What to measure in your first 90 days as CMO

Guide · Enterprise · 5 min read · last verified 2026-07-27

A new CMO's first-90-day measurement plan is the systematic replacement of inherited claims with measured facts. Everything handed to an incoming marketing leader — the dashboard, the agency's quarterly deck, the positioning document, the offhand line that "we're strong in mid-market" — is a claim, produced by people who had reasons to want it believed. Some of it is true. The job of the first ninety days is to find out which parts, using measurements you ran yourself: re-baseline where the company actually appears for its buyers' questions, audit claimed against measured, choose three positions to move by day 90, and re-measure on day 90 with the same instrument.

Inherit the machine, not its claims

The machine is real: a team, a budget, running campaigns, a pipeline with names in it. The claims about the machine are testimony, and testimony from interested parties — the agency describing its own work, the previous leader's deck justifying the previous leader's choices, the sales team's folklore about which content works. Accepting the machine while auditing the testimony is the specific posture of the incoming executive.

This distinguishes the first-90-days problem from two neighbors it gets confused with. It is not market entry — what to do in your first week in a new market covers reconnaissance of unfamiliar terrain, where nobody has told you anything yet. And it is not the operating cadence — how to run a 90-day growth loop covers the repeating cycle a settled team runs. The new-executive problem is stranger than either: familiar terrain, described to you exclusively by people with a stake in the description.

Weeks one to three: re-baseline

Start with the question that no inherited dashboard answers: when this company's buyers ask their questions, where does the company actually appear? Assemble the question set from primary sources — sales call notes, support tickets, onboarding conversations, the phrasings prospects actually use — not from the personas document, which is itself an inherited claim. Then run the scan: for each question, what do search results and AI assistants answer today, who gets cited, and is this company present, absent, or misdescribed?

Record everything with dates and lock the set, following the protocol in how to set an AI visibility baseline. The locked set becomes your personal instrument for the whole tenure: whatever you change in the next ten weeks, day 90's re-scan runs against these exact questions, which makes your first report to the CEO arithmetic instead of anecdote. A market growth intelligence platform — Magrios is built as one — will hold the benchmark and attach an openable source to every finding; the essential thing is that the baseline is yours, dated, and frozen.

Weeks three to six: the claimed-versus-measured audit

Now take the inherited claims one at a time and resolve each against the baseline and whatever other measurement it requires. The working format is a three-column table that will later become the spine of your day-90 report:

Three verdicts, and the third matters most. A claim that is unmeasurable as stated has been steering budget without ever being exposed to evidence, and finding several of them tells you something about the culture you inherited. Deliver the audit as inventory, not indictment — you need this team, and most inherited claims were made in good faith by people who lacked an instrument. The tone that works: "here's what I found, here's what I could and couldn't verify, here's what we'll measure from now on."

Weeks six to twelve: three positions, moved

Choose exactly three positions from the baseline to move by day 90. Three is deliberate: few enough that you personally own each one, enough that a skeptic can't dismiss the results as a fluke. Select for three properties — measurable inside the window with the instrument you built, plausibly movable with the machine as it currently runs, and meaningful to the revenue story the CEO already cares about. A question cluster where the company is absent but a fix is publishable; a misdescription an assistant keeps repeating that correct, well-sourced content could displace; a segment where sales keeps losing to an answer nobody has rebutted.

State each position as a direction, a range, and a review date rather than a point number — the discipline described in how to set growth targets without fake precision. The quick win a new leader actually needs is not launched activity, which any predecessor could also produce, but a measured movement: an answer that cites the company on day 90 where the day-14 scan shows silence. That artifact — before, intervention, after, all dated — is the only quick win that survives scrutiny.

Day 90: re-measure and report

Run the re-scan against the locked set. Then report three things, in order: what the baseline found, what the audit resolved, and what moved. Where a position moved, show the pair of scans. Where it didn't, say which assumption failed — the miss handled as evidence, not embarrassment. Deliver it in a finance-legible format from the start, because this report is the pilot episode of the recurring one: how to report marketing to a CFO describes the four-section structure worth adopting on day one rather than retrofitting in month six. From day 91, the instrument hands over to the team's operating rhythm, and the personal audit becomes a standing loop.

What to leave alone until day 91

The measurement plan is also a restraint plan. The attribution overhaul, the agency review, the rebrand, the martech migration, the full-funnel model — all of these are surgery, and surgery before diagnosis is how new executives convert inherited problems into personal ones. Ninety days of measurement first means every later structural decision arrives with a baseline behind it, and the organization has watched you verify before you cut. The leaders who struggle in the role tend to invert the order: bold moves in month one, justified by the very inherited claims they never checked.