How often you should hear from your agency

Owners usually judge an agency partly on responsiveness, which is fair, and partly on frequency, which is not. Those are different things. A vendor who answers in four hours and reports monthly is serving you better than one who calls every Friday with nothing new.

The hours in your fee are finite. Every hour on a call is an hour not spent producing. That trade is worth making sometimes and it should be made deliberately.

What is the right baseline rhythm?

Four kinds of contact, at four different frequencies.

Weekly during onboarding, for the first four to six weeks only. Setup produces a lot of small questions and access requests, and pushing them into a monthly cycle stalls the work. This period should end on a stated date, not drift.

Monthly, a written report on a fixed calendar day. The tenth, or the second Tuesday. A report that arrives whenever it is ready arrives eventually not at all.

Monthly, a short call. Thirty minutes, tied to the report, with an agenda. Its purpose is questions and decisions, not narration of a document you can read.

Quarterly, a longer planning session. What changed, what to stop, what to start, what the next quarter looks like. This is the one most likely to be skipped and the one most worth keeping.

Ask what proportion of the monthly hours goes to meetings and reporting. It is a fair question and the answer changes how you should think about frequency.

What does a weekly call actually cost me?

Run it as arithmetic rather than as a preference. Use your own numbers.

The meeting line. A weekly thirty minute call is rarely thirty minutes of agency time. Add fifteen minutes of preparation and fifteen minutes of recap and followup, and one weekly call is about one hour of billed capacity per week, so four hours a month.

Say your retainer represents twenty hours. Four hours is 20%derived of everything you buy, spent on the meeting itself. Move to a monthly call and the same arithmetic gives one hour, or 5%derived. The change is worth fifteen percentage points (derived) of your retainer, moved from talking about the work to doing it.

The reporting line, which is usually larger. Ask how long the monthly report takes to assemble. If the answer is four hours and you have also asked for a weekly summary at an hour each, reporting is eight hours against twenty, or 40%derived. At that point you are buying a reporting service with some marketing attached.

Neither figure is automatically wrong. If you are a multi location operator with partners to brief, that reporting is doing real work. The point is that the number exists, it is knowable, and almost nobody asks for it. Ask, then decide.

What should never wait for the monthly report?

Three things, and they should be written into the agreement.

A material drop. Rankings falling across a group of terms, traffic dropping sharply, or conversions stopping. You should hear within a couple of days, with what they know so far, even if the cause is not identified yet.

A policy or platform problem. A suspended profile, a manual action, a disputed listing, a hosting failure. These have deadlines attached and losing two weeks matters.

Anything blocked on you. If work has stopped because they are waiting on photos or an approval, they should tell you that day, and again the following week. Silence here is the failure mode that damages the most engagements, because you find out at month end that nothing happened and both sides feel wronged.

An agency that hides bad news until the monthly report is training you not to trust the good news either.

Why is a weekly number usually meaningless?

Because most of what agencies report weekly moves less than the measurement noise around it, and AI visibility scores are the clearest case.

Researchers at the University of St. Gallen published the measurement study on this in April 2026. They ran identical prompts across ChatGPT, Gemini, Google AI Mode and Perplexity, daily for about 45 days, and separately ran ten repeats of the same prompt on the same day. Two consecutive days shared only 34% to 42% of their cited sources, so roughly 65% of the sources changed overnight with nothing changing on anyone's website. Same day repeats of the identical prompt overlapped on sources 32% to 43%, and ChatGPT was the least stable engine measured. ChatGPT activated web search on only 42.2% of runs, meaning most runs cited nothing at all.

Their prescriptions are specific and you can hold a vendor to them: at least seven runs per prompt per day, a large and varied prompt set, and reporting on a two to four week rolling aggregate rather than week over week. At seven runs the standard error is about 0.10. So a visibility score moving from 40% to 45% in a week is statistically indistinguishable from no change.

This is the debunk to keep on hand, because it is delivered as good news. A weekly report showing the score up five points is not evidence of work. It is evidence that the measurement was taken twice. Ask the vendor for the number of prompts, the number of runs per prompt, and the confidence interval. If they cannot produce those three, the score is noise with a chart around it.

Kohavi and colleagues, in their KDD 2014 paper on running web experiments at Microsoft, Amazon, LinkedIn and Booking, invoke Twyman's law for exactly this situation: any figure that looks interesting or different is usually wrong. Their own experience at Bing was that real successful changes moved key metrics by 0.1% to 1.0% once diluted to overall impact. A weekly report full of dramatic movement is describing instrumentation, not results.

Where this breaks down: a genuine collapse is not noise. If organic sessions fall by half, or the profile stops appearing entirely, that is outside any noise band and it should reach you within days. The rule is not "report monthly no matter what." It is "aggregate the small movements, escalate the large ones immediately."

Is more contact better?

No, and this is the part that runs against instinct.

Frequent checkins push a vendor toward work that shows well on a short cycle. Small visible tasks get prioritized because there is a call on Friday. The slow, structural work that actually compounds gets deferred because it will not have anything to show.

Some of the most valuable work produces nothing to report for weeks. Getting your business described consistently across the sites that mention you is a good example, since consistency across the web is slow, invisible, and genuinely matters. A weekly cadence quietly discourages it.

The opposite pattern is worth watching for too. Reports padded with high volume, low value activity because it fills a page. Profile posts are the usual candidate. In Whitespark's 2026 survey of forty seven local search practitioners, quantity of Google posts scored 43 and keywords in the profile description scored 41, against 227 for the primary category at the top of the same list. What those posts actually do for ranking is narrower than their prominence in reports suggests, and volume of small items is not evidence of progress. Treat that survey as expert opinion rather than test data, which is how its own author describes it.

If you personally want weekly contact, that is a legitimate preference and some owners genuinely work better that way. Say so at the start and expect it to be priced, because it is real time.

What should the monthly call actually cover?

Four items, in this order, in thirty minutes.

What was produced, against the scope. Two minutes, since it should already be in the report.

What the numbers did against the baseline, including anything that went the wrong way and why.

What is next, specifically, with dates.

What they need from you. This should be the last item and it should always exist.

Bring one question of your own to every call, and make it a why rather than a what. Why this page before that one. Why this category. Asking why is how you learn the business, and there is a way to ask about work you do not fully understand without pretending expertise.

If the calls consistently produce no decisions, cut them to quarterly and take the hours back as production. That is a reasonable thing to propose, and a good agency will agree.

What to do this week

Write your current rhythm on one line. Last report date, last call date, last time you heard about something going wrong before you noticed it yourself.

If the third one is never, that is the gap. Send one email asking for the three event rule to be adopted: material drop, policy problem, blocked work, all reported within two business days.

Run the meeting arithmetic on your own fee. Ask for the hours split, calculate the share going to meetings and reporting, and decide whether that is the trade you meant to make.

Then ask the two questions that any agency should find slightly uncomfortable, including this one. First, what percentage of last month's hours went to meetings and reporting, using the actual figure rather than the plan? Second, when did you last tell me something I did not want to hear, and what was the date? SEOD tracks hours by line and still has to answer the second question from memory like everybody else. An agency that cannot answer either one is not measuring itself, and a vendor who has never delivered bad news in a year of work has either been extraordinarily lucky or is managing you.

Then fix the report date. Pick a day of the month and ask for it to be held. Predictability is worth more than frequency, and an agency that can hold a date is showing you something about how they run.

If you are still choosing an agency, get the cadence into the document. A proposal that says nothing about frequency, reporting dates, or response times has left out the terms you will actually live with, which belongs on the list of things worth questioning in a proposal.

Questions about your current cadence?

Email me at eric@seod.com with how often you currently get a report and a call, and whether you have ever heard about a problem from your agency before spotting it yourself. I will tell you which part of the rhythm is missing and the exact sentence to send to fix it. That is the entire reply, and nothing follows it.

There is more on hiring and working with agencies, including what a real discovery process looks like before any of this starts.