Strategy as Logic Not Analysis
Because business lives in the "can be otherwise" world, strategy cannot be derived from data. It must be argued.
The strategist's job is not to process information correctly but to construct the most compelling argument for a possibility that does not yet exist, then identify what would have to be true for it to work, and test the pivotal conditions. That sentence reverses the usual order of corporate planning, which gathers data first and hopes a strategy precipitates out of it. It never does. Data cannot contain a possibility that has not happened yet.
What data is still for
Data matters as evidence about conditions, never as proof of conclusions. "The data shows our strategy is right" is a category error: the data can only show what was, and in the can-be-otherwise domain the past is not a representative sample of the future. The honest formulation is conditional: if these conditions hold, this strategy wins, and here is the current evidence on each condition. That is What Would Have to Be True, and it is the entire method in one sentence.
Why the tools are logical, not analytical
This is why Martin's instruments are logical structures rather than analytical models. The Strategy Choice Cascade forces a coherent argument: five answers that must hold together. The Opposite Test checks an argument contains a real premise. Possibility portraits generate arguments worth having in the first place. None of them computes an answer, because there is no answer to compute; there is a best-reasoned bet, which is different, and pretending otherwise is how analysis becomes a hiding place. Rumelt's crux completes the toolkit: every real strategy has one pivotal challenge that makes it hard, and naming it is what creates shared understanding.
How this connects to OKRs
Our goal system runs on the same epistemology one level down. An OKR is an argued bet, not a derived target: the narrative carries the argument, the Key Results name the evidence that would confirm it, and confidence grades move as the conditions report in. Teams that want the numbers to prove the goal before committing to it are asking analysis to do logic's job, and the quarter usually punishes the confusion.
One line to keep: An OKR is an argued bet, not a derived target; asking analysis to prove it first is asking analysis to do logic's job.
Our synthesis of Roger Martin’s published work, sources credited. Read the originals: they’re excellent.
What is the difference between logic and analysis in strategy? +
Analysis computes an answer from existing data; logic reasons to the most plausible bet in a future that does not yet exist. Strategy lives in the second, so its tools build arguments, not conclusions. Demanding proof first is how analysis becomes a hiding place.
How does this show up in OKRs? +
A good OKR carries an argument in its narrative and names the evidence that would confirm it in its Key Results, with confidence grades moving as conditions report in. Teams that want the numbers to prove the goal before committing are asking analysis to do logic's job.
Reading about method is not the same as running it. We install this system and build the capability that stays.
Turn strategy into executed strategy →