I inherited a wasteful account: the first 30 days
The temptation on day one is to start cutting. Everything looks wrong, some of it obviously is, and cutting feels like progress.
It is also how you delete the one campaign that was quietly producing your best patients. Here is the sequence I actually use, and the reasoning for the order.
On this page
Week 1 — find out what is true
Change nothing. Establish what the account is actually counting.
Open every conversion action and write down what fires it, what its counting setting is, and what share of it is modeled. In inherited accounts this list is usually a graveyard: three overlapping form actions, a call conversion at 60 seconds, a page-view conversion someone added in 2023 and forgot.
Then reconcile against the CRM. Take last month's reported conversions and ask whoever runs intake to find them. The gap is the finding, and it is almost always larger than anyone expects. Check whether GCLIDs are reaching the CRM at all — if not, that is the first thing to build, because every decision after this is guesswork without it.
Also pull change history for six months. It tells you whether you inherited neglect or a series of panicked interventions, and those need different treatment.
Week 2 — stop the obvious bleeding
Only the changes that are safe without attribution data.
- Search terms report, 90 days. Add negatives for queries that are unambiguously not your business — job seekers, students, free-service seekers, competitor research. Safe, because these were never going to convert.
- Geography. Check location settings for "presence or interest" targeting, which quietly buys clicks from people merely reading about your city. Set it to presence.
- Brand exclusions on Performance Max. Stop PMax absorbing people searching your name and reporting them as new acquisition.
- Obvious placement exclusions. Mobile app inventory and content farms in Display, if Display is running at all.
- Fix tracking. Auto-tagging on, GCLID captured on landing pages, hidden field on every form, call tracking storing the click ID.
Notice what is missing: pausing campaigns, changing budgets, restructuring. None of that is safe yet.
Week 3 — structure
Now the CRM data starts arriving and you can see which campaigns produced revenue rather than activity.
The pattern in inherited high-ticket accounts is consistent. A small number of campaigns produce nearly all the confirmed customers, a larger number produce leads that never close, and one or two produce nothing but look excellent on platform metrics. That last group is what has been eating the budget.
Restructure around what the CRM confirms: separate the high-value service lines so they can be bid independently, split brand out so its performance is visible rather than blended, and consolidate the long tail of near-identical ad groups that were splitting data too thin to learn from.
Then rewrite the offer on the landing pages, because in high-ticket healthcare a mismatched offer costs more than a mis-set bid. One form for procedures whose value differs tenfold is the most common finding.
Week 4 — bidding and the uncomfortable report
Move the conversion goal onto the imported CRM outcome, then warn everyone what the report will look like.
Set the imported action as primary, demote form fills to secondary, and re-express targets in cost per acquired customer rather than cost per lead. The number is much larger and it is the real one.
Reported conversions will fall — often by more than half. Cost per conversion will spike on the chart. This is the correct outcome, because the numbers you removed were not describing customers. But a monthly report showing that without context reads as a disaster, and I have watched owners kill a working rebuild in week three of the retraining period because nobody warned them.
Give Smart Bidding 30 to 60 days on the new signal before judging anything. In that window, watch cost per confirmed customer and ignore almost everything else.
What deliberately does not get touched
Restraint in the first month is most of the skill.
| Not touched | Reason |
|---|---|
| Total budget | Scaling before the signal is verified just spends faster in an unknown direction. |
| Low-volume campaigns with unclear outcomes | They may be producing your best customers. Wait for CRM data rather than deleting the evidence. |
| Bid strategies, in week one | Every change restarts learning. Changing them before the conversion definition is fixed wastes the learning period twice. |
| New channels | Microsoft, Meta and everything else wait until the primary channel is honest and profitable. |
| The website, beyond the offer and forms | Redesigns take months and destroy your ability to attribute the change. |
The through-line: the first month buys you a reliable signal. Everything valuable that happens in months two through twelve depends on having it.
VERIFIED EXAMPLE · This is the sequence I used when I inherited a wasteful account at a luxury US healthcare clinic: measurement first, structure third. Cost per admitted patient fell ~78% while spend scaled ~2×, and approved patients per month rose ~9×.
Questions owners ask
Why not cut the obviously bad campaigns on day one?
Because the judgment that they are bad rests on platform metrics, which are the thing you are questioning. In high-ticket accounts the campaign with the worst cost per lead is regularly the one producing the best customers, and once it is deleted the evidence goes with it. Negatives and geography are safe on day one; pausing campaigns is not.
What if the CRM data isn't good enough to reconcile?
Then building it is the first project, ahead of anything in the ad account. A field storing the click ID and a stage that reliably marks confirmed revenue is the minimum. Without those two things every optimization decision is a guess, and you will be having the same conversation in six months.
How long before performance actually improves?
Cost per confirmed customer typically starts moving in months two and three, once bidding has retrained on imported outcomes. Reported platform performance gets worse first. Anyone promising improvement inside 30 days is either fixing something trivially broken or measuring the wrong thing.
Should I tell my team the numbers will drop?
Yes, in writing, before you go live, with the expected magnitude. The most common way a correct rebuild dies is a monthly report showing conversions cut in half landing on a desk with no explanation attached. Set the expectation once and repeat it at the next two reviews.
Can I do this while keeping the current agency?
Sometimes, and it is worth trying. Many accounts are structured badly because the client asked for lead volume and the agency delivered exactly that. Take the reconciliation to them and watch how they respond — resistance to being measured on cost per customer is the tell, not defensiveness about the account.