Kernel of Strategy
Rumelt's kernel says a real strategy has exactly three parts - a diagnosis of the actual problem, a guiding policy for tackling it, and coherent actions that carry it out - and most 'strategies' fail because they skip the diagnosis.
Follow the arrow through three linked boxes: diagnosis, guiding policy, then the actions that carry it out.
Reach for this when…
- Your 'strategy' is a list of goals with no explanation of why they'll work.
- Different departments are pulling in different directions under the same strategy document.
- You suspect the plan is really just last year's plan with new numbers.
How to run it
- Diagnose the situation: name the actual challenge, not the symptom.
- Simplify the diagnosis until it points at what matters most.
- Set a guiding policy: the broad approach for overcoming that specific challenge.
- Design coherent actions: steps that reinforce each other and carry out the policy.
- Check the three fit together: a diagnosis without action is just analysis, action without diagnosis is just busywork.
A worked example
Situation. Rima Akter runs Padma Horticulture, a horticulture exporter in Dhaka, Bangladesh, spread across five export markets and permanently short of cold-chain capacity.
Applied. Diagnosis: the real constraint wasn't demand, it was refrigerated freight capacity. Guiding policy: concentrate on the one market that could fill a container. Coherent action: cut four smaller markets, renegotiate the freight contract, and invest the savings in cold storage.
Result. Within two seasons, the single market she'd concentrated on was outperforming what all five scattered markets combined had delivered, because the cold chain finally held and produce stopped arriving spoiled.
The catch
The kernel exposes bad strategy but doesn't write good strategy for you - the diagnosis step still depends on judgement and evidence, and a wrong diagnosis produces a coherent, well-executed, wrong plan. It also says nothing about how to build buy-in for the guiding policy once you have it.
If you can write the guiding policy before you've done the diagnosis, you haven't diagnosed anything - you've just stated a preference.
Origin: Richard Rumelt