Strategic Marketing Planning: Process, Frameworks, Examples

A.Horizon and reversibility, not scale

A strategic marketing plan answers four questions: where are we now, where do we want to be, how will we get there, and how will we know. The first is a situation analysis covering market, competitors, customers, internal capability and a performance baseline. The rest are objectives, strategy, and the execution and measurement that follow. That is the five-stage process, and it is uncontroversial enough that nobody fails because they got the sequence wrong.

What separates the strategic stages from the tactical ones is not importance or budget, it is what happens if you are wrong. Choosing to serve mid-market instead of SMB reshapes pricing, sales motion, support cost and product roadmap, and unwinding it takes quarters. Choosing to publish three posts a week instead of two costs a week to reverse. Both are real decisions; only one of them needs a plan, because only one of them forecloses options.

The load-bearing word is choice. A strategy is defined by what it declines, and a plan that serves every segment through every channel is a wish list with a budget attached. This is also why the strategic layer gets skipped: declining a segment means giving up revenue you can name, in exchange for revenue you cannot yet, and nobody volunteers to own that trade in a room full of stakeholders. The tactical layer asks for no such thing, which is exactly why teams are more comfortable there.

B.Frameworks are exclusion devices or they are decoration

Three frameworks earn their place, and each of them earns it only in a specific form. SWOT is worth using for its intersections and nothing else: how a named strength attacks a named opportunity, and which weakness a named threat exploits. Four lists of adjectives is the version that gets produced and it commits you to nothing.

STP is the one that does the actual work, because targeting is subtraction. Segmentation divides the market, targeting selects, and the selection is only real if someone can name the segments you are now declining to serve and what you will stop doing for them. Positioning then follows from the choice rather than preceding it, which is the usual order error. Our note on what positioning actually is covers that dependency.

Porter’s Five Forces is a diagnostic rather than a plan. Rivalry, new entrants, substitutes, buyer power and supplier power tell you where your leverage sits and where it does not, which is useful precisely when it rules a strategy out: a channel with no barrier to entry will not become a moat no matter how much of the plan points at it.

The test for any of the three is the same. If the output does not name something you have stopped doing, it produced a list, and a list is what the plan was supposed to replace.

A.Turn each choice into a claim with a number and a date

“Mid-market is our segment” is not a claim, it is a preference. The claim underneath it is that mid-market accounts convert at a higher rate, pay back acquisition cost inside a stated number of months, and retain better at twelve, and each of those has a number attached whether or not anyone writes it down. Writing it down costs an hour and converts the plan from a document into an instrument, because a claim with a number can turn out to be false and a preference cannot.

The same discipline fixes projections. Past performance is the only defensible starting point, adjusted for what you are actually changing. Organic traffic that grew 25% last year with a two-person content team does not grow 50% next year with the same team, and saying so out loud is more useful than the target was. Every number in the plan needs a baseline it came from and a specific reason it will move, which is the difference between a forecast and a hope with a decimal point.

Budget allocation is where the absence of claims shows most clearly. Most budgets are set by precedent, which is last year’s number carried forward, or by trend, which is whichever channel is currently being discussed. Neither is a claim about return. The allocation that is a claim states what each channel returned, what you expect it to return at the new spend level, and at what point diminishing returns are expected to bite. A rigorous ROI measurement framework is what makes that statement checkable. Then reserve 10-20% for channels with no history at all, and treat that line as option value rather than as a smaller version of the main budget: its job is to produce information, so it should be judged on what it ruled in or out, not on its return in-period.

B.The evidence arrives late and pre-biased

The problem with a falsifiable plan is that the facts capable of falsifying it show up after the decisions they should have informed. Revenue is a lagging indicator by construction. Worse, the channel differences that matter most are the ones that appear slowest: two sources can look identical for a month and separate decisively by month three, which is after the 30-day attribution window most paid dashboards run on has already closed. A Cohorts view keyed on acquisition source is where that separation becomes visible at all.

The bias runs the same direction every time. Last-touch credit lands wherever the identifying event happened, which is near the bottom of the funnel, so branded search, direct and email accumulate credit for demand that something upstream created. Averaging more data does not remove a systematic bias, and a multi-touch model inherits it because it redistributes credit only across the touchpoints it can see. Our attribution guide works through the model choices.

One channel is currently invisible rather than merely mis-credited. People research vendors inside ChatGPT, Claude, Gemini and Perplexity and arrive already narrowed down, but those visits usually carry no usable referrer, so they land in the direct bucket and the channel reads as though it does not exist. A plan written today that says nothing about AI assistants is not making a judgement about the channel, it is reporting its instrumentation. The LLM Acquisition report separates those humans out of direct and follows them to signup and revenue, which is the minimum required before the channel can be argued about.

So leading indicators are not a nicety inside the measurement stage, they are what makes the plan checkable on a horizon shorter than the damage. For content, that is organic growth rate and content-sourced pipeline; for a product-led motion, trial signup rate, activation rate and time to value. Each should be paired with the lagging number it is standing in for, and the pairing should be written into the plan, so a leading indicator that stops predicting is itself a finding.

A.Four cadences, one of which has teeth

The four reviews are not the same activity at different frequencies. Each is scoped to a different class of claim, and confusing them is why review meetings feel repetitive. Weekly is execution quality: is this campaign on track, which needs work, which should stop. Monthly is KPI movement against target, tracking MRR growth, net revenue retention, acquisition cost and conversion rate by channel, and producing action items rather than commentary.

Quarterly is the only one that can invalidate the strategy, because it is scoped to the assumptions rather than the results. Has a competitor changed the reference price. Has the segment we chose started buying differently. Has the channel we called a moat been entered. Results reviews can only tell you whether the plan is being executed; the assumption review is the only place the plan itself is on trial. It is also the one that gets rescheduled, because it is the meeting where someone has to say the thing they argued for last year is not working. The annual refresh then rebuilds from a year’s evidence rather than from scratch.

B.What makes the quarterly review survivable

The reason assumption reviews get skipped is usually mechanical rather than political. If answering “did mid-market actually retain better” takes two weeks of analyst time, the question is asked once a year at most, and by then the answer is a post-mortem. Cadence is bounded by the cost of producing the evidence, so lowering that cost is a strategic act rather than an operational one. Describing the metric you want in plain language and having the reports and dashboards built from the question, over tracking that comes from autocapture and a setup pass rather than a quarter of instrumentation work, is what puts a quarterly assumption review inside the realm of things a small team can actually do.

The second requirement is a kill rule written before the evidence arrives. For each claim: the number, the date it will be checked, and what result would cause the plan to change. Without it the quarterly review becomes a negotiation about whether the result counts, and that negotiation is won by whoever proposed the strategy. With it, the review is a check against something the room already agreed to, which is a far smaller act of courage. Picking numbers that can actually settle a question is its own skill; our guide to choosing KPIs covers the common traps.

Note what this does to the plan document. If each claim carries a number, a date and a kill condition, the plan is a page of numbered lines, and the situation analysis, frameworks and channel narratives become working papers behind it. That is the correct proportion. The narrative is what was needed to reach the claims; the claims are the only part anyone will check.