What does a quarterly growth cycle look like?

Week one: measure and lock a baseline — fixed questions, fixed competitor set, fixed scoring. Weeks two to ten: execute three to six prioritised actions, shipped early and logged with dates. Weeks eleven and twelve: re-scan against the locked baseline and sort deltas into moved-as-hoped, moved-for-other-reasons, and did-not-move.

How long until AI visibility work shows results?

Honestly variable. Some changes — a corrected description, a new page entering citations — can register within a single quarter; authority-building outcomes like analyst mentions and review accumulation often take multiple loops. The quarter guarantees you will know what moved, not that everything will.

What if nothing moved after 90 days?

Treat it as diagnostic, not failure. Check the action log first — late shipping is the most common cause. Otherwise the loop is telling you the approach, timescale, or priority was wrong, and the next cycle should change one of those deliberately rather than repeating the same bet.

Should I change my question set each quarter?

Never mid-loop — that destroys comparability. Between loops, evolve it deliberately: retire questions that stopped mattering, add newly discovered blind spots, and keep a stable core so the long-run trend remains readable across quarters.

Is a 90-day loop only for AI visibility?

The structure — locked baseline, short action list, controlled re-measurement — works for any growth surface you can measure repeatedly. AI visibility suits it particularly well because answers drift on their own, which makes undisciplined before-and-after comparisons misleading without a locked benchmark.

How to run a 90-day growth loop

Guide · Continuous Intelligence · 5 min read · last verified 2026-07-27

Reviewed before publication Editorial board Independent commercial review

In shortThe quarterly understand-act-measure cycle: lock a baseline in week one, ship 3-6 actions by week ten, re-scan in weeks eleven and twelve, and read honest deltas.

A 90-day growth loop is a quarterly cycle with three phases: establish and lock a baseline in week one, execute a short list of prioritised actions through weeks two to ten, then re-scan against the locked baseline in weeks eleven and twelve and read what actually moved. It turns growth work from a continuous stream of activity into a sequence of measured experiments. The unit of progress is a measured position change — not a promised revenue number.

Why a quarter is the right length

Ninety days is long enough for cause and effect to plausibly connect. Content shipped in the first half of the window has time to be published, indexed, and picked up by the source pools that AI assistants and buyers draw on. It is also short enough that the priorities set at the start are still the right priorities at the end, and that a failed bet costs one quarter rather than a year.

Be honest about what the length does not guarantee. Some changes surface within weeks; others — analyst mentions, review accumulation, authority building — take multiple loops to register. A quarter is a reporting rhythm that fits how companies already plan, not a physics constant about how fast markets respond. The loop's promise is that you will know what moved, not that everything will move.

Week 1: baseline and lock

The first week is measurement, and it ends with a lock. A usable baseline has four fixed components: the question set (the buyer questions that matter in your category, branded and unbranded), the competitor set you are measuring against, the scoring method, and timestamps. Capture where you stand on each question — presence, position, which sources get cited — and then freeze the lot.

The lock is the point. If questions are added, competitors swapped, or scoring tweaked mid-loop, the end-of-quarter comparison silently becomes fiction — you will be comparing two different measurements and calling the difference progress. Locking is also a commitment device: it removes the temptation, in week twelve, to re-measure with friendlier questions. Keep the metric set small; a handful of numbers you will genuinely re-read beats a dashboard you will not.

Weeks 2–10: act on a short list

From the baseline's gaps, choose three to six actions. Not fifteen — the loop's diagnostic power depends on being able to connect outcomes to actions, and a quarter of scattered effort produces deltas nobody can attribute.

What the actions are depends on your gaps, but they typically look like: shipping or overhauling a comparison page for a question where competitors dominate, publishing a citable asset (original statistics, a definitional guide) for a question where sources are thin, clarifying a pricing or product page that assistants describe wrongly, or closing a presence gap on a review platform that answers repeatedly cite.

Two disciplines matter more than the specific list. Ship early — a page published in week nine has barely entered any source pool by week eleven, so the back half of the window should be for promotion and iteration, not first drafts. And keep an action log with dates: attribution in week twelve depends entirely on knowing what shipped when. An action that cannot be dated cannot be credited.

Weeks 11–12: re-scan and read the deltas

Run the same scan against the locked baseline — same questions, same competitor set, same scoring. Then read the deltas into three honest categories.

Moved as hoped: a gap you acted on closed. Check the timing against the action log before claiming credit, but this is the loop working. Moved for other reasons: answers shifted where you did nothing, or in ways your actions cannot explain — model updates, source churn, and competitor publishing all move answers on their own, and labelling this movement honestly is what keeps the next loop's reasoning clean. Did not move: an action that produced no measurable change. This is not failure; it is the loop telling you the approach, the timescale, or the priority was wrong — each of which is worth knowing at the cost of one quarter.

Resist re-litigating the baseline. The week-one measurement felt fair when it was locked; it does not become unfair because the quarter's results are mixed.

What counts as a result

A result is a measured position change: appearing in answers where you were absent, improved share of voice against the locked competitor set, your pages entering the citations, an incorrect description corrected. These are real, attributable outcomes that compound.

What a 90-day loop cannot honestly promise is revenue. Pipeline lag in most B2B categories is longer than one loop, and visibility is one input among many. Treat anyone promising revenue from ninety days of visibility work with suspicion — and hold your own reporting to the same standard. The defensible claim at quarter's end is 'here is where we moved, here is what we spent, here is the trend', which is also the claim a CFO can actually audit.

Feeding the next loop

The read-out is the next loop's input. Closed gaps rotate out; unmoved gaps get a different approach or a deliberate deprioritisation; movement caused by competitors becomes a tracking priority; and new blind spots surfaced by the re-scan join the queue. This is where the loop compounds — each cycle inherits a cleaner map of what works in your category, which is knowledge no single scan can produce. Magrios structures its understand–act–measure flow around exactly this rhythm: locked benchmark, recommended actions, scheduled re-scan, deltas feeding the next plan. But the loop is a method, not a product feature — a spreadsheet and discipline can run it.

Common failure modes

Five patterns kill most quarterly loops. An unlocked baseline, so the end comparison is unmeasurable. Too many actions, so nothing is attributable. Shipping in week nine, so nothing had time to register. Reading one loop as a verdict on the whole strategy — one quarter is one data point, and direction across loops is the real signal. And metric switching: quietly reporting whatever number improved. The loop only compounds if the measurement is allowed to say no.