product-strategy

name	product-strategy
description	Define a product strategy that connects company vision to where-to-play and how-to-win choices, strategic bets, and quarterly execution. Outputs a strategy one-pager with diagnosis, guiding policy, positioning, bets with kill criteria, and a cascade to OKRs and roadmap.
argument-hint	["company stage","market context","current traction","competitive pressure","time horizon"]
allowed-tools	Read, Write

Product Strategy

A product strategy is an integrated set of choices about where you will compete and how you will win there — choices that reinforce each other and that explicitly exclude alternatives. It is not a vision statement, not a list of goals, and not a roadmap. Richard Rumelt (Good Strategy Bad Strategy, 2011) calls the most common failure "mistaking goals for strategy": "grow ARR 3×" is an ambition, not a strategy. A strategy tells you why you will win and therefore what to stop doing. Michael Porter put the test bluntly in "What Is Strategy?" (HBR, 1996): "The essence of strategy is choosing what not to do." If your strategy document would look sensible with the choices reversed — "we will target the enterprise" vs "we will target SMB" — and nothing else in the document would have to change, you have written a goals memo, not a strategy.

Good product strategy matters because every downstream artifact inherits its quality. Prioritisation frameworks (RICE, WSJF) can only rank options against a value definition — strategy is what defines value. Without it, the roadmap becomes a stack-ranked queue of stakeholder requests, sales-driven one-offs win every trade-off, and the team ships fast in no particular direction.

When NOT to Use This

Premium strategy work has prerequisites. Skip or defer a formal strategy when:

• You are pre-product-market-fit with fewer than ~20 paying customers. You do not yet have the evidence a diagnosis requires. A written strategy will ossify guesses. Run discovery, write a one-page positioning hypothesis, and revisit when retention data exists.
• Leadership is using "strategy work" to avoid a decision. If the real blocker is that two executives disagree about the target segment, a 30-page deck launders the conflict instead of resolving it. Force the decision; then document it.
• The company strategy is genuinely absent. Product strategy cascades from company strategy. If there is none, your first deliverable is a forcing function for the executive team — not a product-level document that invents company direction by stealth.
• You are a single-product, single-segment team under ~15 people with a founder who is the strategy. Write the one-pager to make the implicit explicit; do not run the full quarterly apparatus below.

Strategy vs Vision vs Roadmap

These three are conflated constantly. They answer different questions, change at different speeds, and fail in different ways.

Artifact	Question it answers	Horizon	Changes when	Failure mode when confused
Vision	What does the world look like if we succeed?	3–10 years	Rarely; near-pivot events only	"Vision" used as strategy → inspiring, undecidable
Strategy	Where do we play, and why will we win there?	1–3 years	When evidence breaks the diagnosis	Strategy used as roadmap → dated feature promises
OKRs	Is the strategy working this quarter?	Quarterly	Every quarter	OKRs invented bottom-up → measurement without direction
Roadmap	What are we building, in what order?	Rolling 1–4 quarters	Every planning cycle	Roadmap used as strategy → a queue with a narrative

A roadmap without a strategy is a queue. A strategy without a roadmap is an essay. The cascade section below shows how to keep them connected.

The Decision Framework: Three Choices That Are the Strategy

The structure comes from Playing to Win (A.G. Lafley & Roger L. Martin, 2013), whose strategy cascade is five linked choices: winning aspiration → where to play → how to win → capabilities required → management systems. The middle two — where to play and how to win — are the heart. A third choice, portfolio shape, governs how much you spend defending the core versus finding the next one.

Choice 1 — Where to Play

Posture	Choose when	Honest trade-off
Beachhead: dominate one narrow segment	Pre-~$10M ARR; contested market; evidence of one segment with outlier retention or win rate	TAM story looks small; sales must decline real revenue outside the beachhead
Broad horizontal	Product is genuinely general-purpose AND you hold a distribution advantage (PLG motion, platform, brand)	Shallow differentiation everywhere; collision with the biggest incumbents
Adjacent expansion	Core segment > 50–60% penetrated, or NRR flattening with high logo retention	Splits focus; usually needs a new GTM motion, not just new features
Move upmarket / downmarket	Unit economics broken in the current tier (CAC payback, support load)	Effectively a re-founding: repricing, repackaging, new sales motion, 12–24 months

The most common error is refusing the choice — "our segment is SMBs and mid-market and enterprise" is not a where-to-play. Geoffrey Moore's beachhead logic (Crossing the Chasm, 1991) applies beyond early-stage: dominance of a narrow segment creates references, repeatable sales plays, and pricing power that thin coverage of a broad market never does.

Choice 2 — How to Win

Your how-to-win must name a durable source of advantage, not a feature list. Hamilton Helmer's 7 Powers (2016) is the sharpest catalogue: scale economies, network economies, counter-positioning, switching costs, branding, cornered resource, process power. For product strategy, four dominate:

Winning logic	Works when	Honest trade-off
Counter-positioning	The incumbent's business model (channel, margin structure, cannibalisation risk) blocks it from copying you	The window closes: win before the incumbent's cost of copying falls below its cost of losing
Switching costs / workflow depth	Your product becomes the system of record or embeds in daily workflow	Slow initial land; value compounds late; requires patient capital and low early churn
Network economies	Value genuinely rises with each participant (cross-side or same-side)	Cold-start problem; winner-take-most dynamics punish the #2
Cost / scale economies	Unit costs fall meaningfully with volume and price is a top-2 buying criterion	A race to the bottom for everyone who is not the scale leader

If you cannot state which power you are building and why competitors cannot cheaply neutralise it, you have differentiation-of-the-month, not a how-to-win.

Choice 3 — Portfolio Shape (Bets)

How is investment split between defending the core and finding the next curve? Nagji & Tuff ("Managing Your Innovation Portfolio," HBR, May 2012) found outperforming firms allocate roughly 70% core / 20% adjacent / 10% transformational — a starting anchor, not a law.

Portfolio posture	Choose when	Honest trade-off
Core-heavy (85/10/5)	Core NRR strong and market growing; strategy is share capture	Vulnerable to disruption; growth ceiling is the segment's ceiling
Balanced (70/20/10)	Core healthy but its S-curve visible; adjacent evidence exists	10% transformational is real money that will mostly "fail" — leadership must pre-commit to that
Bet-the-company (50/20/30)	Core is structurally declining or being commoditised	Existential; only with board alignment and 18+ months of runway

Every non-core bet gets explicit kill criteria at funding time (see below). A bet without kill criteria is a permanent department in waiting.

Segmentation That Serves Strategy

Firmographic segmentation — industry, headcount, region — is where most strategy work starts and stalls: it describes the market without revealing where you win. Strategy-grade segmentation cuts the existing customer base by behaviour and economics, hunting for the segment where retention and willingness to pay are outliers. That outlier, plus a structural explanation for it, is a where-to-play candidate; an outlier without an explanation is probably noise.

Cut	Question it answers	Outlier signal worth investigating
Retention / NRR by segment	Where does the product actually hold?	A segment with churn under half the blended rate, or NRR 15+ points above it
Win rate by competitive alternative	Which fight do we win?	Win rate that spikes when the alternative is manual work or a spreadsheet, and craters against a specific rival
Use-case intensity	Who extracts the deepest value?	Accounts using 2×+ the median number of workflows or seats
CAC payback by channel × segment	Where is growth affordable?	A channel-segment pair with payback months below the blended figure
Expansion requests	What do best-fit customers want next?	The same unmet adjacent need appearing across a segment's expansion calls

Two cautions from practice. First, demand a causal mechanism before promoting an outlier to a strategy: Ledgerly's e-commerce segment (Worked Example 1) mattered because settlement-data pain explained the retention gap — a segment that merely happens to retain well this year is a coincidence you are about to over-invest in. Second, segment on data you can keep measuring; a where-to-play defined by attributes you cannot observe in the funnel cannot be operated by sales or growth.

Positioning is then where-to-play made legible to the market. Use Dunford's components in order — competitive alternatives, unique attributes, value with proof, best-fit characteristics, market category — and start honestly: for most products the true competitive alternative is a spreadsheet, an intern, or doing nothing, and positioning against a named rival your buyers never shortlist is positioning for an audience of your own board.

The Strategy Kernel (Rumelt)

Rumelt's kernel is the minimum viable structure for the document itself:

1. Diagnosis — the one or two critical challenges, stated with evidence, that explain the situation. Not a SWOT dump: a judgment about what matters most.
2. Guiding policy — the overall approach to the challenge: your where-to-play and how-to-win, expressed as trade-offs. A useful device is the "even over" statement: "Depth in segment X even over breadth of integrations." Each one names what you will sacrifice.
3. Coherent actions — the 3–5 strategic bets, resourced and sequenced, that implement the policy. Coherent means they reinforce each other; a list of every team's top ask is not coherent action.

Rumelt's signs of bad strategy are a usable review lint: fluff (buzzwords restating the obvious), failure to face the challenge (no honest diagnosis), mistaking goals for strategy, and bad strategic objectives (a dozen "priorities"). Run your draft against all four before sharing it.

Process

1. Write the diagnosis first. Pull retention, NRR, win/loss, and segment-level economics. Interview 5–10 recent wins and losses. Name the 1–2 challenges that explain the data. One page maximum.
2. Segment on behaviour and economics, not firmographics alone. Cut the customer base by retention, NRR, expansion, and use-case intensity. The strategy-relevant segment is where retention and willingness to pay are outliers.
3. Choose where to play. One primary segment/market. Write down the segments you are explicitly not serving and what revenue that forgoes.
4. Choose how to win. Name the power (Helmer) you are building and why the specific competitors in that segment cannot cheaply copy it.
5. Write the positioning. Use April Dunford's components (Obviously Awesome, 2019): competitive alternatives (what buyers use today, often a spreadsheet), unique attributes, value with proof, best-fit customer characteristics, market category.
6. Draft 3–5 "even over" guiding-policy statements. Each must be reversible-sounding — if no sane person would choose the reverse, it is fluff, not a trade-off.
7. Define the strategic bets. For each: hypothesis, resourcing, review date, and kill criteria (a measurable state by a date). Include at least one explicit stop-doing.
8. Cascade to OKRs. Each objective traces to a bet or to the guiding policy; each KR is a measure of the strategy working, with baseline and target. Nothing on the OKR list that the strategy cannot explain.
9. Connect the roadmap. Tag every roadmap item with the bet it funds. Items that fund no bet are either keep-the-lights-on (capped, e.g. ≤ 20–30% of capacity) or cut.
10. Communicate on a drumbeat and review quarterly. One-pager, not a deck, as the canonical artifact. Quarterly: check each bet against its kill criteria; annually or on diagnosis-breaking evidence: rewrite.

The Strategy Cascade

flowchart TD
    M["Mission<br/>Why the company exists"] --> V["Vision<br/>The world in 3-5 years if we win"]
    V --> D["Diagnosis<br/>The 1-2 critical challenges, with evidence"]
    D --> WTP["Where to Play<br/>Segment, market, channel choices"]
    D --> HTW["How to Win<br/>Named source of durable advantage"]
    WTP --> GP["Guiding Policy<br/>3-5 'even over' trade-off statements"]
    HTW --> GP
    GP --> B1["Bet 1<br/>+ kill criteria"]
    GP --> B2["Bet 2<br/>+ kill criteria"]
    GP --> B3["Stop-doing decision"]
    B1 --> OKR["Quarterly OKRs<br/>Baseline -> target per KR"]
    B2 --> OKR
    B3 --> OKR
    OKR --> RM["Roadmap<br/>Every item tagged to a bet"]
    RM --> REV["Quarterly strategy review"]
    REV -->|"Evidence supports bet: double down"| OKR
    REV -->|"Kill criteria hit: reallocate"| GP
    REV -->|"Diagnosis broken: rewrite"| D

Output Format: The Strategy One-Pager

# Product Strategy — [Product/Company] — [Year–Year]
**Owner:** [Head of Product]  **Status:** Draft / Ratified [date]  **Next review:** [quarter]

## Diagnosis (the challenge)
[3–6 sentences. The 1–2 critical challenges, each anchored to evidence:
retention/NRR by segment, win-loss data, market shift. No SWOT dumps.]

## Where We Play
- **Primary segment:** [specific: size band, vertical, geography, channel]
- **Explicitly not serving:** [segments declined, and the revenue forgone]

## How We Win
- **Advantage:** [the named power and its mechanism]
- **Why competitors cannot cheaply copy it:** [1–2 sentences per key competitor]

## Positioning (Dunford)
| Component | Statement |
|---|---|
| Competitive alternative | [what best-fit buyers use today] |
| Unique attributes | [2–3, tied to the advantage] |
| Value + proof | [quantified outcome + evidence] |
| Best-fit customer | [observable characteristics] |
| Market category | [the frame of reference] |

## Guiding Policy — even-over statements
1. [X] even over [Y]
2. [X] even over [Y]
3. [X] even over [Y]

## Strategic Bets
| Bet | Hypothesis | Resourcing | Kill criteria (state + date) | Review |
|---|---|---|---|---|
| 1. [name] | If we [action], then [measurable outcome] | [teams, %] | Kill if [metric] < [threshold] by [date] | [Qx] |
| 2. [name] | ... | ... | ... | ... |
| Stop: [name] | We stop [activity] to fund the above | frees [capacity] | n/a | [Qx] |

## Cascade to OKRs ([quarter])
- **O:** [objective traceable to a bet]
  - KR1: [metric] from [baseline] to [target]
  - KR2: [metric] from [baseline] to [target]

## What would change our mind
[2–3 observations that would invalidate the diagnosis and trigger a rewrite]

Strategic Bet Funding Memo (one per bet)

# Bet: [name]
**Sponsor:** [exec]  **Kill-decision owner:** [named person, ideally ≠ sponsor]
**Funded:** [date]  **Scheduled review:** [date]  **Duration of commitment:** [quarters]

## Hypothesis
If we [action], then [measurable outcome] by [date], because [causal mechanism].

## Link to strategy
Funds guiding-policy statement #[N]; deepens [named advantage/power].

## Resourcing
[teams, headcount, budget — and where it comes from]

## Leading indicators (checked monthly)
- [metric]: [baseline] → [expected trajectory if the hypothesis is true]

## Kill criteria — states and dates, written now, not at review
- Kill if [metric] [</>] [threshold] by [date]
- Kill if [qualitative state — e.g., no design partner converts to paid] by [date]

## Pre-committed reallocation
If killed, people and budget move to [named destination].

## What graduation looks like
[the measurable state at which this bet stops being a bet and joins the core]

The two fields that do the real work are kill-decision owner and pre-committed reallocation. Separating the kill call from the sponsor removes the conflict of interest, and naming the reallocation destination in advance turns a kill from a loss into a transfer — which is what makes it politically survivable (see Fieldstone, below).

Worked Example 1: Ledgerly — Series A, Choosing a Beachhead

Situation. Ledgerly is a fictional 42-person Series A company selling invoicing and AP automation to SMBs. ARR $3.4M, growing 6% month-over-month but decelerating; monthly logo churn 3.1%; sales cycle 68 days; win rate 14% in deals where the buyer compares against the dominant SMB accounting ecosystem's add-ons.

Diagnosis. Segment analysis (step 2) showed the averages were hiding the strategy. Multi-channel e-commerce merchants — selling on Shopify plus at least one marketplace — were 22% of customers but 41% of revenue, with 0.9% monthly churn (vs 3.1% blended) and 128% NRR (vs 96% blended). Win/loss interviews explained why: these merchants drown in marketplace settlement data (fees, reserves, refunds across channels), and incumbent tooling reconciles it in batch, weeks later, usually via an accountant. The critical challenge: Ledgerly is spending a generalist's CAC to lose generalist deals, while accidentally winning a segment it does not name, package, or price for.

Where to play. Multi-channel e-commerce merchants, $1M–$20M GMV, US, acquired primarily through the Shopify App Store. Rationale: this is the only segment where the data showed outlier retention and a structural reason for it — not a segment chosen because it sounded large. Explicitly not serving: services SMBs and single-channel sellers. Sales was directed to disqualify them, walking away from roughly $400K of active pipeline. That number went in the strategy doc deliberately: a where-to-play that costs nothing excludes nothing (Porter's test).

How to win. Counter-positioning (Helmer). Ledgerly's real-time settlement ledger ingests channel APIs directly and recognises revenue continuously. The incumbents distribute through an accountant channel whose revenue comes partly from manual reconciliation work; automating it in real time would cannibalise the channel that sells them. Rationale: the team chose counter-positioning over "more features" because features are copyable in quarters, while the incumbent's channel conflict is a business-model constraint that buys Ledgerly an estimated 2–3 year window. The one-pager stated the window honestly: "This advantage decays; we must own the segment before it does."

Guiding policy (even-over statements).

1. Depth in Shopify and Amazon settlement data even over breadth of integrations (declining 40+ integration requests in the backlog).
2. Merchant self-serve even over accountant channel partnerships (the reverse was genuinely arguable — accountants drive SMB software adoption — which is what made it a real trade-off).
3. Accuracy and trust in the ledger even over new module velocity.

Bets and kill criteria.

Bet	Hypothesis	Kill criteria
Real-time settlement ledger as hero capability	Merchants connecting 2+ channels in trial convert and retain at outlier rates	Kill if < 40% of new e-commerce trials connect a second channel within 14 days, measured at end of Q2
Shopify App Store as primary acquisition channel	Marketplace-led CAC payback beats outbound	Kill if CAC payback > 18 months after two full quarters
Stop: outbound selling to generalist SMBs	Frees 3 AEs and $30K/mo spend to fund the above	n/a — a stop-doing, reviewed for regret at Q3

Cascade to OKRs (two quarters). O: Become the default finance stack for multi-channel merchants. KR1: e-commerce segment ARR $1.4M → $2.6M. KR2: win rate against the real competitive alternative — spreadsheets plus a bookkeeper, per Dunford's analysis, not the incumbent suite — from 31% → 50%. KR3: segment NRR 128% → 135%. Note KR2: positioning work revealed most best-fit buyers were not choosing between Ledgerly and the incumbent at all; the OKR was rewritten to measure the fight Ledgerly was actually in.

Worked Example 2: Fieldstone — Growth-Stage, Escaping Commoditisation

Situation. Fieldstone is a fictional field-service management (FSM) platform: $28M ARR, 240 people, Series C. Annual growth decelerated 72% → 31% → 19% across three years. Three funded competitors discount 25–30% in mid-market deals; blended NRR slid from 112% to 104%.

Diagnosis. Horizontal mid-market FSM had commoditised — feature parity across four vendors, procurement-led deals, price as the tiebreaker. But trades contractors (HVAC, plumbing, electrical) — 34% of the base — showed 119% NRR, used 2.3× more workflows per account, and cited permitting/compliance needs in 71% of their expansion conversations. Critical challenge: Fieldstone is fighting a price war on a battlefield where it has no structural advantage, while sitting on a defensible position it has not fortified.

The where-to-play decision — three options, two rejected.

Option	Evidence	Decision and rationale
Move upmarket to enterprise	1 win in 9 enterprise pursuits; 9-month cycles; suite-class incumbents attached to CRM platforms own the segment	Rejected — no how-to-win existed; "upmarket" was an aspiration wearing a strategy costume
Horizontal platform / app marketplace	Only 11 of 62 integrations exceeded 5% attach; no evidence of cross-side network effects	Rejected — network economies require demand for the network, and the data showed none
Verticalise on trades contractors	119% NRR, 2.3× workflow depth, compliance pain no horizontal rival serves	Chosen — the only option where existing evidence, not hope, showed both retention and a copyable-only-slowly moat

How to win. Switching costs (Helmer), deepened by a compounding content asset: jurisdiction-specific permitting and compliance libraries built per city and county. Rationale: the team chose switching costs over price or brand because the compliance library gets harder to replicate every quarter it grows — a horizontal competitor would need years of jurisdiction-by-jurisdiction work that its broad roadmap cannot justify. This reframed engineering priorities: the "boring" permitting-data pipeline outranked two flashier AI features in the next two planning cycles, because it fed the moat and they did not.

Portfolio shape. Balanced 70/20/10: 70% core FSM (defend mid-market cash flow — the vertical bet is funded by not losing it), 20% trades verticalisation, 10% one transformational bet: Fieldstone Capital, embedded lending for contractor equipment purchases.

A kill criterion firing. Fieldstone Capital's funding memo specified: kill if < $2M originated volume or loss rate > 4% by end of Q3. Q3 actual: $1.1M originated, 6.2% losses. The bet was killed on schedule and six engineers moved to the compliance-library team. Rationale for pre-committing: the exec sponsor was the CEO, exactly the situation where a struggling bet normally survives on seniority. Writing the kill criteria at funding time — Annie Duke's "states and dates" formulation (Quit, 2022) — converted a political fight into a calendar event. The kill was announced at all-hands as the strategy working, not failing.

Communicating the strategy. The canonical artifact was a one-pager, not the 40-slide offsite deck. Every PM spec and quarterly review opened by naming which bet the work funded. A baseline survey found 31% of product staff could state the where-to-play; after two quarters of repetition at all-hands, sprint reviews, and onboarding, 85% could. Fieldstone treated that comprehension number as a leading indicator: a strategy the team cannot recite is a strategy the roadmap will quietly ignore.

Worked Example 3: Harborline — When the Diagnosis Breaks Mid-Cycle

The first two examples show strategies being set. This one shows the review loop in the cascade diagram doing its job — the arrow from quarterly review back to diagnosis.

Situation. Harborline is a fictional customer-support QA platform: $9M ARR, 70 people. Its ratified strategy rested on a how-to-win of superior scoring accuracy — proprietary models that graded support conversations better than the manual sampling its buyers used. Where-to-play: support orgs of 50–500 agents in regulated industries. The strategy's "what would change our mind" section, written a year earlier, included: "General-purpose foundation models reach parity with our scoring accuracy at materially lower cost."

The break. Within two quarters, foundation-model-based competitors launched auto-QA at roughly one-fifth of Harborline's price, and Harborline's own bake-offs confirmed near-parity on standard scoring. Trial-to-paid conversion fell 24% → 15%; two renewals cited a cheaper AI-native alternative. At the quarterly review the leading indicators tripped the pre-written clause. Rationale for acting on it: because the invalidating observation had been written down in advance, the debate was not whether the diagnosis had broken — only what to do next. Teams that skip the "what would change our mind" section end up litigating the evidence itself, usually for two or three quarters longer than the market gives them.

The rewrite. New diagnosis: scoring accuracy has commoditised; the durable asset is not the grade but what happens after it. The choices moved accordingly:

Element	Old strategy	Rewritten strategy	Rationale
How to win	Accuracy advantage (a capability rivals just matched)	Switching costs: QA workflow as system of record — calibration sessions, coaching loops, audit trails regulators accept	Accuracy was a wasting asset; workflow embeddedness deepens with use and survives model parity
Where to play	Support orgs 50–500 agents, regulated industries	Unchanged — narrowed messaging to compliance-audit use cases	The segment still had the outlier retention; only the reason they stayed had changed
Scoring engine	Proprietary models as the moat	Commodity input — swapped to foundation models, cutting model spend ~60%	Counter-position on the former moat: sell "bring any model; we make it auditable" against rivals still selling accuracy
Bets	Bet 2 was accuracy benchmarking content	Killed; funding moved to audit-trail and calibration features	Benchmarking marketing amplifies a fight Harborline had stopped trying to win

The lesson. The rewrite took six weeks, not a year, because the machinery existed: a dated diagnosis, pre-written invalidation triggers, bets that could be individually killed, and a one-pager whose changes could be diffed line by line at all-hands. Strategy durability is not the absence of change — it is a low cost of change when the evidence demands it.

Connecting Strategy to OKRs and Roadmaps

The cascade only works if traceability is enforced in both directions:

• Downward: every objective names its bet; every KR has a baseline and target; every roadmap item carries a bet tag. Untagged work is either keep-the-lights-on (explicitly capped — 20–30% of capacity is typical) or it is cut.
• Upward: quarterly reviews score the bets, not just the KRs. A green KR on a dying bet (activity without validation) and a red KR that taught you the diagnosis was wrong (a productive miss) must be distinguishable — Grove-lineage OKRs (popularised by John Doerr, Measure What Matters, 2018) are a measurement system for the strategy, not a substitute for one.
• Kill criteria are the hinge. Format: state + date — "kill if [metric] below [threshold] by [date]," written at funding time, reviewed on the calendar, with a named decision-owner. Criteria invented at review time will be negotiated to fit the result.

Communicating Strategy

A strategy is only real at decision points — a sales escalation asking for an off-segment feature, a sprint trade-off, a hiring plan. Communication succeeds when the person at that decision point can apply the trade-off without asking leadership. That implies three practices:

• Lead with the diagnosis, not the conclusions. Rumelt's kernel ordering is also the persuasive ordering: people accept a trade-off whose cause they have seen. Announcing "we are dropping segment X" without first showing the churn and win-rate evidence produces compliance at best and quiet sabotage at worst.
• Say the not-serving list out loud, repeatedly. The strategy's first real test is usually a large off-segment deal. If sales hears the exclusions for the first time during that escalation, the strategy loses.
• Repeat past your own boredom, and measure it. Comprehension — can a randomly chosen PM state the where-to-play and the top even-over statements? — is a leading indicator of whether the roadmap will actually follow the document.

Vehicle	Audience	Cadence	Purpose
One-pager (canonical, ratified, dated)	Everyone	Reread at every planning cycle	Single source of truth; diffable when it changes
Strategy narrative (2–3 pages of prose)	Product, engineering, design	Once at ratification; in onboarding	The reasoning and rejected options, not just conclusions
All-hands segment with bet scoreboard	Whole company	Monthly or quarterly	Repetition; public bet status including kills
Bet tags on specs, briefs, reviews	Working teams	Every artifact	Traceability where decisions actually happen
Comprehension pulse (3 questions)	Product org	Quarterly	Detects strategy drift before the roadmap shows it

Anti-Patterns

Symptom	Why it fails	Do instead
"Our strategy is to grow 3× and delight customers"	Goals and fluff, not choices — nothing is excluded, so nothing is decided	Write the diagnosis and the even-over trade-offs first
Where-to-play lists every segment	A market description, not a choice; sales optimises locally and the product blurs	One primary segment, plus a written not-serving list with forgone revenue
How-to-win is a feature list	Features are copied in quarters; no durable advantage is named	Name the power (Helmer) and why rivals structurally cannot copy it
Strategy deck refreshed annually, never consulted	Artifact theater; the real strategy is whatever the roadmap does	One-pager as canonical; bet tags on every roadmap item; quarterly bet reviews
Bets without kill criteria	Failing bets survive on sunk cost and sponsor seniority	States-and-dates kill criteria written at funding time
Strategy set by extrapolating the roadmap	Direction becomes an artifact of last year's backlog	Cascade top-down: diagnosis → policy → bets → roadmap
OKRs written bottom-up, then "aligned"	Teams measure what they were already doing; strategy becomes decoration	Objectives derive from bets; KRs measure strategy progress with baselines
Twelve "strategic priorities"	Rumelt's bad strategic objectives — priority spread is priority absence	3–5 bets maximum; everything else is explicitly deferred
Strategy known only to leadership	Teams cannot apply trade-offs they have never heard	One-pager, drumbeat repetition, and measure comprehension

Checklist

## Product Strategy Review Checklist

### Diagnosis
- [ ] Names 1–2 critical challenges (not a SWOT dump)
- [ ] Every claim anchored to evidence: retention/NRR by segment, win-loss, market data
- [ ] Passes Rumelt's lint: no fluff, faces the real challenge, goals ≠ strategy

### Choices
- [ ] Where-to-play names one primary segment AND what is explicitly not served
- [ ] The not-serving list has a real cost attached (forgone revenue/pipeline)
- [ ] How-to-win names a durable power and why each key rival cannot cheaply copy it
- [ ] Positioning covers Dunford's components incl. the true competitive alternative
- [ ] 3–5 even-over statements; each has a genuinely arguable reverse

### Bets
- [ ] 3–5 bets, each with hypothesis, resourcing, and review date
- [ ] Every bet has kill criteria in state + date form, written at funding time
- [ ] At least one explicit stop-doing decision
- [ ] Portfolio shape (core/adjacent/transformational) is deliberate, not accidental

### Cascade
- [ ] Every OKR objective traces to a bet; every KR has baseline → target
- [ ] Every roadmap item carries a bet tag; untagged capacity is capped
- [ ] Quarterly review scheduled; kill criteria owned by a named decider

### Communication
- [ ] Canonical artifact is a one-pager, ratified and dated
- [ ] "What would change our mind" section exists
- [ ] Comprehension is checked: can teams state where-to-play and how-to-win?

10 Rules

1. Diagnosis before direction — always. A strategy written before the diagnosis is a preference with formatting.
2. If reversing the choice changes nothing else in the document, it was never a choice. Real strategy propagates: flip the segment and the positioning, bets, and roadmap must all break.
3. The not-serving list is the strategy. Budget for it to hurt — a where-to-play that costs no revenue excluded nothing.
4. Name the power or admit you don't have one. "Better UX" is not a moat; "their channel economics punish them for copying us" is.
5. Write kill criteria when the bet is funded, never at review time. Criteria written after the results exist will be written to fit the results.
6. Celebrate a clean kill louder than a mediocre launch. The first killed bet sets the price of intellectual honesty for every bet after it.
7. One page, ratified and dated, beats forty slides. Decks are performed once; one-pagers are consulted. The canonical artifact must fit in working memory.
8. The roadmap is downstream of strategy or it is the strategy. Untagged roadmap items are votes against your own document.
9. Position against what buyers actually use today. Per Dunford, the competitive alternative is usually a spreadsheet, an intern, or nothing — not your feature-matrix rival.
10. A strategy the team cannot recite does not exist. Repeat it past the point of your own boredom, and measure comprehension like any other leading indicator.

References

• Richard Rumelt, Good Strategy Bad Strategy (2011) — the kernel: diagnosis, guiding policy, coherent action; the four signs of bad strategy.
• A.G. Lafley & Roger L. Martin, Playing to Win (2013) — the five-choice strategy cascade; where-to-play / how-to-win.
• Michael E. Porter, "What Is Strategy?", Harvard Business Review (1996) — strategy as trade-offs and fit; choosing what not to do.
• Hamilton Helmer, 7 Powers (2016) — the taxonomy of durable competitive advantage.
• April Dunford, Obviously Awesome (2019) — positioning components, starting from the true competitive alternative.
• Geoffrey A. Moore, Crossing the Chasm (1991) — beachhead segment logic.
• Bansi Nagji & Geoff Tuff, "Managing Your Innovation Portfolio", Harvard Business Review (May 2012) — the 70/20/10 portfolio finding.
• John Doerr, Measure What Matters (2018) — OKRs as practised from Andy Grove's Intel lineage.
• Annie Duke, Quit (2022) — kill criteria as "states and dates"; pre-commitment against sunk cost.