you know this account better than whoever left it behind.
Step 1: the access checklist
Before you can check anything, you need access. Ask for all of this at once:
- Meta Business Manager: partner access to the ad account, the page, the pixel or dataset, and the catalog if the business sells products
- Google Ads: user access or a manager account link, plus viewer access to Google Analytics 4 and Search Console
- Website or tag manager access
- CRM access, or a recent export if access is not possible right away
- Shop platform access, if applicable
- Existing creative material and a brandbook, if one exists
Step 2: the tracking audit
This is the core of the work. Five checks, each answerable from what is on your screen, no deep platform digging required.
What counts for bidding. Which conversion actions are marked primary and actually feed the bid strategy, and which are secondary and only exist for reporting? A campaign can optimize for the wrong thing for months without anyone noticing.
Double counting. Is the same action, one purchase, one lead, counted by more than one conversion action? Does a pixel event and a server-side event for the same action both fire with no deduplication between them? This is the most common finding, and it makes every cost number look better than it is.
Missing events per funnel step. Is there an event for each meaningful step, or does the funnel jump straight from a page view to the final conversion? Without the steps in between, you cannot tell where people drop off.
Deduplication between browser and server. If both a browser pixel and a server-side connection exist, are they set to deduplicate on the same event ID? If not, the account may be counting some actions twice.
Platform numbers versus CRM or shop numbers. Compare what the platform reports against what the CRM or the shop counts itself, for the exact same period. A small gap is normal: timing, attribution windows, browser blocking. Set your own tolerance before you compare. Anything larger is a finding, not a rounding error.
Report each finding in three lines: what is broken, why it matters, and what the fix is. A list of problems with no fixes is a complaint, not an audit.
Step 3: record the baseline
Before anything changes: spend, your goal metric (CPL or ROAS) and volume per channel, over the last 90 days. This is what every later comparison gets measured against. Record it before you start steering, not after.
Step 4: account file and guardrails
Write a short file with the account IDs, the active structure and known pitfalls. Write a guardrails file alongside it: budget ceilings, CPL or ROAS limits, and a clear line between what a partner or assistant may change alone and what always needs the owner's sign-off first. Do this while the takeover is still fresh. Nobody writes guardrails after the first mistake.
Step 5 and on: naming and kickoff
Propose a naming convention, for example [BRAND]_[Funnel]_[YYYY-MM]_[Concept], so reporting is still readable six months from now. Pull everything into one kickoff document: the baseline, the tracking findings with the fix plan, a 90-day plan, the communication rhythm, and who does what. Close with a monitoring proposal: which metrics, how often, and at what threshold you raise a flag. A proposal, nothing created automatically.
The skill
The Account Takeover Check runs this entire first-day playbook. It has no access to any account, no API connection, no login. It works from what you supply: screenshots, exports, pasted lists.
Download the Account Takeover Check (zip)
In Claude Code, copy the folder to ~/.claude/skills/. In the Claude app, upload the zip in the skills settings. The README in the zip covers both, and a fully worked fictional tracking audit is included in English.
If a problem keeps coming back even after the fix, dig into the process behind it with Five Whys. Once tracking is trustworthy, use the volume gate to see which ads actually win.
What this does not do
The skill never changes anything in an ad account, a CRM or a website, and it never sends anything on your behalf. It does not tell you which menu to click, because that differs by account and by interface language. It tells you what to check and why it matters. Not affiliated with or endorsed by Meta or Google.
One next step
Ask for this account's access checklist today. Once you can answer two of the five tracking checks, you already know more about this account than the previous owner told you.