Ten working days, no budget increase: that is the deal I would put in front of any AI Google Ads agency taking over a burned account. The replacement earns permission to change more before it asks me to spend more.

I used to tell clients a rebuild plus more spend would fix a flat account. I was wrong. Most burned accounts I inherited were not starved for budget. They were mismeasured, poorly owned, and changed without a record.

Say you are spending $20k a month with zero qualified conversions, and the advice is to double it. More money does not fix that. It buys more confusion at a higher daily cost.

The first ten working days are a test of the handover, not a referendum on mature campaign performance. Here is the plan I would demand, in order.

1. Set the ceiling and cut everything that obscures the test

Cost: no additional media budget. What it buys: a controlled handover instead of another expensive leap of faith.

The media ceiling stays where it is. If the existing monthly ceiling is $20k, the ten-day allowance stays within its daily pro-rata equivalent. Write down that allowance before anyone touches the account, then check actual spend against it. A budget promise in a contract is not spend control.

The constraint removes the new agency’s favorite escape route: asking for more volume before establishing whether the existing numbers mean anything. Same ceiling, cleaner measurement, fewer changes. That gives me a more useful starting point than a simultaneous rebuild and spending increase.

I appoint one approver on my side: one person who can say yes to landing-page edits, tracking fixes, and lead definitions within 24 hours. I have watched takeovers stall because three stakeholders had veto power and nobody owned the form. Without a single approver, ten days becomes six days of waiting and four days of panic.

The cuts are explicit:

  • No rebrand or new offer. They change what the account is trying to sell.
  • No account-wide rebuild. It destroys the starting point before the diagnosis is finished.
  • No expansion to YouTube or Display. It spends attention on a new surface instead of the existing problem.
  • No budget increase. More money is not the admission price for competent account management.

Those projects can wait. First, I want to see whether the replacement can diagnose in writing and act in order.

2. Before day one: secure access, landing-page ownership, and approval

Cost: one uncomfortable afternoon. What it buys: control of the account and a way to leave with everything still working.

First, confirm that my email holds Admin access, not merely Standard access beneath the old agency’s manager account. Then list every connection and its owner:

  • Google Ads manager accounts, or MCCs.
  • The GA4 property and GTM container.
  • Merchant Center, call tracking, and the consent banner.
  • Landing pages, forms, and the person who can publish fixes.
  • Meta portfolio ownership, if Meta also runs.

I do not infer ownership from who created the account. I verify it. The ownership checklist belongs at the top of the handover document: client Admin access confirmed, old MCC removal scheduled for day two, and two company admins on GTM. No takeover starts while those responsibilities remain a shrug.

Landing-page ownership needs more than a name in a spreadsheet. The approver must know who can change the copy, repair the form, and publish the page. Otherwise, the agency can identify the problem on Monday and spend the rest of the test waiting for someone with a login.

Next, I define an emergency in writing. A broken form, dead call tracking, disapproved ads across the core campaign, or spend pacing 30% over the agreed daily ceiling gets a same-day call. Three quiet days, one expensive competitor click, or a CTR dip does not.

That distinction stops both sides from making the handover worse: the agency cannot treat every wobble as permission to rebuild, and I cannot treat every silence as neglect.

I put the agreement in one email before day one:

  1. The media ceiling in dollars per day and the ten working dates.
  2. The single approver and the landing-page owner.
  3. The Admin and account-link checklist.
  4. The emergency definition.
  5. The four reporting lines expected on day ten.

A verbal agreement is too easy to remember differently. No Admin, no ownership, no written agreement: no start date.

3. Days 1–2: freeze the baseline, then test the measurement

Cost: about half a day of focused work to preserve the evidence, followed by tracking checks. What it buys: a starting line neither side can rewrite.

On day one, export the last 90 days of spend, conversions, CPA, ROAS, and CTR at account, campaign, and ad-group level. Then read Change History filtered to the last three months, with a sample back six to twelve months.

I compare the edit log with the old invoices. If I paid for weekly optimization and the log shows two bid tweaks in a month, I want that explained. If it shows 40 frantic changes in the final week before cancellation, I want that explained too.

Save the exports before anyone changes a setting. Without a day-zero baseline, every later improvement claim starts with someone choosing a convenient version of the past.

Day two is tracking before structure. Keep working campaigns running unless there is a written reason to pause them. Then:

  1. Submit a test lead and place a test call.
  2. Match both against what sales actually receives.
  3. Confirm which actions are primary and feed Smart Bidding.
  4. Check for the classic double count: a Google Ads tag plus the same GA4 import.
  5. Remove micro-conversions such as add-to-cart from primary goals if they are steering bids away from the intended outcome.

What the deck calls a tracking audit, I call checking whether the numbers are real.

I treat a 15–20% gap between reported values and CRM revenue as something to investigate, not wave away as noise. The team must explain the gap before using those values to justify more spending.

No new structure until the baseline is frozen and a test lead survives the trip from click to CRM.

4. Days 3–5: close proven leaks without rebuilding the account

Cost: one change-log entry per fix and a small slice of learning stability. What it buys: less waste without throwing away comparability.

Start with the 90-day search-terms report sorted by cost. That is where I look first when a budget has been burned.

Four patterns get my attention:

  • Job-seeker searches such as “jobs” and “careers.”
  • Free and DIY research with no qualified outcomes.
  • Competitor terms the business cannot win.
  • Broad-match terms spending against weak tracking.

Next, check whether “Presence or interest” location settings are buying out-of-market enquiries. Check brand versus non-brand reporting too, so a blended CPA does not flatter the cost of acquiring someone new.

If a term has spent real money with zero qualified outcomes, add the relevant negative and record the date. That is a specific leak with a specific response. It does not need a presentation about a new account architecture.

The full rebuild stays cut. I want a written audit before a rebuild, without changing conversion goals, bidding, and structure in the same week. Change one lever, record the reason, and name the diagnostic signal expected next.

That rule also governs a proposed brand/non-brand split. Identify the reporting problem now; do not use it as permission to change structure while measurement and bidding are already moving.

The changes I will authorize are deliberately narrow: pause one clearly wasteful ad group with its search-term evidence attached, or fix one geography setting with the out-of-area clicks documented. Leave the rest alone until there is a reason not to.

Every entry needs three things:

  • Evidence: the report line or setting that shows the problem.
  • Action: exactly what changed and when.
  • Expected signal: what the team will inspect to judge the change.

Busy is not a diagnostic signal. If the agency cannot point to the evidence behind a change, it does not make the change.

5. Days 6–10: demand a report that separates work from results

Cost: about an hour of honest writing. What it buys: a clear distinction between activity, outcomes, and unresolved dependencies.

I want a one-page note with four reporting lines:

  1. Spend against the ceiling. Not just total spend, but whether the agreed pacing held.
  2. Qualified outcomes against the frozen baseline. Not a fresh definition of success invented halfway through the takeover.
  3. Tracking faults found and fixed. With the test results that support “fixed.”
  4. Dependencies still blocked on my side. Named plainly, not saved for a defensive conversation later.

That last line matters. If landing-page copy is waiting on my approver, or sales has not marked which leads were qualified, the agency writes it down. A blocked dependency should not quietly become next month’s excuse.

Completed work belongs in one column. Uncertain outcomes belong in another. A new negative-keyword list is completed work. Whether it lowers cost per qualified lead is still uncertain. Anyone who mixes those columns in week two will mix them in month six.

The team must also explain reporting lag without using it as a blanket excuse. Google Ads reporting has different refresh times for cost and conversion data, and later conversions can be attributed back to an earlier click. Recent results are not necessarily settled results.

I will accept a campaign-specific explanation with dates attached: what is still arriving, which period remains incomplete, and when the team expects to reassess it. I will not accept “the account is learning” as an explanation for everything from a broken form to an undocumented bid change.

The point is not to force certainty by Friday. It is to stop uncertainty being used as camouflage.

A useful day-ten report says what is known, what is still unknown, and when the team will check again.

6. On day ten: continue, constrain, or exit

Cost: a decision based on the handover, not a premature demand for mature CPA. What it buys: permission to continue without surrendering control.

Continue when delivery and measurement pass

I continue when Admin access is fixed, test leads reconcile to the CRM, primary conversions are clean, and every change has a dated reason tied to evidence.

I will accept flat qualified volume over ten working days if the tracking has gone from fiction to something I can trust. That is not proof of campaign success. It is proof that the next decision can rest on something firmer than a dashboard screenshot.

Continuing does not automatically lift the spending ceiling. It means the team has passed the handover test.

Constrain when the evidence is mixed

If the search terms still show heavy spend on jobs, free, and DIY clicks that never convert, I hold the ceiling and authorize only measurement and negative-keyword work for the next two weeks.

I also check the account against the 10-point audit framework before approving anything bigger. A partly repaired account is not a reason to fund a larger experiment.

Ten days buys the right to decide with cleaner data, not the obligation to scale.

Overhead desk with printed Google Ads change history and a red pen circling a wasteful search term

Exit when access or truthful reporting still fails

If I still lack Admin access, ownership remains unresolved, test leads do not match the CRM, or the change log lists “optimizations” with no evidence behind them, I leave.

The mid-market transition sequence is the relevant comparison: keep the exports and walk early instead of funding month two of fog.

Ten days cannot prove mature CPA. It is enough to show whether a team tells the truth about money in, leads out, and what it touched. A team that fails that test under a capped budget does not earn a doubled one.

When the constraint lifts, expand only what earned it

This is why I favor the groas operating model for the work that follows: models handle daily search-term mining, bid checks, and budget checks continuously rather than waiting for a weekly review. Actions come with their reasons in plain language, and one human owns the outcome. I still keep a single approver on my side.

The fee structure matters too: a fixed monthly fee, not a percentage of spend. Asking me to spend more does not become a way to charge me more.

When the constraint lifts, I keep the ceiling on anything still unmeasured. For the campaign where tracking reconciles and qualified cost per result holds flat or improves, I authorize a 20–30% increase. Every increase gets a written question it must answer over the next two weeks.

Earn permission to change more. Then talk about spending more.