Strategy & competition
Rumelt's Kernel of Good Strategy
Rumelt's test of whether a strategy exists at all: a diagnosis that names the critical challenge, a guiding policy for dealing with it, and a set of coherent actions that carry the policy out. A document missing any of the three is goals, vision or fluff, whatever its cover says.
Also known as The Kernel, Good Strategy / Bad Strategy, Diagnosis, Guiding Policy, Coherent Action. First set out by Richard P. Rumelt in 2011; the primary source is cited in full below.
- Format
- Structural model
- Level
- Corporate · Business unit
- Best for
- Plan execution
- Decision stage
- Diagnose · Decide · Plan
- Difficulty
- Intermediate
- Time to apply
- A focused day to draft a kernel; the diagnosis behind it can take weeks of evidence-gathering to earn.
Plate · The model
The components
Diagnosis
A judgement that defines or explains the nature of the challenge, simplifying the situation to its critical aspects and often reframing it. It determines what the strategy must overcome.
Signals of strength
Names the critical challenge in a sentence or two · Simplifies without falsifying: an informed insider would recognise the situation · Identifies obstacles, including internal ones
Guiding policy
The overall approach chosen to cope with the diagnosed challenge. It directs and constrains action like a guardrail, and draws on or creates advantage.
Signals of strength
Rules some plausible responses out, not just in · Creates or exploits an identifiable advantage · Its opposite is a position a sane rival might actually take
Coherent action
The set of feasible, coordinated actions, resource commitments and policy changes that carry out the guiding policy. Coordination is itself a source of advantage.
Signals of strength
Actions reinforce one another rather than compete · Named resources are committed to each action · Nothing on the list undercuts anything else on it
When it earns its keep
- The strategy documents in front of you are full of ambitions, values and targets, and you suspect there is no strategy in them. The kernel is the sharpest available test.
- The organisation faces a genuine crux, a challenge whose nature is unclear, and effort is scattered across initiatives that do not add up.
- You have run diagnostic frameworks, Five Forces or scenario work for instance, and need to convert the findings into a policy and a coordinated set of actions.
- A proposed plan needs a coherence check: whether its actions reinforce one another or quietly conflict, and whether real resources sit behind them.
And when it doesn't
- The situation is stable and well understood, and the task is disciplined execution of a known playbook. Forcing a fresh diagnosis onto a solved problem manufactures drama.
- You need a full architecture of market choices across segments, channels and geographies. The cascade in Playing to Win does that work; the kernel deliberately does not.
- It is being used as a template. Rumelt is scathing about template-style strategy, and three headings filled with fluff are still fluff.
- The exercise is really consensus goal-setting. A kernel requires leadership willing to name one challenge and exclude responses, which a harmony-first process will not produce.
How to run it
Before starting, gather the inputs the analysis depends on:
- Evidence about the situation: performance data, competitor and market movement, and the structural changes bearing on the business.
- A candid account of obstacles, including internal ones such as inertia, politics and sacred-cow activities, since these are often the real challenge.
- A clear picture of the resources, capabilities and advantages that effort could be concentrated behind.
- Leadership prepared to choose, which means prepared to disappoint the owners of everything the strategy will not do.
- 1
Diagnose the challenge
Work out what is going on before deciding what to do. A good diagnosis simplifies the overwhelming complexity of the situation to its critical aspects and names the obstacle to progress, often reframing the problem entirely. Rumelt treats this as most of the work: without diagnosis there is no strategy, only wish lists and hope.
- 2
Formulate the guiding policy
Set the overall approach for dealing with the diagnosed challenge. Like guardrails on a road, the guiding policy directs and constrains action without prescribing every step, and it should create or exploit some source of advantage. If its opposite would be obviously absurd ('serve customers badly'), it is not a policy, it is fluff.
- 3
Design coherent actions
Specify the coordinated steps, resource commitments and policy changes that carry out the guiding policy. Coherence is the test: the actions must be consistent with the policy and with each other, and each must have real resources attached. A list of departmental initiatives stapled together fails this step by definition.
- 4
Screen for the hallmarks of bad strategy
Check the result against Rumelt's four hallmarks: fluff (inflated language dressed as insight), failure to face the challenge, mistaking goals for strategy (ambition and targets presented as if they were a plan of action), and bad objectives, either blue-sky wishes or an impracticable dog's dinner of demands. Most strategy documents fail at least one.
- 5
Concentrate and sequence
Apply focus and leverage: pick proximate objectives, targets close enough to be feasibly hit, and concentrate resources on them rather than spreading effort across everything the diagnosis touched. Strategy is as much about what the organisation will now stop doing as what it will start.
Reading the result
A short statement of strategy in three parts: the diagnosis of the critical challenge, the guiding policy chosen to deal with it, and the coherent actions, with resources attached, that will carry the policy out, plus an honest verdict on whatever document the kernel replaced.
- Read it backwards: each action should trace to the policy, and the policy to the diagnosis. A break in that chain means decoration has crept in.
- Test the guiding policy by negation. If no serious competitor would ever adopt the opposite, the policy says nothing.
- Check what has been excluded. A kernel that stops nothing has concentrated nothing.
- Hold the diagnosis lightly: it is a judgement about a moving situation, and the strategy is only as good as the diagnosis is current.
A worked example
A timber merchant chain replaces its growth goals with a strategy
A 14-branch timber and builders' merchant chain across the north of England has a strategy that reads: grow revenue 25% in three years, be the customer's first choice, invest in our people. Like-for-like sales are falling as national chains and online specialists take trade accounts. A new managing director suspects the document is goals in a strategy costume and rebuilds it with the kernel.
- Diagnosis
- Interviews with 40 lapsed trade accounts point away from price and marketing: customers defect after stock-outs on core structural timber and sheet materials. Availability is inconsistent because each branch sets its own range and buying is fragmented across 14 branch managers. The crux is reliability, and it is an internal, self-inflicted obstacle, which is why previous marketing spend changed nothing.
- Guiding policy
- Win and keep the repeat trade professional within 25 minutes of a branch by being the most reliable stockholder of core materials in the region, and stop competing for one-off retail custom and price-led contract tenders. The negation test passes: a sane rival could choose breadth of range or lowest price instead.
- Coherent action
- Cut the range by 30% to a guaranteed core list; convert two mid-sized branches into stocking hubs with twice-daily transfers; promise next-morning delivery on the core list; centralise buying of core lines; launch a trade account app showing live branch stock; end Sunday retail opening. The actions fund each other: the range cut releases the working capital that fills the hubs, and the retail exit frees counter staff for trade service.
- Screen for the hallmarks of bad strategy
- The old document fails all four: 'first choice for customers' is fluff, the 25% target is a goal mistaken for strategy, nothing in it faces the availability problem, and its objectives are a dog's dinner of departmental wishes. The new kernel names a challenge, excludes options and commits resources, though the screen flags one risk: the diagnosis rests heavily on lapsed-account interviews.
The read. The kernel concentrates the whole business on availability, which is also its exposure: if defections are really price-led, deeper range cuts will make things worse. So the managing director set a proximate objective, raise core-list fill rate from 91% to 98% within two quarters, and made account retention the tracked condition. Six months in, fill rate and retention move together, which is the diagnosis earning its keep rather than being taken on faith.
Pitfalls
- Writing the diagnosis to fit an action already favoured. The kernel is read top-down but is only honest if it was built top-down too.
- A guiding policy that excludes nothing, such as 'delight customers profitably'. If the opposite is absurd, the policy is fluff.
- Coherent action drafted as a wish list of departmental initiatives with no resources named and no test of whether the actions conflict.
- Mistaking ambition for strategy: revenue targets, vision statements and values pasted over a challenge nobody has faced.
- Treating the diagnosis as permanent. Challenges move, and a kernel built on last year's diagnosis coordinates effort against the wrong obstacle.
What the critics say
Martin argues that strategy is choice-making before it is problem-solving: the kernel starts from a challenge, offers little help generating and comparing where-to-play and how-to-win options, and can produce strategy that is reactive, framed by the problem in view rather than by an aspiration about winning.
Martin, R. L. (2024) 'Good Strategy/Bad Strategy & Playing to Win', Roger Martin on Medium.
The kernel presents one approach to strategy for all settings. Reeves, Haanaes and Sinha argue that strategy-making must match the environment, and that challenge-led, plan-then-act logic fits classical, predictable settings far better than adaptive or shaping environments, where continuous experimentation does the work a single diagnosis cannot.
Reeves, M., Haanaes, K. and Sinha, J. (2015) Your Strategy Needs a Strategy. Boston: Harvard Business Review Press.
The book argues from retrospective cases, Apple in 1997, Nvidia, Desert Storm, in which the good kernel is visible with hindsight. It is easier to recognise a diagnosis after events have vindicated it than to produce one in the fog, and Rumelt offers judgement and examples rather than a repeatable method for getting the diagnosis right.
Sources and further reading
- Rumelt, R. P. (2011) Good Strategy/Bad Strategy: The Difference and Why It Matters. New York: Crown Business.
- Rumelt, R. P. (2011) 'The perils of bad strategy', McKinsey Quarterly, June 2011. ↗
- Rumelt, R. P. (2022) The Crux: How Leaders Become Strategists. New York: PublicAffairs.
- Martin, R. L. (2024) 'Good Strategy/Bad Strategy & Playing to Win', Roger Martin on Medium. ↗