How often should someone touch your Google Ads account?

Blocking a junk search term and deciding whether to pause a keyword are not the same activity. One is reading a report that was complete yesterday. The other is reading performance that will not finish arriving for another month. Run both on the same schedule and you are guaranteed to be too slow on one and too fast on the other.

What sets the clock for any class of work is how fast that class of data matures. Search terms land a day late and are then final, so daily is both the floor and the ceiling. A keyword's real performance is unreadable until a 28 day window has closed behind it, because bookings arrive days after the click that caused them. Smart Bidding spends roughly two weeks relearning after every target change, so a target changed weekly is never actually measured.

Move faster than the evidence and you are not optimising. You are adding noise, then reading your own noise back as a result. That is the failure mode nobody sells you, because it looks exactly like effort.

I wrote earlier about why an account needs a cadence at all, and why Google's incentive and yours point in opposite directions. This is the companion piece. Not why, but what: the whole schedule my team works to, class by class, and who decides each one.

Most agencies cannot keep a schedule like this, and the reason is honest: a human with a spreadsheet cannot read search terms across a hundred accounts every morning and pause the right thing on a 28 day window without missing some. So they quietly stretch the loops. Negatives become monthly. Ad reviews become whenever. The client never sees the gap, and the account pays for it.

I solved this the hard way. Over several years I built my own system, purpose made for clinic Google Ads, that does the reading for my team: every search term on every account overnight, every keyword's 28 day window, every ad's impressions, every campaign's bidding state, every landing page rating, every conversion signal. It is the instrument. My team is the one making the decisions, and the instrument means nobody on it ever has to choose which clinic's account gets skipped this week.

Where a class of work is rule bound, brand protection negatives for example, the rule is mine, written with my team from years of clinic search data, and the system applies it the same way on every account, every day, with the working recorded. Where a class of work is a judgment call, a strategist on my team decides it the morning it surfaces, with the reason written down. The system never gets to improvise, and the team never gets to be late.

That is the whole trick, and it is not a trick. It is a marketing team that decided the schedule should be set by the data, and a founder who built what it took to keep it.

Three classes of work run every day on every account I manage, and all three run daily for the same reason. There is nothing left to wait for.

Rule bound negatives come first. Brand protection, price shoppers, obvious junk that has nothing to do with the services you offer. Search terms reach the report about a day after the click, and once they are there they do not improve with age. Daily is the floor because yesterday's junk is already paid for, and daily is the ceiling because checking twice in one day shows you the same data twice. My team's rules run on every account before the clinic opens.

Then the judgment calls. Plenty of search terms are genuinely arguable. A term that reads like a competitor's brand but is also how half the suburb describes the service. A term that is wrong for this clinic and right for the one down the road. Those get decided the same morning they surface, by a strategist on my team, with the reason written down. Every day one sits in a queue is a day it keeps spending.

Third, conversion tracking faults. Checked daily, fixed on detection. Tracking is the one class with no maturation period at all: either the booking was counted or it was not, and every hour a broken signal runs is an hour of bidding trained on a lie.

Most classes of work sit on a weekly clock, and weekly is not a compromise. It is what a week of evidence buys you.

New keywords need seven or more days of search term evidence before you can tell intent from coincidence. Add them daily and you fill the account with terms that got two clicks on a Tuesday. Keyword pauses need longer again: a 28 day window, because conversion lag means a keyword that looks dead on day nine may simply be waiting for its bookings to land. Pausing earlier is guessing, and a pause without a live replacement is just a smaller account.

Ads are the class people get most wrong. My team reviews every ad group's evidence weekly, but the verdict does not move weekly. An ad needs two to four weeks of impressions before its numbers mean anything, so every ad group gets revisited on a four to six week loop rather than churned every Monday. New variations are written by my team, screened for AHPRA before they go anywhere near the account, and shipped as additions. Editing an existing ad in place destroys its performance history, so an edit has to earn itself with evidence in a way an addition never does.

Budget reallocation is weekly and budget neutral: no more than a fifth of the account's budget moves in any week, always from a losing theme to a winning one, always inside the total you have already agreed. Landing pages get a weekly look for one specific signal, which is Google rating the landing page experience below average. And your fifteen minutes sits here too, on a sample of what changed.

Bidding is the class where impatience is most expensive. My strategists review targets weekly and change them at most once a fortnight per campaign, because Smart Bidding spends about two weeks relearning after every change to a target. Change it weekly and every number you read is a relearn rather than a result. The account never settles, and the person managing it never finds out whether the last change worked.

Experiments run monthly for a duller reason: statistical power. A clinic campaign does not generate enough traffic for a split to resolve in a fortnight, and a split that has not resolved returns an opinion dressed up as a result. Monthly gives the test enough behind it to be worth acting on.

Sitelinks and assets are the opposite case. They barely depend on data at all, so they get a monthly sweep plus whatever events demand them: a new landing page, a new service, a site rebuild. There is no evidence to wait for and no benefit in fiddling weekly.

Some work has no cadence, and putting it on one is the mistake. Campaign pauses and enables, and any change to your budget total, are event driven and they are yours. New practitioner starting in six weeks, a room going quiet, a season turning: those are your events, not mine. I recommend, I show the working, you decide. Reallocation inside the total is my job. The total itself is only ever yours.

Landing pages sit in the same bucket at the moment that matters. I watch weekly, and if Google starts rating the landing page experience below average my team builds the replacement. It goes live only with your sign off, because it is your clinic's front door and a page swap is visible to every patient who lands on it.

A site rebuild is the sharpest version of this. The day a clinic rebuilds its website, every ad destination in the account is potentially pointing at a page that no longer exists, and no schedule will catch that in time. That is an event. My team sweeps every destination the day it happens.

An account managed on impulse is easy to spot, and it rarely looks neglected. It looks busy. There is a change log full of activity and a performance line that wanders sideways for a year.

The tell is that nothing in it was ever measured. Targets moved before the relearn finished. Ads were rewritten before they had impressions. Keywords were paused on nine days of data and quietly added back a month later. Every individual change was defensible. Together they cancelled each other out and cost a year.

That is the whole standard I hold my team to, and it is three separate failures, not one. Late is money already gone: a junk search term left running for three days is three days you paid for. Silent is a change nobody can check: work that happened without a reason written down is indistinguishable from work that did not happen. Waiting is the subtle one: a change made before its evidence matured, which reads as diligence and behaves as damage.

So when you ask how often someone should touch your account, the useful version of the question is narrower. How often should each class of work be touched, who decides it, and can they show you the record. If the answer you get is a single frequency, you are talking to someone who has one routine rather than fifteen clocks. If the answer to who decides is anything other than a team for the rules, a strategist for the judgment calls, and you for the money, the schedule was never the real problem.

And if you ask an agency for the schedule they run and they cannot hand you one, you have your answer. This is mine.