DIGNALEGI

Worth reading. A personal index selected by Ratipcan Uysal.

Digna Legi guide / Product strategy

How to Build a Product Strategy That Earns Its Next Commitment

Build a product strategy that links causal assumptions to evidence, stages commitment, and defines when to continue, revise, or stop.

By Ratipcan Uysal12 min read

A torn paper route splits at a mechanical test press; one branch continues and the other stops at a yellow gate.

A product strategy can be falsifiable and still be badly designed.

Imagine a team that makes a clear prediction, funds a twelve-month roadmap, and agrees to judge the result at the end. The strategy can be proven wrong. It also requires a year of payroll, opportunity cost, and organisational attention before the important evidence arrives. By then, accuracy is a poor consolation.

The stronger question is not only:

What evidence would prove us wrong?

It is also:

What can we learn before the next irreversible commitment?

Most strategy documents separate those questions. The narrative explains why the company should act. The roadmap specifies what the team will ship. The budget commits people and time. The measurement plan appears at the end, after the major choices have already been made.

That order is backwards. Evidence should govern commitment, not decorate it.

The central failure in product strategy is commitment outrunning evidence.

A useful product strategy is an allocation argument under uncertainty. It identifies a consequential change, chooses where to play, explains how the product should alter customer behaviour, and names what the company will give up. It then connects increasing levels of commitment to increasing levels of evidence.

The objective is not to eliminate uncertainty. It is to make uncertainty cheaper, earlier, and capable of changing the decision.

Start with the choice, not the aspiration

“Win the enterprise,” “become the platform,” and “lead with AI” describe preferred futures. They do not yet allocate anything.

Michael Porter’s What Is Strategy? makes choice and trade-offs central to strategy. A distinct position requires a different system of activities, and that system gains meaning partly from what the company declines to do. Todd Zenger adds a useful second lens in What Is the Theory of Your Firm?: a strategy should contain a theory about how a company’s resources and capabilities will create value.

For a product team, those ideas produce three questions:

  1. Which customer and problem deserve concentrated attention now?
  2. Why is this company unusually able to change that situation?
  3. What work, customer, or opportunity receives less attention as a result?

The third question is the test. If every existing priority survives, the strategy has added vocabulary without making a choice.

This is why diagnosis must precede the roadmap. Reforge's Product Work Beyond Product Market Fit separates several kinds of product work that are often treated as one queue. A feature problem asks whether a capability creates enough lasting customer value. A growth problem asks what prevents existing value from spreading. A scaling problem asks which bottleneck will constrain the product or organisation. An expansion problem asks why the company can win in a new market or product.

Those problems demand different evidence. A retention problem will not yield to a launch count. A distribution problem will not be solved by improving a feature that customers never encounter. A scaling problem may require operational proof before it needs a new proposition.

Write the choice in a sentence:

We will concentrate on this customer and constraint, because this change in the world creates an opening that our specific capability can exploit. We will reduce or defer this competing work.

If the last field remains empty, the choice is unfinished.

Turn the choice into a causal chain

A strategy needs more than a position. It needs an account of how the position can produce a result.

Use this structure:

For this customer, facing this constraint, changing this part of the product through this route to adoption should produce this observable behaviour, which should create this business consequence.

Each field closes an escape route.

  • Customer prevents “everyone” from passing as a segment.
  • Constraint demands a diagnosis rather than a category such as onboarding or collaboration.
  • Product change identifies the mechanism rather than naming a technology.
  • Route to adoption explains how the customer encounters and begins using the change.
  • Behaviour specifies an action that can be observed.
  • Business consequence explains why the behaviour deserves investment.

The route to adoption is routinely omitted. Brian Balfour’s Product Channel Fit argument is useful here: products adapt to the constraints of their channels. A collaborative product expecting invitation-led growth needs value that can be shared quickly. A product relying on search needs demand, indexable supply, and a reason that supply persists. A high-touch enterprise product needs economics that can support sales, security review, implementation, and service.

A causal chain should therefore cross the boundaries between product, distribution, and economics. “Customers will love it” stops at sentiment. “The feature will increase engagement” names a metric without explaining why it moves. “Sales will sell it” transfers the missing mechanism to another function.

A roadmap lists outputs. A strategy explains why those outputs should change behaviour.

Draw the chain in plain language. Then attack its weakest link.

Find the assumption that can break the decision

Teams often produce long assumption registers. The list creates coverage and hides priority. A typo in an integration name and a false belief about customer willingness to switch do not deserve equal attention.

Rank assumptions using three dimensions:

  • Consequence: If this is false, does the strategy fail or merely need adjustment?
  • Uncertainty: What credible evidence already exists?
  • Irreversibility: How much will be committed before the assumption can be corrected?

The most dangerous assumption combines high consequence, weak evidence, and expensive reversal. That is the assumption the strategy should test first, even when another part of the plan is easier to build.

This changes what “small” means. A small release is not necessarily a small test. A polished prototype shown to friendly customers may be cheap to produce yet incapable of testing switching behaviour. A manual service may look operationally ugly while testing whether the customer will complete the full workflow and pay for the outcome.

Progress equals important uncertainty removed per unit of commitment, not volume shipped.

Rita McGrath and Ian MacMillan’s Discovery-Driven Planning was developed for ventures in which conventional forecasting is unreliable. Their core move is to make assumptions explicit and plan from required outcomes backwards, with checkpoints that expose mistaken assumptions before the full plan is spent. It is a planning method rather than controlled evidence that one process wins everywhere, but its logic is sound for product bets with sparse historical data.

Ask of every proposed milestone:

Which consequential assumption becomes easier to believe or reject when this milestone is complete?

If the answer is “none,” it is a delivery milestone, not an evidence milestone. The distinction matters because delivery can proceed flawlessly while the strategy remains untested.

Make commitment wait for evidence

Evidence does not arrive in one form. A useful sequence moves from cheap signals to costly proof while keeping each step close to the disputed mechanism.

Problem evidence
Observe the constraint in real behaviour, existing workarounds, failed attempts, or costly delay. Interviews can reveal language and context, but stated enthusiasm is weak evidence of future adoption.
Mechanism evidence
Test whether the proposed change alters the behaviour it is meant to alter. This may require a concierge service, a narrow workflow, a prototype with a real commitment, or a deliberately limited product.
Repeatability evidence
Check whether the result survives across customers, time, and ordinary operating conditions. A heroic founder-led success may prove possibility without proving a system.
Economic evidence
Determine whether acquisition, delivery, support, risk, and retention make the result worth scaling. A useful behaviour can still be a poor business.

These are not universal gates. Some regulated or infrastructure products need feasibility and safety evidence before customer testing. Network products may need a minimum level of participation before their mechanism can be observed. Enterprise sales cycles may make a complete economic test slow. The sequence should follow the causal chain, not a ritual.

The governing principle is stable: larger and harder-to-reverse commitments require stronger evidence.

For each stage, specify:

  • the assumption being tested;
  • the observable result;
  • the maximum spend of time, money, and attention;
  • the decision that the result unlocks;
  • the owner and decision date.

Evidence has value only when it changes what the company is prepared to commit.

Without a linked decision, a metric becomes commentary. Teams collect it, discuss it, and continue with the existing allocation.

Bound the loss before estimating the upside

Early product bets tempt teams to model an attractive future and discount it with a confidence percentage. The calculation looks disciplined. Its inputs are often invented.

Jason Cohen's Lost Confidence identifies a practical distortion in confidence-weighted prioritisation. Familiar, incremental work feels predictable, while differentiated bets contain genuine uncertainty. Multiplying impact by subjective confidence can therefore favour the work the company already understands, even when that work has little strategic upside.

When probabilities cannot be calibrated, shape exposure instead:

  • What is the maximum loss the company can absorb without damaging its core obligations?
  • What is the smallest complete test capable of reaching the disputed behaviour?
  • Which assets, relationships, or capabilities remain valuable if the thesis fails?
  • What opportunity is being displaced while the test runs?
  • What evidence is required before the next tranche of commitment?

Saras Sarasvathy’s effectuation theory contrasts expected-return logic with decisions based on affordable loss, available means, partnerships, and adaptation to contingencies. The theory was developed around entrepreneurship and should not be treated as a universal product operating model. Its affordable-loss principle is still useful when a market is too uncertain to support precise probabilities.

Offensive and defensive work diverge at this point. Reforge's Product Strategy Is Really About Offense vs. Defense treats them as different investment logics. An offensive bet seeks enough upside to change the trajectory and may require concentrated resources. Defensive work seeks to contain a material downside such as reliability, capacity, compliance, or competitive exposure.

The stopping rules differ. An offensive bet stops or changes when its mechanism fails to produce the expected behaviour. Defensive work stops when risk reaches an agreed threshold. Without that threshold, reliability, polish, parity, and internal efficiency can expand indefinitely.

A loss limit makes an uncertain bet fundable without pretending uncertainty has disappeared.

Write revision triggers before staffing the plan

Once a team has been hired, a launch announced, and a leader’s reputation attached, evidence no longer enters a neutral system.

Hal Arkes and Catherine Blumer’s sunk-cost experiments documented a greater tendency to continue after investing money, effort, or time. In their studies, prior investment also affected judgments about a project’s likelihood of success. The research does not prove that every product organisation will escalate every commitment. It establishes a credible bias that strategy processes should anticipate.

Precommitment is the defence. Before the investment, write three types of trigger:

Continue
Which result earns the next level of commitment?
Revise
Which result preserves the diagnosis but contradicts the proposed mechanism, segment, channel, or operating model?
Stop
Which result makes the expected upside too small, the downside too large, or the causal claim no longer credible?

Place each trigger near the mechanism. If the strategy assumes that security review blocks enterprise deployment, measure completed deployments after removing that constraint. Pricing-page visits are too distant. If the strategy assumes that collaboration creates distribution, measure invitations that produce active collaborators. Button clicks are too early.

Mike Fisher's Red-Teaming Your Strategy makes a related organisational point: challenge should be designed into the process rather than depend on a brave dissenter appearing at the right moment. The intensity should match the bet. A reversible interface decision does not need a formal review. A large, ambiguous, hard-to-reverse commitment does.

The trigger also needs an owner with authority. A threshold that nobody is obliged to act on is a warning, not a rule.

A strategy is not testable until contrary evidence can change the allocation.

Keep the choice stable and the mechanism adaptable

Falsifiability can be misused. A team may react to every weak signal by changing direction, confusing responsiveness with learning. Strategy then becomes a sequence of explanations for the latest metric.

Henry Mintzberg and James Waters offer the necessary counterweight in Of Strategies, Deliberate and Emergent. Their research places deliberate and emergent strategy at opposite ends of a continuum. Realised strategy can contain intended choices and patterns that become visible through action.

This means a product strategy needs two kinds of stability:

  • Choice stability: enough continuity in the target customer, problem, and advantage for the organisation to build coherent capabilities and for evidence to accumulate.
  • Mechanism flexibility: permission to change the feature, workflow, channel, sequence, or operating model when evidence shows that the current route does not work.

Do not protect a tactic by calling it strategy. Do not abandon a sound diagnosis because the first implementation failed.

The distinction is easiest to manage with explicit levels of revision:

  1. Tactic revision: change the implementation while preserving the mechanism.
  2. Mechanism revision: change how the product is expected to produce the behaviour while preserving the customer problem and strategic choice.
  3. Strategy revision: change the customer, problem, advantage, or major allocation.

This prevents two opposite errors. One is dogmatism, where every failure is blamed on execution. The other is strategic thrashing, where every disappointing week produces a new thesis.

Write one page that governs the roadmap

Before making slides, write a strategy record that a sceptical executive and a delivery team can inspect in one sitting.

Change
What changed in the customer, market, technology, channel, or company? Separate observation from interpretation.
Choice
Where will the company concentrate, and what will receive less attention?
Advantage
Which capability, asset, relationship, or constraint makes this company suited to the choice?
Causal chain
Which product change should alter which behaviour, through which route to adoption, and why should that matter economically?
Critical assumption
Which uncertain claim can invalidate the largest amount of planned commitment?
Evidence sequence
What problem, mechanism, repeatability, and economic evidence should arrive, in what order?
Commitment schedule
What is committed now, what remains optional, and what result earns the next tranche?
Revision triggers
Which evidence causes the team to continue, revise, or stop? Who decides, and when?

The roadmap follows this record. Its job is to order the work that buys evidence and builds the chosen position. If a roadmap item cannot be connected to the causal chain, the risk being contained, or the evidence being purchased, it does not belong in the strategic plan.

The separate guide on roadmapping with decisions rather than dates shows how to carry that logic into Explore, Choose, and Commit states without hiding genuine delivery constraints.

The one-page constraint also reveals false agreement. People can applaud an ambitious deck while holding incompatible beliefs about the customer, mechanism, acceptable loss, and time horizon. Precise fields make those disagreements discussable before delivery turns them into cost.

Our view: make evidence and commitment advance together

The usual advice says to make strategy clear, focused, and measurable. All three are necessary. They are insufficient because they leave allocation outside the logic of learning.

Our judgment is stricter:

A product strategy is credible only when every material increase in commitment must be earned by evidence about the causal mechanism.

This does not mean running an experiment for every decision. Mature products often have operating knowledge strong enough to support direct action. Some bets cannot be decomposed without destroying the value being tested. Safety, legal, and platform obligations can require investment before market evidence. In those cases, the strategy should state why staged evidence is unavailable and bound the exposure by other means.

The standard applies most strongly when four conditions coincide: the upside is strategic, the mechanism is uncertain, the commitment is large, and reversal is costly. That is where a confident narrative is most seductive and least sufficient.

What would change our view? Evidence that explicit staged commitments systematically cause teams to underinvest in coherent, long-horizon positions would weaken it. So would evidence that precommitted revision triggers produce metric gaming or premature abandonment more often than they prevent escalation. The correct response would be to narrow the method to particular uncertainty profiles, not to defend it as doctrine.

For now, the practical sequence is clear:

  1. Make a real choice.
  2. Explain the causal chain.
  3. Find the assumption that can break the decision.
  4. Buy evidence in the order that reduces irreversible risk.
  5. Let each level of evidence earn the next level of commitment.
  6. Decide in advance what causes continuation, revision, or exit.

The goal is not a strategy that predicts the future. It is a strategy that can learn before the company loses the option to change.

For adjacent cases and arguments, continue with the Product Strategy reading list.

Sources and further reading