create-gtm-strategy

Files (1)

• SKILL.md

SKILL.md (first lines)

# GTM strategy

A GTM plan is not a document problem. Almost any competent model can produce a plausible-looking one in a single pass, and that is exactly the trap: plausible GTM plans fail silently, because nothing in them announces which parts are known and which were invented. The value here is in the discipline around the writing, not the section headings.

Three habits do most of the work:

1. **Ask before drafting.** The answers reshape the plan; guessing produces a document that has to be rebuilt.
2. **Label every claim.** Distinguish what is known, assumed, modeled, and targeted. Unlabeled confidence is where strategies go wrong.
3. **Audit the draft against itself.** Numbers that contradict each other are the most common defect in strategy documents and the easiest to catch.

The deliverable is a **document the user can edit and version** — not chat prose, and not a rendered view. Plans get revised over weeks.

---

## Stage 0 — Establish the situation before writing anything

Do not start drafting on a one-line brief. Ask, using a multiple-choice tool if one is available so the user can answer in seconds rather than composing paragraphs.

Ask about what you cannot infer and what would change the plan's structure. Usually four things:

- **Focus.** Which segment, vertical, or market leads? If they name several, that is a finding — most companies cannot run three in parallel, and saying so is more useful than accommodating it.
- **Horizon and shape.** A dated launch, a rolling market entry, or a fixed sprint? This determines whether phases are weeks or quarters.
- **Resources.** Who sells, how many of them, what budget, and what delivery capacity. Capacity is frequently the real ceiling and is frequently omitted.
- **Existing proof.** What has the company already done that a stranger could verify? This is the most commonly under-asked question and often the most decision-changing. See Stage 2.

Skip questions you can answer from context, files, or the conversation. Asking about something the user already told you erodes trust in the rest of the work.

**In unattended or scheduled runs**, ask nothing. Choose the most reasonable interpretation, state it plainly at the top of the plan, and proceed.

Expect answers to invalidate parts of your draft later. When they do, revise the affected sections rather than patching around them — a plan whose premise moved needs its premise rewritten, not a footnote.

---

## Stage 1 — Evidence discipline

This is the part that separates a useful plan from a confident one. Adopt a labeling convention and state it at the top of the document:

- **(fact)** — sourced, with the source named
- **(assumption)** — believed but unproven; should be tested
- **(modeled)** — arithmetic from stated inputs, where the inputs are estimates
- **(target)** — a goal being set, not a prediction
- Unmarked text is a recommendation

Then hold the line on four rules.

**Never invent specifics.** No prices, client names, results, headcounts, timelines, or contract terms that the user has not supplied. Where a number is needed but unavailable, write `[confirm]` and say what it blocks. A plan with three honest blanks is more useful than one with three invented figures, because the blanks become a to-do list.

**Attribute figures the company did not measure itself.** "The client reported a 40% reduction" is honest and publishable; presenting the same number as the company's own measurement is not. Buyers read attributed figures as *more* credible, not less, because attribution names who did the measuring.

**Mark market sizing as modeled.** TAM/SAM/SOM built from assumptions is a shape argument, not a measurement. Say so, show the inputs, and note that the figures should not be used externally without sourcing. Where possible, make SOM a function of the company's actual capacity to deliver rather than a share of market — for most companies the ceiling is what they can serve, not what exists.

**Watch for your own inference running ahead of evidence.** The most common way this work goes wrong is treating something you reasoned toward as something you were told, then building a pillar of the strategy on it. Before any recommendation carries structural weight, check whether its basis was stated or inferred. If inferred, either flag it as an assumption or ask.

---

## Stage 2 — Write the plan

Adapt the structure to the situation. Two shapes recur.

**Product or feature launch** — a dated event, tiered or usage pricing, phases in weeks, adoption and activation metrics, competitive displacement.

**Services, consulting, or agency market entry** — no launch date, project or retainer pricing that resists tiering, phases in quarters, delivery capacity as the constraint, references and proof as the scarce asset.

Pick the shape that fits and say which you picked. When it is genuinely a hybrid, choose the dominant motion rather than averaging the two.

### Sections that earn their place

Adjust to the situation, but these carry most of the value:

**Executive summary.** What to do, why, what not to do, the biggest risk, what happens next. Written last, read first.

**Market and opportunity.** Sizing (modeled, with inputs shown), timing factors labeled as assumptions, and the competitive position stated as a hypothesis unless a real scan has been run. If a scan exists, reconcile the positioning with it — shipping a claim that your own research contradicts is a preventable error.

**Segmentation and ICP.** Priority order with reasoning, qualification criteria, and **anti-personas**. Anti-personas are frequently the most useful part: naming who not to pursue prevents more waste than naming who to pursue. Be specific about why each is wrong — wrong price expectation, wrong buying process, wrong data access.

**Personas.** Typically four in B2B: economic buyer, champion, end-user lead, and gatekeeper. For each: what they are measured on, what they fear, what convinces them, and where they are reached. The gatekeeper is routinely omitted and routinely kills deals.

**Proof and evidence inventory.** *Include this even though most GTM frameworks omit it.* List what the company has actually delivered, and for each: is it nameable, are there numbers, who measured them, and can it be published. Then state the gap plainly. A company whose every figure is client-reported or anonymous has a different problem from one with no delivered work at all, and the strategies diverge sharply. This section frequently reframes the entire plan.

**Value matrix.** Pain point → how the company addresses it → the one-line message. Grounded in named workflows and outcomes, not capability claims. Often the strongest section, because it is the hardest to fake.

**Positioning.** A statement, value pillars with the evidence behind each, and the honest status of that evidence. If a pillar's proof does not exist yet, say so — that converts it from a claim into a build item.

**Pricing.** For products, a tier structure may fit. For services and bespoke work it usually does not, and forcing one misrepresents how the work is sold. The better frame for those: the buyer's problem is not a missing price list, it is that they cannot tell whether this is a small or a large decision, and that uncertainty is what stops the conversation. Solve for self-qualification — a published fixed-price entry offer, a stated floor shown alongside a typical range, and a transparent formula for the larger work. A floor published alone anchors every buyer at the floor.

**Demand generation.** Channels with role, expected time to first result, and cost expressed in the resource that is actually scarce — often founder or seller hours rather than cash. Say explicitly what you are *not* doing and why; a channel list without exclusions is a wish list.

**Content strategy.** Pillars, core messages, and channels. Keep the detailed asset calendar in a separate document if it runs long — it moves on a different cadence from the strategy.

**Sales enablement.** Battle cards, discovery script, objection handling. Include an internal business case the champion can forward unedited — commonly forgotten, disproportionately useful. Be wary of recommending a public ROI calculator for complex services: it implies the engagement is simple.

**Partnerships.** Who refers and why, the first concrete move, and priority. Judge partnerships on sourced opportunities, not meetings held.

**Metrics.** Three levels — business outcomes, GTM performance, leading indicators. For each: definition, why it matters, an initial benchmark, and review frequency. Then name **the one metric that governs**, and explain why the obvious candidates are wrong for this situation. Keep the set small enough to actually review.

**Phased plan.** Match the horizon. Each phase gets a goal, activities, deliverables, and a decision milestone.

**Risks.** Probability, impact, mitigation. Rank by expected damage rather than listing alphabetically.

**Budget.** Percentages against a stated envelope. If the real investment is people rather than cash, say so and size it — a plan that looks cheap because its costs sit in someone else's budget is not cheap.

**Open decisions.** What leadership must choose, with a recommendation, the alternatives, and a deadline. A plan with no open decisions is either finished or dishonest.

### Match depth to the decision

Length should track the size of the bet and the number of people who must act on the plan, not the number of headings in the framework above. A founder making a single market-entry call needs something they will actually re-read; a leadership team allocating across several initiatives needs more. A fourteen-person firm does not need fifteen thousand words.

Two habits keep this honest. When a section has nothing evidenced to say, say so in a line rather than filling it — an unanswered question belongs on the open-decisions list, not padded into a paragraph of plausible prose. And write the executive summary so someone can act on it without reading the rest, because most readers will not.

---

## Stage 3 — Audit the draft against itself

Do this before delivering. It is the highest-return step in the whole process and it takes minutes.

Read the draft as an adversary looking for arithmetic that does not reconcile. In practice the recurring defects are:

- **A metric target that the budget cannot produce.** Compute it. A CAC target divided into the stated budget and engagement count frequently comes out three times over.
- **Two tables describing the same spend differently.** Pick one as authoritative and derive the other from it.

From hugobernardo/skills-library, MIT.

More from hugobernardo/skills-library