Most annual marketing plans at $10M–$50M companies get built the same way: whoever is available opens a spreadsheet two weeks before the board meeting, lists the channels the company already runs, and puts numbers next to them. The board nods. Q1 arrives. One number moves — pipeline is light, or the webinar underperforms — and nothing in the document tells anyone what to do about it. The plan was an artifact, not an instrument.
The fix isn’t a better template. It’s the order of decisions. Plans fail because they start at the tactics and reason backwards to the goal, which means the tactics were never actually chosen — they were inherited. If you’re the founder, CFO, or de facto marketing owner who just got handed the annual marketing plan, this is the sequence. Six decisions, in this order, and the AI question that most plans will get wrong this year.
What number is marketing actually on the hook for?
Start from the revenue number the business has already committed to. Not the one you hope for. The one the board has seen, the one finance is modeling against. Then answer the question almost no company without a CMO has answered explicitly: what share of that revenue is marketing responsible for sourcing?
If nobody has said, that’s not a detail to fill in later. That’s the first meeting. Get the CEO, the head of sales, and whoever owns the P&L in a room and force a number onto paper. Referrals, outbound, expansion, partnerships — none of that is marketing’s line. What remains is. A marketing plan without a CMO fails most often right here, before a single channel gets discussed, because nobody agreed on what the plan was for.
How much to spend against that number is a separate question, and the sizing method — pipeline math, not a percent of revenue — lives in its own post. This one assumes you’ve landed on a budget and cares about what makes the plan hold together once you have one.
How do you turn a revenue number into a plan?
Work the math backwards, using conversion rates you can evidence — not the ones you’d like to have.
Revenue target implies pipeline. Pipeline implies opportunities. Opportunities imply leads. At each step, use your actual historical rates. Pull them from the CRM, however messy. If your close rate on marketing-sourced pipeline was 20% last year, the plan uses 20%, not the 30% you intend to hit after the sales process improves. You can write the improvement in as a separate initiative with its own owner. You cannot bake it into the arithmetic and call it a plan. If you don’t trust the rates your CRM is giving you, start here.
A worked example, with round numbers that are illustrative, not benchmarks:
Say marketing is on the hook for $3M in new revenue. Your average deal is $50K, so that’s 60 closed deals. Historically you close 20% of qualified pipeline, so you need $15M in pipeline — 300 qualified opportunities. If 10% of leads become qualified opportunities, you need 3,000 leads across the year. Now every channel conversation has a denominator. “Should we sponsor the conference?” becomes “can the conference produce leads at a cost that pencils against 3,000?” That’s a different conversation than the one most planning spreadsheets host.
If the backwards math produces a lead number your budget can’t plausibly buy, you’ve learned something in November instead of June. That’s the entire point of doing it in this order.
How do you choose channels — and what does each one owe you?
Only now do you pick channels. And every channel gets two things: a job and one leading indicator you will read monthly.
A job means a specific role in the funnel, stated in a sentence. “Paid search exists to capture demand from buyers already searching for the category.” “The newsletter exists to keep 2,000 mid-funnel contacts warm until they’re in a buying window.” If you can’t write the sentence, the channel is in the plan out of habit, and habit is how budgets die quietly.
A leading indicator means the number that moves before revenue does. Not the lagging outcome — the early signal. Qualified demo requests from paid, reply rate on outbound-adjacent content, sourced pipeline per event. One per channel. When you review monthly, you read that number, and the plan says in advance what a good month and a bad month look like. This is what turns a plan from a list into an instrument: it doesn’t just say what you’ll do, it says how you’ll know. Which numbers actually predict revenue, and which only describe it, is here.
How much do you reserve for tests — and what kills them?
Set aside a defined slice of the budget for things you haven’t proven yet. The exact share matters less than the discipline around it: every test gets a kill condition and a double condition, written before the test starts.
“We’ll run the partner-webinar experiment for one quarter. If it sources fewer than a set number of qualified opportunities, we stop. If it beats our blended cost per opportunity, we double it in Q3.” Written in advance, these decisions are analysis. Written after the fact, they’re politics — whoever championed the test defends it, whoever funded it attacks it, and the meeting produces nothing.
Tests without pre-committed exit criteria aren’t tests. They’re small permanent programs waiting to be forgotten. The cleanest test a mid-market company can run — the holdout — is here.
What are you going to stop doing?
A plan with no subtractions is a wish list.
Somewhere in your current marketing activity there’s a channel that survives on inertia: the trade publication placement nobody has attributed a deal to since it started, the monthly blog cadence that exists because stopping feels like admitting something, the event you sponsor because you sponsored it last year. The annual plan is the one moment of the year when killing these costs nothing politically. Use it.
Name the subtractions in the plan itself, with the money they free up and where it goes instead. This does two things. It funds the tests without asking for a bigger budget. And it signals to everyone reading the plan — board included — that this is a set of decisions, not a compilation.
What review cadence goes in the plan itself?
Most marketing planning processes end at the document. The document should end with the operating system: what gets read monthly, what gets read quarterly, and which number triggers a re-plan rather than a shrug.
Monthly: the leading indicators, one per channel, thirty minutes, decisions only. Quarterly: pipeline against the backwards math from section two — are the conversion rates holding, is the lead volume tracking, do the tests get killed or doubled. And then the tripwire: name in advance the miss that forces a re-plan. If marketing-sourced pipeline runs meaningfully below plan for two consecutive months, you don’t note it in a deck. You reopen the plan, re-run the math, and reallocate. What the quarterly version looks like when it reaches the board is here.
Write this cadence into the plan document. Feedback loops that live in someone’s good intentions don’t survive Q1. Feedback loops written into the plan get honored, because the plan is what everyone agreed to.
Where does AI belong in next year’s plan?
Not where most plans will put it.
The pattern to avoid: a line item called “AI tools,” sized as a category, justified by the general sense that the company should be doing something. That’s how a stack of subscriptions gets bought, three people trial them for a month, and nothing in the funnel changes. Budgeting for AI as a category is budgeting for software, and software was never the constraint.
Instead, name the workflows. Go through the plan you’ve just built and identify which existing work gets faster with AI in it: first drafts of campaign copy, research on target accounts before outreach, repurposing one piece of content into six formats, summarizing sales calls into messaging insight. For each workflow, write down two things — the specific speed or volume gain you expect, and the person accountable for output quality. That second part is the one that gets skipped. AI-drafted work that ships without an accountable human reviewer degrades the brand one competent-sounding paragraph at a time, and nobody notices until a prospect does. How to pick the first workflow is here.
The test for whether AI belongs in your plan: can you point to a line in the backwards math that moves because of it? More content into the mid-funnel nurture, faster turnaround on campaign launches, more accounts researched per week. If it moves a number that’s already in the plan, it earns its place. If it’s in the plan because it’s the year everyone’s supposed to have an AI line, it’s decoration.
Why a shorter plan is a stronger plan
A plan built in this order comes out shorter than the one it replaces. That’s not a compromise. A forty-page deck can’t be argued with — nobody can find the assumption they disagree with, so nobody objects, so the plan gets approved and ignored in the same meeting. A plan that fits the six decisions above on a few pages exposes every assumption: the share marketing owns, the conversion rates, each channel’s job, the kill conditions, the subtractions, the tripwire. Anyone in the room can point at a number and say “I don’t believe that.” Good. That argument, held in November, is worth more than a quarter of polite execution against a document nobody believed.
Building this without a CMO in the seat is genuinely doable — the sequence above is most of the job. Where it gets harder is the judgment inside each step: which conversion rates to trust, which channel jobs are realistic, which test deserves the budget. That’s the work — sitting alongside the business as the senior marketing head it doesn’t have on payroll, holding the plan accountable to the pipeline math month after month. If you’re walking into planning season holding the spreadsheet, you don’t need a bigger document. You need a sharper argument, and a clear read on whether it’s time to bring that seat in.