How to build an annual marketing plan from evidence

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

An annual marketing plan built from evidence is a short document of dated assumptions: what you currently believe about where buyers are being lost, what you intend to do about it, what range of movement would count as progress, and the calendar dates on which each belief gets re-checked against a fresh measurement. It sits at the top of a three-layer cadence. The weekly growth review steers, the quarterly loop aims and fires, and the annual plan chooses the territory both of them operate in. It earns its keep by being the layer that changes least often — and by stating, in writing, the conditions under which it should change early.

What an annual plan is not: a forty-slide deck. Deck-length planning tends to be theater, an artifact optimized for the meeting where it is presented rather than for the year it is supposed to govern. The test is blunt. If nobody could open the plan in July to settle an argument about money or priorities, it was a performance, not a plan.

What the annual layer decides that shorter cadences cannot

The weekly review makes one kill-or-keep call on live work. The quarterly loop locks a baseline, executes a short list, and re-scans against it. Neither can answer the questions that need a year of runway: which parts of the market you will measure at all, how money is divided across the places demand actually shows up, what band of movement the year should produce, and what would make you tear the whole approach up ahead of schedule.

That gives the annual document exactly four jobs — scope, targets, budget, and re-plan points. Everything else in it is commentary, and most of the commentary can be cut.

Scope: name the question sets you will compete on

A year of marketing is bounded by the buyer questions you choose to be findable for. The annual plan names those questions as sets — grouped by segment, product line, or region — and states which sets are in scope for measurement and investment, and which are explicitly out. Scope is where evidence enters first: the sets you commit to should come from research into what buyers in your market actually ask, not from a brainstorm about what you wish they asked. Each set should be pressure-tested against the five-questions framework; the annual layer is where you decide which slice of the market those five questions get asked about.

The discipline lives in the exclusions. A plan that covers every plausible question set covers none of them well, and it teaches you nothing, because thin effort spread everywhere produces no readable signal anywhere. Writing a set out of scope is not neglect; it is the precondition for learning something from the sets you kept.

Targets: ranges with review dates, not point predictions

An annual point target is fake precision with a deadline attached. A single number written in January pretends to knowledge nobody has in January. The honest alternative is a range tied to a mechanism: for each in-scope question set, the band of position change you expect if the planned work actually ships, plus the date that band gets checked. The reasoning behind ranges — and what to do when reality lands outside them — is covered in how to set growth targets without fake precision. The annual plan's specific contribution is scheduling: every target carries the date of the quarterly re-scan that will test it, so no number can drift through the year unexamined.

Budget: fund measured demand surfaces, not last year's habits

The default annual budget is last year's budget with the figures nudged. Evidence-based budgeting starts from a different question: where, observably, do your buyers conduct their research? Search results, AI assistants, communities, trade publications, analyst conversations, events — these demand surfaces can be mapped before money is committed, and the budget should follow the map rather than the org chart of last year's spend.

Two cautions keep this honest. First, the map is dated. Which surfaces matter — including which AI assistants your buyers currently consult, and how those assistants behave — is an observation with a date on it, not a fact about the year; budget lines should inherit the re-plan points below instead of being fixed twelve months out. Second, absence of measurement is not absence of demand. A surface you have never measured belongs in scope as a small, instrumented bet before it belongs in the budget as a mainline.

Re-plan points: schedule the rewrite before you need it

An annual plan that can only be revised through an argument will be revised by whoever argues loudest. Better to schedule the argument. The natural rhythm is one re-plan point after each quarterly re-scan: four dates on which the plan's assumptions are compared with fresh measurement and the document is amended in writing, with the amendment dated and signed. This is the reading a locked benchmark exists to give — the same question set, measured the same way, before and after — and it is what a Magrios re-scan feeds into the annual layer.

Alongside the scheduled points, list the triggers that justify an early rewrite: a major competitor entering the category, a question set proving far larger or smaller than assumed, or a channel crossing the exit criteria you wrote down when you entered it. Pre-committed triggers convert a shock into a procedure.

The document itself: assumptions with review dates

The working format is one table and a few pages of context. The table has six columns: the assumption, the evidence behind it, the action it justifies, the range that would count as progress, the re-check date, and one named owner. If an assumption has no evidence, it stays in the table — labeled as a bet, with a smaller budget and an earlier re-check date. Bets are legitimate; disguised bets are not.

This format is deliberately hostile to theater. A slide can hide an unsupported claim inside a diagram; a table row with an empty evidence cell confesses immediately. The plan's job is to be consulted, argued with, and amended — twelve months of decisions checked against it, four amendments layered onto it. By December it should look used.

The three layers, assembled

Each layer protects the others. The annual plan keeps the weekly review from thrashing between fashions; the weekly review keeps the annual plan from calcifying into scripture; the quarterly loop supplies both with the only thing that settles arguments — measurements taken before and after the work.