Good Strategy Bad Strategy

Good Strategy/Bad Strategy gives you Rumelt's kernel test - a real diagnosis, a guiding policy, coherent actions - to tell an actual strategy apart from a wish list dressed up as one.

Diagnosis leads, then the guiding policy it produces, then the coherent actions that carry it out, each step feeding the next.

Rumelt's kernel: three parts a real strategy needs, in order.

Reach for this when…

The "strategy" is a list of goals with no explanation of how they'll be reached.
Nobody in the room can say what specific challenge the plan is actually answering.
Initiatives compete with each other instead of reinforcing one direction.

How to run it

Write the diagnosis: what specifically is the challenge, stated plainly.
Write the guiding policy: the overall approach for dealing with it.
List the coherent actions that carry out the policy.
Check the actions actually reinforce each other rather than pulling apart.
Check against bad-strategy hallmarks: fluff, no real diagnosis, goals mistaken for strategy.
Rewrite the kernel until it survives the check.

A worked example

Situation. Matias Herrera runs a household goods retail chain in Valparaiso, Chile, and had a strategy document that was mostly a target list: grow revenue 20%, open five stores, improve NPS.

Applied. Matias ran the kernel test. The real diagnosis turned out to be that his stores were losing customers to faster online delivery, but only on the categories with thin in-store margins.

Result. The guiding policy became: concede fast-moving low-margin lines to online, own the categories customers need to see and touch before buying. Layouts, staffing and new openings all followed - the target list disappeared.

Youssef's missing piece was the diagnosis: customers weren't leaving broadly, they were leaving on thin-margin lines.

The catch

The kernel test is brutally good at exposing goal-lists dressed as strategy, but it doesn't write the diagnosis for you - a wrong diagnosis passes the test just as cleanly as a right one. It also assumes someone in the room is willing to name the real problem, often the politically hardest part.

If you can swap your diagnosis with a rival's and it still reads fine, it isn't a diagnosis.

Origin: Richard Rumelt

connecteddale