What a SaaS marketing strategy actually is

First, the test: is it even a strategy?

Richard Rumelt's kernel is the most widely accepted test of whether a document is a strategy at all. It has exactly three parts, and all three are load-bearing.

Rumelt also named four failure modes, and they map almost perfectly onto what founders complain about in AI-generated strategy:

• Goals mistaken for strategy — “grow signups 20%” is a target, not an approach.
• Fluff — abstraction that restates the obvious in longer words.
• Failure to face the challenge — no diagnosis anywhere in the document.
• A dog's dinner of objectives — a list of everything, with no trade-off.

MKT1 names the same anti-pattern from the practitioner side: “Random Acts of Marketing” — check-the-box work copied from other companies with no strategic alignment.

Strategy vs plan vs tactics

ICP, defined rigorously

“ICP” is used loosely enough to be useless. The distinctions below have broad practitioner consensus and each one changes what you store.

The clearest formulation: an ICP describes a company; a buyer persona describes a person. Personas live inside your ICP. The ICP guides go-to-market; personas guide messaging, content and sales conversations.

April Dunford's twist is worth honouring, because it disciplines the whole thing: her best-fit customer is defined by characteristics that predict caring about your differentiated value — not by firmographics for their own sake. A firmographic filter that does not predict value-fit is noise.

The buying committee — six roles

Load-bearing for messaging, because the homepage usually speaks to the champion or the user, while the pricing and security pages speak to the economic buyer and the blocker.

• Champion — feels the pain most, advocates internally. Usually an IC or manager, not the economic buyer.
• Economic buyer — approves budget. VP or C-level.
• Technical evaluator — fit and integration. CTO, head of IT, architect.
• End user — daily use; determines whether it sticks.
• Operations owner — owns the process being changed.
• Blocker / gatekeeper — procurement, legal, security, finance.

The anti-ICP is required, not optional

Strong consensus here, best put as: an ICP without disqualifiers is a wish list. Typical axes — wrong tech stack, no budget authority, size outside range, a regulatory blocker, wrong region, a buying motion you cannot service, free or student domains.

The composite field schema

Synthesised from MKT1, FullFunnel and the ICP-scoring literature. Tags show what a website can supply.

Positioning

Dunford's definition: how your product is a leader at delivering something that a well-defined set of customers cares a lot about. Her five components matter less than the order she derives them in — the order is the method.

Geoffrey Moore's template is the same six slots without the derivation attached:

On market category strategy there are three choices — head-to-head in an existing category, “big fish, small pond” (a subsegment you can lead), or creating a new category. Dunford is emphatic that new-category creation is usually the wrong call: roughly 90% of recently-public tech companies positioned into existing markets.

Field schema

Messaging architecture

The “message house” shape is near-universal: one value proposition as the roof, three to five pillars as the columns, proof under each.

Wynter's five layers make it scorable

Peep Laja's team derived these from thousands of message tests. They are diagnostic-ordered: failure at layer N makes layer N+1 irrelevant. That turns messaging from a document into a set of pass/fail gates, each with a named remedy.

Two ideas that punch above their weight

MKT1's Perceptions. Three to five narratives you want to drive, written from the audience's point of view, each of which must fail a single test: could a competitor claim this? If yes, it is not a perception. This is the best generic-output filter in the entire body of research, and it is directly implementable as a validation step.

Moesta's four forces. A switch happens only when push (dissatisfaction with today) plus pull (appeal of the new) outweighs anxiety (fear of the new) plus habit (comfort of today). Founders over-index on push and pull and ignore the other two — which is precisely why objection handling and risk reversal belong in the schema as first-class fields rather than an appendix. The forces that lose deals are the ones nobody writes down.

Channel strategy

Two credible camps, and they genuinely disagree. Both force a trade-off, which is the part that matters.

Bullseye — rank, then focus

From Traction. Three rings: brainstorm all 19 channels and generate a plausible idea for each (the discipline is that you may not skip one because it “feels wrong”); rank the ~3 most promising on expected impact, confidence and cost-of-test; then commit resources to the one showing real signal. The core claim is that at any given stage one channel dominates, and the job is to find it cheaply.

Tests are structured: hypothesis (if / then / because), action steps, a fixed deadline, and a success metric chosen in advance.

MKT1 — build a mix

Explicitly contradicts Bullseye. Channels reinforce each other like a team with different roles, organised as six growth engines rather than 19 tactics: Inbound, Outbound, Product Virality, Events, Ecosystem, Lifecycle. Selection is driven by GTM motion — self-serve leans inbound and product-led; sales-led leans outbound, events and ecosystem.

Field schema

Stage advice is consistent across sources: under roughly $1M ARR, founder-led outbound, referrals and founder-led social dominate. SEO and content pay back over 6–18 months and need sustained investment. Pre-PMF, the purpose of a channel is learning, not volume. The most-cited mistake is picking channels because they worked at a previous company.

Content strategy

Content pillars are the end of a chain, not the start of one. Getting the chain right is what stops a content calendar from masquerading as a strategy.

Three distinctions that are easy to collapse and expensive to collapse:

• Messaging pillar ≠ content pillar. A messaging pillar is a claim about you; a content pillar is a topic area you have authority in. They overlap; they are not the same object.
• Perception ≠ content pillar. Pillars are chapters; perceptions are throughlines that run across chapters.
• Strategy ≠ calendar. A content calendar presented as a strategy is the single most common fake deliverable in this category.

Two MKT1 rules worth encoding: the 30% juice rule — about 30% of content is product-centric, 70% is problem, market and vision — and mileage, which splits repurposing (one idea, many assets) from distribution (many channels). Both are required, both are planned at creation time, and there are no dead ends: every asset leads to a next one.

Roadmap item fields — adopt verbatim

With coverage checks across the set: ICP coverage map, funnel-stage distribution, and the 30% juice ratio.

What a website can and cannot tell you

This is the constraint that should shape a website-input product more than any framework on this page. Most of a strategy is derivable. The parts that are not are not evenly distributed — they cluster in exactly the places that make a strategy specific.

The three structural gaps

These are not gaps you can close with a better crawler or a better prompt. They are gaps in what a website is.

Five founder questions that unlock the rest

Ranked by value added ÷ effort to answer. Everything else can be inferred and marked as hypothesis.

1. What do customers use instead of you — including doing nothing?Unblocks the first and most important step of positioning, and it is the one thing a website structurally cannot supply.
2. Describe your last three won deals and last three lost deals, one line each.Yields best-fit characteristics and disqualifiers from real evidence rather than inference.
3. What triggered your best customers to start looking?The single largest ICP gap. Turns a static profile into a targetable moment.
4. Which customer type gets value fastest?Resolves prioritisation — the ranking a website can never give you.
5. Who must sign off, and who tries to block it?Populates the buying committee and, with it, the objection set.

Where the experts genuinely disagree

These are not ambiguities to resolve by picking the more popular source. Each one is a real fork with consequences for what you build.

Actionable vs useless

What founders actually complain about in AI-generated strategy, and the design rules that fall out of each complaint.

The complaints

• Trendslop — whatever buzzwords are currently popular, in place of a trade-off.
• No trade-offs — it lists everything. Rumelt's dog's dinner.
• No business context — generic because the input was generic.
• Reformulated inputs — it hands back your own website copy as if it were insight.
• Undifferentiated claims — fails the competitor-claim test on every line.
• Stops at “what” — no how, no owner, no first step, no metric.

The rules

1. Emit a diagnosis, not a description.One named primary obstacle, argued from evidence.
2. Force at least one explicit trade-off.A deprioritised segment, a skipped channel, a claim abandoned. A required field with a validator.
3. Cite the evidence.Every derived claim links to the page and sentence it came from. The strongest available answer to “this is generic” — it demonstrably is not, because it quotes their own site.
4. Run the competitor-claim test on every message.If a competitor could say it, flag it.
5. Separate asserted from verified.Website-derived hypothesis vs founder-confirmed vs customer-validated. Never present an inference as a finding.
6. Give every pillar a first action.With an owner, an effort estimate, and a success signal.
7. Name the alternatives specifically.Wynter’s data is unambiguous: unnamed alternatives mean a failed differentiation layer.
8. Refuse to invent.An empty, named gap beats a confident fabrication every time.

The composite object model

Everything above, assembled. The shape of a strategy that would survive a founder reading it closely.