How do you write a product strategy for a startup?

Strategy is the plan to win, written as guiding policies: we will, and we will not. Three to five bets, in sequence, each with the critical question that has to be true for it to hold. The 'we will not' list is the strategy; the 'we will' list is just the roadmap's parents. This is the level where flexible on the details lives.

What are strategic bets and critical questions?

Strategy is the plan to win, written as guiding policies: we will, and we will not. Three to five bets, in sequence, each with the critical question that has to be true for it to hold. The 'we will not' list is the strategy; the 'we will' list is just the roadmap's parents. This is the level where flexible on the details lives.

Why should a strategy say what you will not do?

Strategy is the plan to win, written as guiding policies: we will, and we will not. Three to five bets, in sequence, each with the critical question that has to be true for it to hold. The 'we will not' list is the strategy; the 'we will' list is just the roadmap's parents. This is the level where flexible on the details lives.

Strategy · Product Strategy · Product Thinking

§The question: what is the plan to win?

§The mission said why. The vision said what the world looks like. says how, and how is the level where most teams write a list of good intentions and call it done. This chapter is about writing it so that it excludes.

§ 6.1Guiding policies: we will, we will not#

§Rumelt's kernel has three parts: a diagnosis of the situation, a guiding policy for dealing with it, and coherent actions that carry the policy out. The diagnosis, at this point in , is the evidence framework from Part I. The coherent actions are the roadmap two chapters down. The strategy itself is the guiding policy, and the form that makes it a policy rather than a hope is two lists.

§We will. Three to five , in sequence. Each is a claim about how the company intends to reach the vision, specific enough that a competitor could choose differently.

§We will not. The things the company could plausibly do, that others in the space do, that it is choosing not to. This list is shorter and more important. It is the part a reasonable person could disagree with.

Weak strategy	Strong strategy
"We will build a best-in-class loyalty experience, expand across brands, and monetize through payments."	"We will let the customer build her own profile before we ask anything of the operation. We will have the operation read that profile in one restaurant before it writes anything. We will introduce check-in only once information is flowing, and payments only on top of check-in. We will build tiers on real visits, not tokens. We will not pay customers for their data. We will not launch payments before check-in works. We will not ask the operation to type during peak hours."
"We will use AI to deliver the best answers."	"We will answer the question, with sources, and not be a chatbot. We will not optimize for time spent."

Weak strategy

Strong strategy

"We will build a best-in-class loyalty experience, expand across brands, and monetize through payments."

"We will let the customer build her own profile before we ask anything of the operation. We will have the operation read that profile in one restaurant before it writes anything. We will introduce check-in only once information is flowing, and payments only on top of check-in. We will build tiers on real visits, not tokens. We will not pay customers for their data. We will not launch payments before check-in works. We will not ask the operation to type during peak hours."

"We will use AI to deliver the best answers."

"We will answer the question, with sources, and not be a chatbot. We will not optimize for time spent."

§The second strong example paraphrases the position Perplexity's founder has stated publicly: an answer engine, not a chat , and a refusal to chase engagement metrics that would pull it toward one. Whether or not it is the right strategy, it is a strategy, because a competitor in the same market has chosen the opposite on both counts.

§ 6.2Bets in sequence#

§Three to five bets, not eight, and in order. The sequence is not a timeline; it is a dependency chain. Each bet, if it holds, makes the next one possible, and each is a thing the company is betting the next quarter on.

§Lafley and Martin's frame, where to play and how to win, is useful for the first two bets of almost any startup: the first bet is usually a segment and the second is usually the mechanism by which that segment is won. The hospitality group's first bet was the customer alone, because she depends on no one. The second was one restaurant's operation, reading only, because five of six had asked for exactly that and reading is cheaper to adopt than writing. The third and fourth followed from the first two holding.

§ 6.3Critical questions: what has to be true#

§Every bet carries a question: what has to be true for this bet to hold? Not a goal, a condition. "Customers will complete a profile without an incentive" is what had to be true for the hospitality group's first bet. "Servers will glance at a screen between tables" is what had to be true for the second. "The operation will maintain the data once it flows" is what had to be true for the third.

§The is the link between strategy and everything below it. The goals in the next chapter are how you will know the condition held. The roadmap is what you build to find out. Write the question next to the bet, and a bet whose question you cannot write is not a bet. It is a wish with a sequence number.

§ 6.4Flexible on the details#

§This is the level Bezos meant by details. The mission and vision are held stubbornly; the strategy is where evidence is allowed to change the plan, and should. When a critical question comes back false, the bet changes, or its position in the sequence does, and the roadmap under it is regenerated. Nothing above it moves.

§That is why the sequence matters. A team that has ordered its bets by dependency can lose bet two and know exactly what falls: bets three and four, and nothing else. A team with an unordered list loses a bet and does not know what else it just lost.

§ 6.5Where it comes from#

§Each bet cites its evidence. The first bet in the hospitality strawman pointed at six of six customers completing profiles and the vision's "without being asked." The "we will not pay for data" pointed at the same six and at the mission. The "we will not launch payments before check-in" pointed at the six-of-six failure. A bet with no citation is the room's opinion, and it may be right, but the strawman should say so, so the team knows which bets are evidence and which are conviction.

§ 6.6Committed ahead of the work#

§There is one more property that separates a strategy from a running list of decisions, and Vince Law names it well: a strategy is a set of principles and decisions, informed by reality and caveated with assumptions, that you commit to ahead of development. The word is commit. A bet you can quietly abandon the first hard week was never a bet; it was a preference. Writing the "we will not" list down, in advance, is what makes the commitment real, and it is the thing that keeps a team out of what Law calls the feature factory: shipping tactical work with energy and no guiding policy above it. The ladder exists so that the month's work descends from a commitment made before the month began, not from whatever felt urgent on Monday.

§ 6.7What comes out#

§Two lists. We will: three to five bets, in dependency order, each with its critical question and its citation. We will not: the exclusions, each with the reason. One page. The next chapter attaches a number to each bet.

6. Strategy