A product strategy fails in one of three places, and naming which one is almost the whole of the work. The strategy has by this point been chosen, tested, funded, sequenced, measured, communicated inside the organisation and positioned outside it. Sooner or later the results arrive and disappoint somebody. The person who wrote the strategy is the worst placed person in the building to judge it, and also the only one with enough of the reasoning in their head to judge it correctly. That person is usually the product manager.
The sections below start with the question that has to be settled before any other. Reading the indicators for a verdict comes next. The rest of the page locates a failure in the part of the kernel it belongs to. It then separates a deliberate change of direction from thrashing and ends with the review rhythm.
The question that comes before every other one
Before anything else, somebody has to ask whether enough time has passed for the strategy to be right. Only after that does it make sense to ask whether the strategy is wrong.
This is the most commonly skipped step and the most expensive. A strategy aimed at enterprise retention sets a lagging measure that takes four quarters to read. At month five the number has not moved, and somebody senior asks whether the approach is working. A strategy that was always going to take a year then gets replaced by one that will also be abandoned at month five.
The defence was built earlier, when the indicators were chosen. A leading indicator that was supposed to move by now and has not is evidence. A lagging indicator that could not possibly have read yet is no evidence at all, and saying so plainly in the review is a large part of what the role is for.
One more distinction is worth stating. An unpopular strategy is not a failing strategy. Non goals make people unhappy by design, so discomfort from the teams whose work was declined is the strategy operating correctly.
Reading the indicators for a verdict
If enough time has genuinely passed, three readings are possible, and each of the three means something different.
The leading indicators moved and the lagging ones followed. The strategy is working, and the remaining question is whether to press harder on it.
The leading indicators moved and the lagging ones did not. This is the most informative failure available. The team successfully changed the thing it believed would drive the outcome. The outcome did not follow, so the belief connecting the two was wrong. The diagnosis or the guiding policy has been falsified, and cheaply.
Neither moved. The least informative result, because it cannot tell a wrong strategy from one that was never really carried out. Settling it means looking at what the organisation actually did with its time.
Locating the failure in the kernel
Rumelt's kernel from the opening module gives the three places a strategy can break. Each one has a different response, and treating them as one failure is what produces a second strategy that fails the same way as the first.
| What the evidence shows | Where it failed | The response |
|---|---|---|
| Buyers do not have the problem the strategy was built around | Diagnosis | Return to the market. Nothing built on top of it can be salvaged |
| The problem is real and the chosen way of winning does not move it | Guiding policy | Keep the diagnosis, change the policy. This is what a pivot means |
| The policy was sound and the organisation never did the work | Coherent action | Keep both and fix allocation, sequencing and focus |
| Everything holds and the slow measures have not read yet | Nothing has failed | Say so, and hold the direction |
The third row is the common case and the most often misread. A strategy can choose parts data over scheduling and then spend three quarters shipping scheduling improvements, because the largest customer kept asking. That strategy has not been tested. It has been abandoned quietly, and it is about to be replaced by a new strategy that will meet exactly the same pressure.
The question that separates those two rows is uncomfortable and specific. It is what share of engineering time actually went to work that served the guiding policy. If the honest answer is under half, the strategy was never on trial.
Changing direction once, deliberately
If the evidence genuinely points at the diagnosis or the policy, the change should be a single deliberate move with a decision behind it.
Eric Ries gave the useful definition in The Lean Startup, published in 2011. A pivot is a structured change of one element of the strategy, and it keeps everything learned so far. It is a considered move and never a reaction. The strategy keeps its diagnosis and changes its policy, or keeps its segment and changes how it wins, and it holds on to the evidence that produced the original choice.
Thrashing looks different and is easy to recognise from outside. Direction changes arrive every quarter, each one triggered by the most recent piece of bad news, and none of them runs long enough to produce evidence. The organisation is busy, nothing compounds and the teams stop believing any stated direction will survive to the next review.
Two disciplines keep a change on the right side of that line. The change is written down as a change, naming what is being kept and what is being dropped and why, so that the reasoning survives. And the money already spent stays out of the decision entirely. It is sunk and cannot be recovered, and the only live question is what the remaining investment buys compared with the alternatives.
The review rhythm
A rhythm puts the decision on a schedule, so it arrives on a date the organisation set in advance. Nobody has to raise the alarm to get it looked at.
A workable version has three levels. Leading indicators get read monthly. If an indicator that should have moved has not, that is a prompt to ask why, and never on its own a reason to change course. The strategy's assumptions get revisited quarterly, alongside the question of whether any of them has been settled or disproved since the last look. And once a year the whole kernel is reopened deliberately, with the diagnosis re examined against what the market now looks like.
The annual review is the easiest of the three to skip, and it is the one that matters most, because a strategy rarely announces that it has expired. The condition it was built to address gets resolved, or a competitor's move makes it irrelevant. The strategy carries on regardless, because nobody scheduled the moment to ask.
A strategy is a diagnosis of a situation, a guiding policy for dealing with it and a set of coherent actions that carry the policy out. It succeeds or fails at one of those three, and every review that produces a useful answer is a review that said which. The course began with that structure because it is the same instrument at both ends. It builds a strategy, and then it judges the one that was built.