Skip to content
Ashish Kumar AgnihotriOperating Partner

The 90-day plan

The COO 90-Day Plan: A Real Operator’s First Three Months

Every 90-day plan template promises listening tours and stakeholder maps. Real operating work follows a harder arc: diagnose in month one, install in month two, govern in month three — with named artefacts appearing at each stage. This is the plan I actually run when I take an operating seat, including what I deliberately refuse to do early.

The first ninety days decide the ceiling of an operating engagement, and they are usually spent wrong in one of two ways. The eager way: changing things in week two, before anyone has measured what is actually happening — reorganising on anecdote, optimising the loudest complaint rather than the binding constraint. The cautious way: ninety days of listening that produces a beautifully observed document and no operational difference. The arc that works runs between them — diagnose, install, govern — with each month producing artefacts the next month stands on. Nothing in month two is guesswork, because month one measured; nothing in month three is fragile, because month two installed.

The sequence is the same whether the seat is full-time or fractional; what differs is compression and focus. A fractional operator at two or three days a week cannot absorb everything, so the plan is built against a mandate — the two or three operating problems the engagement exists to move — rather than against the whole org chart. That constraint is an advantage in disguise: the discipline of working only on what was diagnosed as binding is exactly what most first quarters lack. By day ninety the question is not “has the new COO settled in” but “which numbers moved, and what runs now that did not run before.”

This plan is not theoretical. It is the arc behind nineteen years of operating work: the diagnosis-first discipline that took a newsroom from a capped output to roughly 400 stories a day — a fourfold lift — by fixing flow before adding people; the installation of staged quality that moved a score from 95% to 99% across 2,000+ campaigns and 450 clients; the governance that cut a billing cycle from roughly two months to fifteen days across 75 entities. The details change with the business; the arc — diagnose, install, govern — has never needed to. What follows is each month in working detail, artefact by artefact.

In depth

What you need to know.

The shape of the plan: diagnose, install, govern

Month one exists to replace opinion with measurement: where work actually flows, where it waits, what the few numbers that matter currently say. Month two converts that diagnosis into installed machinery — an operating cadence, a scorecard, unambiguous ownership — chosen because the diagnosis demanded them, not because they are what COOs install. Month three makes the machinery self-sustaining: quality checks built into the flow, playbooks written down, a governance rhythm that runs without the operator chairing every session. Each phase produces artefacts the next depends on, which is what makes the sequence non-negotiable. Install before you diagnose and you build the wrong things confidently. Govern before you install and there is nothing to govern. The plan is less a calendar than a dependency chain with dates attached.

Days 1–30: follow the work, not the org chart

Diagnosis means tracing how work actually moves — order to delivery, brief to publish, invoice to cash — and finding where it waits, because in most struggling operations the waiting dwarfs the working. The org chart will not tell you this; queues form between boxes, not inside them. So the first weeks are spent following real items through the pipeline, sitting in the meetings where decisions do or do not get made, and pulling the numbers the business already trusts rather than building new dashboards. At Republic World, this was the step everyone wanted to skip — leadership read low output as a headcount problem, and the pipeline map showed stories sitting in invisible queues between desks. The fourfold lift that followed was designed in those first weeks of looking.

Days 1–30: the artefacts month one must produce

By day thirty, three things exist in writing or the month was tourism. First, the diagnostic: a short, board-ready document stating where the operating model strains, what each strain costs — trapped working capital, rework, decisions queued behind one person — and the fix sequence, with reasoning. Second, the baseline: the two or three numbers the engagement will be judged on, captured as they stand today — a cycle time, a first-pass quality rate, a decision latency — because improvement claimed without a baseline is storytelling. Third, the mandate confirmed or corrected: the problems the diagnosis actually found, agreed with the CEO, replacing the problems everyone assumed. That third artefact matters most. A surprising share of first quarters fail by solving the brief instead of the business.

Days 31–60: install the operating cadence

The weekly operating review is the engine of everything that follows, and installing it well is harder than scheduling it. The design: one hour, the leadership that owns the work, the scorecard on the table, and a standing agenda that moves from numbers to exceptions to decisions — every item leaving either decided, or assigned with a date. What it is not: a status theatre where updates are read aloud to people who could have read them silently. The first two or three sessions feel slow because the discipline is unfamiliar; by the fifth, decisions that used to wait weeks for an ad-hoc meeting are closing in the room. Cadence is the difference between an operation that steers and one that drifts between crises — which is why it is installed before anything else is optimised.

Days 31–60: ownership and the scorecard go live

Two installations run alongside the cadence. Ownership: every critical process gets one named owner — not a committee, not a department — with the split between what they decide alone and what they escalate written down. Most mid-size operations have never done this precisely, and the exercise surfaces the gaps where work has been falling between two people who each assumed the other had it. The scorecard: the baselined numbers from month one, live on a single source of truth, updated on a rhythm everyone knows. Deliberately few numbers — a leadership team does not need a dashboard it has to study; it needs a handful of figures it trusts and acts on. By day sixty the operation can, for the first time, see itself weekly. That visibility is what month three governs with.

Days 61–90: govern — quality in the flow, playbooks, board rhythm

Month three hardens the gains. Quality checks move into the flow, staged where errors are most likely and still cheap to correct, weighted by exposure — the logic that protected more than USD 20 million in client billings at a global advertising network, and that lifted a quality score from 95% to 99% across 2,000+ campaigns. Playbooks get written: the critical processes documented as they now run, in editable form the team owns, so the model survives personnel changes and, in a fractional engagement, survives the operator. And the governance rhythm completes: weekly operating review chaired increasingly by the team, a monthly review with the CEO or board on trajectory against baseline, escalation paths tested by real exceptions. By day ninety the machinery should run for a fortnight without the COO touching it.

Quick wins versus foundations: sequencing honestly

Quick wins are real and worth taking — but only the ones the diagnosis surfaces, because they cost nothing structural: an approval step that exists for a reason nobody can state, a report prepared weekly that nobody reads, a decision waiting on a person who never needed to be in the loop. Removing these in month one buys credibility for the harder work, and credibility is operating currency. The quick wins to refuse are the ones that mortgage the foundations: reorganising before flow is mapped, adding headcount to an unmeasured process, tooling purchases as a substitute for ownership, public targets set before baselines exist. The test is simple — a genuine quick win still looks like a win in month six. Anything that will need undoing later is not quick; it is merely early.

The mistakes that sink first ninety days

The recurring failures are worth naming because every one is avoidable. Prescribing before diagnosing — arriving with the playbook already written and fitting the company to it. Boiling the ocean — working the whole org chart instead of the two or three binding constraints, so effort spreads thin and nothing moves visibly. Building parallel machinery — new dashboards nobody trusts instead of numbers they already do. Winning the meeting and losing the floor — impressing leadership while the people who run the work were never enlisted. Skipping the baseline, which makes every later claim of progress unverifiable. And mistaking presence for progress — a calendar full of reviews, artefacts nowhere. The common thread is sequence violated: each mistake is a month-two or month-three move performed in month one.

Questions

Common questions.

Run the arc: diagnose, install, govern. Month one, trace how work actually flows, find where it waits, baseline the two or three numbers that matter, and write the diagnostic. Month two, install the machinery the diagnosis demanded — a weekly operating cadence, a scorecard on a single source of truth, named ownership of critical processes. Month three, harden it: quality checks in the flow, playbooks documented, a governance rhythm that runs without the COO chairing everything. The order is the plan; each month produces the artefacts the next one stands on.

Day 30: a written diagnostic — where the model strains, what it costs, the fix sequence — plus baselined measures and a mandate confirmed against evidence. Day 60: the weekly operating review running with a standing agenda that closes decisions; the scorecard live on numbers the business already trusts; every critical process with one named owner. Day 90: playbooks in editable form, quality checks staged inside the flow, and a monthly leadership review tracking trajectory against baseline. If a first quarter cannot point at these objects, it produced attendance, not an operating model.

Same arc, narrower aperture. A full-time COO eventually absorbs the whole operating surface; a fractional operator at one to three days a week runs the plan against a defined mandate — the two or three problems the engagement exists to move — and deliberately leaves the rest alone. Two further differences: everything is built for transfer from day one, because the artefacts must outlive the engagement; and the fixed cadence forces better mechanics — decisions batch to the weekly review instead of leaking across the week. The constraint tends to sharpen the quarter, not weaken it.

Take the quick wins the diagnosis surfaces; refuse the ones that compete with it. Removing a pointless approval step, killing an unread report, unblocking a decision that waited on the wrong person — these cost nothing structural and buy credibility for the harder installations. What sinks first quarters is the other kind: reorganising before flow is mapped, hiring into an unmeasured process, buying tools as a substitute for ownership. The test: a real quick win still looks like a win in month six. Anything you would have to undo later was not quick — just early.

Two ways, both agreed before day one. Artefacts: does the machinery exist — diagnostic written, cadence running, scorecard trusted, ownership named, playbooks documented? These are inspectable facts, not impressions. Trajectory: are the baselined numbers moving — a cycle time shortening, a first-pass quality rate climbing, decision latency falling? Ninety days is rarely enough to finish the movement, but it is enough to see direction on at least one measure. Presence, meetings attended and decks produced measure cost, not progress. If neither artefacts nor trajectory can be shown, the quarter failed, however busy it felt.

Sequence violations, almost every time. Prescribing before diagnosing — arriving with the solution and retrofitting the company to it. Working the whole org chart instead of the binding constraints, so nothing moves visibly. Building new dashboards nobody trusts instead of using numbers people already do. Skipping the baseline, which makes all later progress claims unverifiable. Engaging leadership while never enlisting the people who actually run the work. And confusing a full calendar with progress. Each is a month-two or month-three move performed in month one — the arc exists precisely to prevent them.

The plan hands over to the operating rhythm it installed. Quarters two and three deepen rather than widen: targets set against the baselines now that trend data exists, the second tier of process improvements sequenced from the diagnostic, the team chairing more of the cadence while the COO moves to the exceptions and the board-facing layer. In a fractional engagement this is also where cadence often scales down — two or three days a week easing toward one or two — because a well-built quarter makes the operator progressively less necessary. That taper is the success condition, not a loss of momentum.

Judge any plan — including this one — against five criteria. Sequenced: diagnosis before installation, installation before governance, with dependencies respected. Specific: artefacts named per phase, not themes and intentions. Baselined: measures captured before anything changes, so progress is verifiable. Focused: built against the two or three binding constraints, with the discipline to leave the rest alone. Transferable: everything installed in a form the team owns by day ninety. Templates with listening tours and stakeholder maps are not wrong, merely insufficient — the best 90-day plan is the one that leaves machinery running, not a document filed.

The next step

A short conversation settles most of this — and a fixed-fee diagnostic settles the rest.

Book a consultationThe diagnostic

Keep exploring

How to hire a fractional COOThe buyer’s guide — the 90-day answer is the interview.COO KPIs and the operating scorecardThe numbers the plan baselines and steers by.The fixed-fee diagnosticMonth one, available as a standalone engagement.When to hire a COOThe signals that a company needs this quarter run.Fractional COO servicesThe engagement this plan runs inside.What is a fractional COO?The role behind the plan, in plain language.