River
Y CombinatorBacked by Y Combinator
FREE TEMPLATE

Product Strategy Document Template

A statement is a choice only if somebody argued for the opposite. This template goes looking for that person, by name, one line at a time.

Free download  ·  No account needed

A strategy is a set of choices, and a sentence is only a choice if somebody in the building seriously argued for the opposite. That single test separates a strategy from a description of the present tense with "we will" in front of it. No strategy template in circulation applies it, because applying it requires knowing who was in the room and what they argued for.

So the Choice Register carries three columns no template has. Rejected Alternative holds the option that was genuinely considered. Advocated By holds the name, because an alternative nobody argued for was never an alternative, it was a strawman written to be knocked down. Switch Trigger holds the observable thing that would make the choice wrong. Fill all three and it is a choice. Fill none and it is a description, and it will survive every review precisely because there is nothing in it to attack.

In the worked example, eleven statements produced three choices and eight descriptions. The count is the finding: three real choices is a defensible strategy, and three presented as eleven is a document where nobody can tell which three matter. Two of the descriptions contradicted the choices outright, and four of the claims underneath them rested on research nobody had opened since 2023. The strategy is what the roadmap pack follows from and what the prioritization framework scores against, which is why a document of unfalsifiable statements quietly breaks both.

Eleven drafted statements, tested one line at a time

Choice Register, the dates on the evidence underneath it, and the assumptions each choice rests on.

Choice Register

Illustrative, for a fictional construction management product called Meridian. Every statement in the draft, quoted as written, tested on one question: did somebody argue for the opposite.

Statement as draftedRejected alternativeAdvocated bySwitch triggerVerdict
Our primary buyer is the general contractorSubcontractor-first, reaching the GC through the tradesVP Sales>1/3 new ARR non-GC, 2 quartersChoice
We are the system of record for project financialsField execution first, financials stay in the ERPHead of ProductFinancial data arrives via a rival's integrationChoice
We enter Canada and the UK in FY2027Deepen the US Southeast, 38% win rate vs 22%CROSoutheast win rate below 30%Choice
We will deliver an exceptional user experiencenone existsnobody-Description
We will invest in AI to accelerate workflowsnone as writtennobody-Description
We will build an open platform and partner ecosystemnone as writtennobody-Description
We will maintain best-in-class security and compliancenone existsnobody-Description
We will focus on mid-market firms of 50 to 500 staffenterprise, or the long tailnobody this cycle-Description
We will continue to serve existing customers with excellencenone existsnobody-Delete
We will be mobile-first for field usersdesk-first for the office roles who signnobody explicitly-Contradicts financials
We will grow through a product-led self-serve motionsales-led, committee, security reviewnobody this cycle-Contradicts GC buyer

Three choices, eight descriptions. Report the ratio before critiquing any single line, because it is arithmetic rather than taste. The descriptions are not badly written: they are usually the best-written lines in the draft, which is what a sentence with nothing contested in it reads like. They get reclassified rather than deleted, three to operating principles, three to initiatives, two gone. The two contradictions are decisions waiting to happen: a 90-day committee evaluation is not a self-serve motion, and the accountant reading budget-to-actual on two monitors is not a mobile-first user.

Evidence Register

Every factual claim in the draft, with the date of its source rather than the date of the slide. Anything over eighteen months old is flagged, and market sizing is flagged twice.

Claim in the draftSupportsSource dateAgeWhat the current edition says
Market is $9.8B growing at 12%expansionMar 202341 moSame publisher now says 8.4%
80% of specialty contractors use spreadsheetsfinancials202253 moCurrent edition says 54%
Labour shortage of 501,000 workersframing202338 moPublished annually, never refreshed
Largest rival holds ~40% of the GC segmentexpansion202338 moPublic filings, free to recheck
UK firms show lower software adoptionexpansion202426 moTrade article, primary source unknown
Southeast win rate 38% vs 22% averagethe option that lostlast quarter3 moCurrent
20 daily field users per project vs 2 monthly on financialsthe option that lostlast quarter3 moCurrent
GC procurement runs 90 days with committee sign-offresolves contradiction 2last quarter3 moCurrent

Five stale, five current, and every stale claim is external while every current one is internal. That is the pattern: the numbers about the outside world are the ones nobody owns. The 41-month market sizing is the entire quantitative basis for entering two countries, and the strongest current number in the register argues against the choice the company actually made. Both facts belong in the narrative rather than out of it.

Assumption and Risk Log

One question per assumption: what observation would show this is false, and by when will we have it. A threshold and a date, or it is a belief.

AssumptionUnderFalsification test, with its thresholdBy whenIf false
The GC consolidates onto one platform rather than best-of-breed per tradebuyer3+ of the top 40 accounts add an overlapping tool in a yearEnd Q3Scope narrows to what we win outright
Financial data creates a higher switching cost than daily field usageproduct24-month retention gap under 5 points between the two cohortsEnd Q4The choice inverts
A US-shaped product travels to the UK with modest localisationexpansion2 of the first 5 discovery calls cite contract structureEnd Q22027 becomes a build year
Our AI work is differentiating rather than table stakesinitiative3+ named rivals ship equivalent capability in 12 monthsQuarterlyStays funded, stops being called differentiation
Mid-market is not squeezed from abovesegmentEnterprise vendor deal count under 500 staff doublesEnd Q4Segment defence becomes the strategy
Senior domain expertise is hireable at planned cost abroadexpansion2 of the first 3 hires land above bandEnd Q3One country at a time, not two
Buyers pay a premium for an integrated financial recordproductASP at parity or worse against point solutionsEnd Q3Pricing changes, which changes more than the product
Compliance requirements do not fragment further by stateframing2+ states add distinct reporting rules in a yearAnnual reviewEvery dated commitment moves

The cheapest test in the log sits underneath the most expensive choice: five discovery calls settle a two-country expansion, and they should happen before the engineering does. The most consequential test is the retention comparison, because the choice the whole product rests on is contradicted by the freshest number in the evidence register and nobody has yet measured the thing that would settle it. The last row is honest about being a review item rather than a check, which is better than leaving it in the log looking checkable.

What is in the pack

01

Choice Register

One row per drafted statement with the rejected alternative, the name of the person who argued for it, and the observation that would make the choice wrong.

02

The count, not a critique

Choices against descriptions, totalled and reported before any individual line is discussed, because a ratio is arithmetic and a line edit is an opinion.

03

Contradiction pass

Every statement compared against every other, plus against the roadmap, to find the pairs that imply different products. Named and decided rather than hedged into coexistence.

04

Evidence Register

Every factual claim carrying the date of its source, with anything over eighteen months chased to the current edition and the difference written down.

05

Assumption and Risk Log

A falsification test with a real threshold and a date for each assumption, so the ones that are actually permanent get labelled instead of looking checkable.

06

What We Are Not Doing

The rejected alternatives turned around, each with its advocate, its strongest case, a re-entry condition, and the sentence you say when somebody asks.

How it works

  1. 1

    Send everything

    The draft, the board deck, the last strategy nobody followed, the research underneath it. Contradictory versions help rather than hurt.

  2. 2

    Test every statement

    Could somebody have written the opposite, did somebody argue it, and what would make this wrong. Three questions, stop at the first no.

  3. 3

    Date every claim

    Each factual claim gets its source and that source's date, and the stale ones get chased to whatever the current edition says.

  4. 4

    Write the three documents

    The narrative people read, the choices with the losing case stated properly, and the not-doing list with a re-entry condition on every entry.

Frequently asked questions

What do I need before this is useful?

A draft, however rough, plus whatever research the numbers in it came from. Send the contradictory versions too, because the disagreements between two drafts are usually more informative than either one. If you have a blank page instead, start from whatever decision is currently being argued about most: a live disagreement is a choice with both advocates already present.

Are you saying most of my strategy is worthless?

No, and the descriptions are usually the best-written lines in the document. A sentence with nothing contested in it reads cleanly, which is exactly what makes it hard to spot. Descriptions get reclassified rather than deleted: standards like security become operating principles, bets like an AI investment become initiatives under the choice they serve, and only the genuinely empty ones go.

What counts as a rejected alternative?

An option a competent colleague argued for, with their name attached. That name is the whole test: an alternative the author invented in order to reject it is a strawman, and nobody owns a strawman. Where an alternative is live but nobody has said so aloud, go find that person first. It is usually the most valuable hour in the process.

Why does the not-doing section matter so much?

Because trade-offs are what make a position defensible. Porter's What Is Strategy? argues that trade-offs "create the need for choice and purposefully limit what a company offers", and his example is an airline that competed two ways at once and paid a straddling penalty. Same failure as a draft committing to both a committee buyer and a self-serve motion.

How is this different from a strategic plan?

A plan is a list of initiatives with dates and owners, comforting because it describes things you control. Roger Martin's Don't Let Strategy Become Planning observes that most people asking for strategy want that document, where strategy "is manifested as a long list of initiatives with timeframes associated and resources assigned". This pack files the initiatives under the choices they serve.

What happens after the document is written?

The registers stay useful, which is most of the point. When somebody asks why you are not doing something, the answer is already recorded: what it was, who argued for it, why it lost, and what would bring it back. And when the same declined request arrives three times in a month, that is a switch trigger firing rather than three separate conversations.

Does it write the OKRs and the roadmap too?

No, deliberately. Those are downstream and they are separate systems: measurable outcomes belong in the OKR and outcome pack, and where the effort actually went last period comes out of the portfolio and investment review. Collapsing strategy into a roadmap is the most common failure in this genre, because it produces a document that cannot be wrong.

Find out how many of your statements are choices

Send the draft and the research underneath it. The first thing back is the count, and the name of whoever argued for the option that lost.

Test my strategy