A real strategy has three parts: a diagnosis of the specific obstacle you're up against, a guiding policy for overcoming it, and a set of coherent actions that carry the policy out. If your document only lists goals, KPIs, and aspirations without naming the crux you're solving for, you're holding a wish list, not a strategy.

Quick answer: Rumelt's kernel test asks whether a document names a specific diagnosis, a guiding policy that follows from it, and coherent actions that follow from the policy. No diagnosis usually means no strategy — just a budget with ambitions attached.

Most strategy documents fail this test not because their authors are careless, but because goal-setting is easier and more comfortable than diagnosis. Naming a target ("become the category leader," "grow ARR 40%") feels productive. Naming the actual obstacle in your way — the thing that's been true and unaddressed for two years — feels like admitting a problem. That discomfort is exactly why so few documents clear the bar.

What Richard Rumelt's Strategy Kernel Actually Is

The strategy kernel is a three-part test from strategist Richard Rumelt's 2011 book Good Strategy Bad Strategy: a real strategy contains a diagnosis, a guiding policy, and coherent action — and each part must logically follow from the one before it. Strip any one part out and what remains is a goal, a slogan, or a plan without a rationale.

Rumelt, a longtime strategy professor at UCLA Anderson, built the kernel after years of consulting engagements where he noticed the same pattern: executives could produce beautifully formatted strategy decks that, on close reading, contained no actual strategy — just restated ambitions. The kernel is his diagnostic tool for telling the two apart.

  1. Diagnosis — a clear-eyed read of the situation that identifies the single most important obstacle standing between you and your goal, simplifying a messy, complex situation down to its critical features.
  2. Guiding policy — the overall approach chosen to deal with the obstacle identified in the diagnosis. It rules out a wide range of other approaches, which is what makes it a policy rather than a wish.
  3. Coherent action — the specific, coordinated steps, resource commitments, and policies that carry out the guiding policy. Actions must reinforce each other, not sit as an unrelated checklist.

The kernel is deliberately sequential. You cannot write a credible guiding policy without a diagnosis to respond to, and you cannot claim coherent action without a policy for those actions to cohere around. Most bloated "strategy" documents skip straight to step three, or worse, list a set of goals and call the list itself a strategy.

Why the sequence matters more than the content

A guiding policy borrowed wholesale from a competitor or a trend report ("we will become AI-first") isn't wrong because AI is a bad direction — it's wrong because it wasn't derived from your diagnosis. The same policy could be correct for one company and irrelevant for another; the diagnosis is what makes it yours.

This is the fastest way to pressure-test any strategy document already sitting in your drive: find the sentence that names the obstacle. If you can't find one, everything downstream — goals, initiatives, roadmaps — is arbitrary, however confidently it's written.

The Diagnosis: Naming the Crux Before You Name a Goal

A diagnosis answers one question honestly: what is actually stopping us, stated specifically enough that it rules out most possible responses. It is not a SWOT list, not a market-sizing slide, and not a restatement of the goal in different words — it's a claim about cause, made in plain language a skeptic could argue with.

Rumelt has described reviewing hundreds of corporate and public-sector strategy documents over his career and finding the same gap again and again: pages of objectives, almost nothing that names why the organization hasn't already achieved them. That gap is the single most common failure mode in strategy writing, and it's usually invisible to the people who wrote the document, because a goal feels like an explanation when you've been staring at it long enough.

A usable diagnosis has three properties:

  • It's specific to your situation, not a generic industry statement ("the market is competitive") that would be equally true of any company in your space.
  • It names a mechanism, not just a symptom — "churn is rising" is a symptom; "customers hit a capability ceiling at month four and switch before renewal" is closer to a mechanism.
  • It's falsifiable — someone with the same data could look at it and say "no, I think the real problem is X" and you could argue the point with evidence.

Diagnosis work is where structured discovery tools earn their keep. A Jobs to Be Done interview protocol can surface the mechanism behind a symptom — why customers actually hire and fire your product — better than a survey ever will; our Jobs to Be Done complete guide walks through running that discovery end to end. Mapping the emotional highs and lows across a customer's lived experience, as in a customer journey exercise, often exposes exactly where the mechanism breaks down.

For diagnosing competitive and market position rather than customer behavior, Wardley Mapping gives you situational awareness — where each component of your value chain sits on an evolution axis from genesis to commodity — which is often the difference between diagnosing "we're behind" and diagnosing precisely why and where. Our guide to seeing the whole competitive board with Wardley Mapping covers the technique in depth.

A diagnosis is not the same thing as a problem statement in a PRD. A PRD problem statement justifies one feature. A strategic diagnosis explains why an entire category of effort — the last four quarters of roadmap, say — hasn't moved the number you care about.

Guiding Policy: The Approach You Commit To by Ruling Out the Rest

A guiding policy is the overall approach for tackling the diagnosed obstacle — and its defining feature is what it rules out, not what it includes. If your "policy" is compatible with almost any initiative your team might propose, it isn't a policy; it's a mood.

This is where Rumelt's kernel rhymes with a second well-known framework: A.G. Lafley and Roger Martin's Playing to Win (2013), built from Lafley's years running Procter & Gamble. Their strategy choice cascade forces two decisions before anything else — where to play and how to win — and both are explicitly framed as choices that exclude alternatives. "How to win everywhere, for everyone" is treated in that book as a description of having no strategy at all, the same failure Rumelt calls mistaking a goal for a policy.

Michael Porter made the same point earlier and more bluntly in his 1996 Harvard Business Review essay "What Is Strategy?" — still one of the most widely assigned readings in MBA strategy courses — arguing that strategy is fundamentally about trade-offs: choosing what not to do is as much a part of strategy as choosing what to do. A guiding policy without an explicit trade-off attached to it hasn't earned the name.

A guiding policy should be testable against a short checklist:

QuestionIf the answer is "no"What that signals
Does it respond directly to the diagnosis?It's disconnected from the stated obstacleYou've written an aspiration, not a policy
Does it rule out real alternatives?Almost any roadmap item is still "compatible" with itIt's a mood, not a choice
Could a competitor with different resources copy it verbatim?Yes, easilyIt isn't specific to your situation or capabilities
Can one sentence state it without buzzwords?It needs a paragraph of qualifiersIt's fluff dressed as strategy

Good guiding policies often sound almost narrow when you first write them down — "win the mid-market segment currently underserved by enterprise tools, by trading feature breadth for a two-week implementation" is a policy precisely because it makes some other appealing options (chase enterprise, match feature-for-feature) explicitly off the table. Our broader product strategy guide covers how a guiding policy of this kind should shape everything downstream, from roadmap sequencing to what you say no to in a sales conversation.

Coherent Action: The Part That Kills Most Strategy Documents in Practice

Coherent action is the set of concrete, coordinated steps — resource commitments, org design, product bets, sales motion changes — that actually carry out the guiding policy, and the test is coordination, not quantity. A long list of "strategic initiatives" that each team pursues independently, without reinforcing each other, is not coherent action even if every item on the list is individually reasonable.

This is usually where a genuinely good diagnosis and policy still collapse in execution. A guiding policy of "win on implementation speed" implies specific, mutually reinforcing actions: a leaner onboarding flow, a sales process that doesn't over-promise customization, an engineering roadmap that protects simplicity over feature parity. If your actual roadmap simultaneously funds a heavyweight customization framework because a sales rep asked for it, the actions aren't coherent — they're fighting the policy.

Signs coherent action has broken down:

  1. Two teams are resourced to pursue approaches that would undercut each other if both succeeded.
  2. The roadmap reads as a merged wish list from every function rather than a set of moves that reinforce one bet.
  3. No one can say which current initiative would be cut if the guiding policy were taken seriously.
  4. Quarterly planning starts from "what do we have capacity for" instead of "what does the policy require."

This is precisely the gap our piece on the strategy-deck-to-daily-execution gap covers — the mechanism by which a coherent-sounding slide turns into an incoherent quarter, and what closes it. A kernel that's airtight on paper still needs a translation layer into sprint-level decisions, or the diagnosis and policy work was wasted.

The Hallmarks of Bad Strategy: How to Spot One in Your Own Document

Rumelt names a small set of recurring failure patterns that are worth checking your own strategy doc against directly, because they hide in confident, well-formatted language rather than obviously sloppy writing. Bad strategy is rarely stupid — it's usually fluent, which is exactly why it survives review.

HallmarkWhat it looks like on the pageFast test
FluffBuzzword-dense restatement of the obvious dressed up as insight ("leverage synergies to drive customer-centric innovation")Remove every abstract noun and adjective — is there a concrete claim left?
Failure to face the challengeConfident language about the destination, silence about what's blocking the path thereAsk "why haven't we already done this?" — if the doc can't answer, it hasn't faced the challenge
Mistaking goals for strategy"Strategy" = a target plus a deadline ("$50M ARR by FY27") with no mechanism attachedAsk "how," not "what" — a real strategy answers "how," a goal only answers "what" and "by when"
A long list of disconnected objectivesTen to fifteen "strategic priorities," each defensible alone, none clearly ranked or connectedAsk which three would be cut under resource pressure — if there's no answer, there's no priority

The most common of these in product organizations is mistaking goals for strategy, because product teams are trained — correctly, for execution purposes — to set measurable targets. OKRs, RICE scores, and roadmap KPIs are excellent tools for coordinating execution once a guiding policy exists. Used as a substitute for one, they produce a document that looks rigorous and contains no actual strategic reasoning.

Fluff deserves special attention because it's the hardest hallmark to self-detect: vague, elevated language reads as sophistication to the person who wrote it and as evasion to everyone else. A vision statement is a legitimate, different artifact from a strategy — it should be aspirational and durable.

But when fluff creeps into the strategy section, it's usually covering for a diagnosis the team hasn't done. Our guide to writing a product vision people actually repeat draws that line clearly: a vision earns vagueness because it's meant to outlast any one obstacle; a strategy kernel doesn't get that pass.

A Rewrite Example: Turning a Wish List Into a Kernel

The fastest way to internalize the kernel test is to watch a typical wish-list "strategy" get rewritten against it, obstacle first. Below is a composite, illustrative example built from the pattern that shows up constantly in real strategy decks — not a specific company's actual document.

Before (wish list, not a strategy):

Our strategy is to become the leading platform for mid-market operations teams by delivering best-in-class innovation, deepening customer relationships, and driving sustainable growth. Priorities: launch AI features, expand into two new verticals, improve NPS, grow ARR 35%, reduce churn.

Every clause is a goal or an ambition. Nothing in it explains why the company hasn't already captured this position, and nothing rules out any initiative a team might propose — it would be equally compatible with almost any roadmap.

After (kernel format):

  • Diagnosis: Mid-market operations teams evaluate us against enterprise incumbents on feature count, but our actual win rate comes from implementations that finish in under three weeks versus the incumbents' three-to-six-month rollouts. We've been diluting that edge by matching incumbent feature requests deal-by-deal, which slows our own implementation time and erases the only advantage that reliably wins deals.
  • Guiding policy: Compete on implementation speed as the primary axis, not feature parity. Explicitly decline enterprise-style customization requests that would extend time-to-value past two weeks, even when a specific deal would prefer them.
  • Coherent action: Cap the onboarding configuration surface engineering will support; retrain sales on a two-week-implementation pitch instead of a feature-comparison pitch; kill the two roadmap items that exist solely to match incumbent feature lists; set the sales incentive structure to reward time-to-first-value, not just contract size.

Notice what changed isn't the ambition level — both versions want growth — it's that the second version can be argued with. Someone could reasonably say "I think the real obstacle is pricing, not implementation time," and that disagreement would be productive, because there's now a specific claim on the table instead of a mood.

Where Prodinja Fits: Making the Diagnosis Concrete

The hardest part of this whole exercise is rarely writing the guiding policy — it's doing diagnosis work rigorous enough to earn one. Prodinja's Systems Engineering canvas is built for exactly that step: it lets you map the core constraint in your system and the causal loops feeding it, what's reinforcing the obstacle and what's balancing against your fixes, before you commit language to a guiding policy.

Treating diagnosis as a canvas you build rather than a paragraph you write is a useful forcing function for the same discipline Rumelt is describing: you can't map a loop you haven't actually identified.

Key Takeaways

  • A strategy needs all three kernel parts — diagnosis, guiding policy, coherent action — and each must logically follow from the one before it; a goal list alone is not a strategy, however detailed.
  • Diagnosis is the part almost everyone skips. It names the specific mechanism blocking progress, not a generic market observation or a restated goal.
  • A guiding policy is defined by what it excludes. If it's compatible with nearly any roadmap item your team might propose, it's a mood, not a policy.
  • Coherent action means coordination, not a checklist length. Initiatives that don't reinforce each other — or that quietly work against the guiding policy — aren't coherent even if each is individually sound.
  • Bad strategy is fluent, not sloppy. Fluff, avoiding the real challenge, and mistaking goals for strategy all read as confident, well-formatted writing — that's exactly why they survive review.
  • The fastest self-check: find the sentence in your document that names the obstacle. If there isn't one, everything after it is arbitrary.
  • Diagnosis tools compound. Jobs to Be Done, customer journey mapping, and Wardley Mapping all exist to make the diagnosis step evidence-based instead of a guess dressed as insight.

Frequently Asked Questions

What is Rumelt's strategy kernel?

The strategy kernel is Richard Rumelt's three-part test for a real strategy from his book Good Strategy Bad Strategy: a diagnosis of the core obstacle, a guiding policy for addressing it, and coherent action that carries the policy out. All three must be present and logically connected — missing any one means the document is a goal list, not a strategy.

What's the difference between a goal and a strategy?

A goal states a destination and a deadline ("$50M ARR by next year"); a strategy explains the mechanism for getting there despite a specific, named obstacle. Rumelt calls confusing the two — "mistaking goals for strategy" — one of the most common hallmarks of bad strategy, because measurable targets feel rigorous even with no diagnosis behind them.

What are the hallmarks of bad strategy?

Rumelt identifies several recurring patterns: fluff (buzzword-heavy language with no concrete claim), failure to face the challenge (naming a destination without naming the obstacle), mistaking goals for strategy, and long, unranked lists of "strategic objectives" that don't logically connect to any diagnosis. All four tend to read as confident and well-produced rather than obviously weak.

How do you write a good strategic diagnosis?

Write a diagnosis by naming the single most important, specific mechanism blocking your goal — not a symptom, not a generic market condition — stated concretely enough that someone could disagree with it using evidence. Tools like Jobs to Be Done interviews, customer journey mapping, and Wardley Mapping help surface that mechanism instead of guessing at it from inside the building.

Is a vision statement the same as a strategy?

No. A vision statement is meant to be aspirational and durable, describing a future worth working toward; a strategy kernel is meant to be specific and time-bound, diagnosing a current obstacle and committing to one approach for beating it. Vagueness that's appropriate in a vision statement becomes fluff — one of Rumelt's hallmarks of bad strategy — the moment it shows up in the strategy itself.