good-strategy-bad-strategy

Files (6)

• SKILL.md
• bad-strategy.md · references/
• case-studies.md · references/
• dynamics-inertia.md · references/
• kernel.md · references/
• sources-of-power.md · references/

SKILL.md (first lines)

# Good Strategy Bad Strategy

A framework for creating and auditing strategy, distilled from Richard Rumelt's *Good Strategy Bad Strategy: The Difference and Why It Matters*. Good strategy has a simple underlying logic — an honest diagnosis of the critical challenge, a guiding policy for overcoming it, and coherent actions that carry the policy out. Use this skill to detect the four hallmarks of bad strategy and to replace goal lists and vision decks with a working kernel.

## Core Principle

**Strategy is coherent action backed by an honest diagnosis — not goals, vision, or wishful thinking.** A goal ("20% growth") names an ambition; a strategy explains how the ambition will be achieved given the actual obstacles. Bad strategy is not the absence of strategy but an active substitute for it: buzzword fluff, refusal to name the challenge, and laundry lists of initiatives. The heart of strategy work is choice — concentrating effort and resources on the one or two pivotal objectives whose accomplishment unlocks everything else.

## Scoring

**Goal: 10/10.** Score strategies, plans, and strategy documents by walking the eight rows of the Quick Diagnostic and counting how many pass. Report the current score and the specific changes needed to reach 10/10. The bands below name what each tier looks like; the row count keeps the rating reproducible run to run.

- **9-10 (8 rows pass):** Complete kernel — honest diagnosis, choiceful guiding policy, coordinated resource-backed actions — aimed at a pivot point, with an explicit list of what will not be done
- **7-8 (6-7 pass):** Kernel present but one element weak: thin diagnosis, a policy that rules little out, or actions not yet coordinated and funded
- **5-6 (4-5 pass):** The challenge is named, but the plan is a list of independent initiatives and some goals masquerade as strategy
- **3-4 (2-3 pass):** Mostly goals, targets, and vision statements; no diagnosis; fluff in key passages; nothing ruled out
- **0-2 (0-1 pass):** Pure bad strategy — buzzword fluff, dog's-dinner objective lists, denial of the real challenge

## Framework

### 1. The Kernel of Good Strategy

**Core concept:** Every good strategy shares the same structure: a **diagnosis** that defines and simplifies the critical challenge, a **guiding policy** — the overall approach chosen to overcome the diagnosed obstacles — and **coherent actions**: coordinated, resource-backed steps that carry out the policy. A document missing any of the three is not yet a strategy.

**Why it works:** A diagnosis replaces the overwhelming complexity of reality with a simpler story that highlights what is critical, often by analogy to a known pattern. The guiding policy channels effort by ruling out vast realms of possible action — like guardrails, it directs without dictating every move. Coherent actions turn intent into coordinated force; most plans fail by jumping straight from ambition to a list of independent initiatives.

**Key insights:**
- The diagnosis is the strategy's pivot: Gerstner reframed IBM's 1993 challenge from "mainframes are dying, break the company up" to "our advantage is integrated capability; the obstacle is internal coordination" — and everything downstream changed
- A guiding policy is not a goal or a vision — it is an approach ("ride wave X by concentrating on Y"), and a real one feels like a choice with losers
- If a competitor could paste your guiding policy into their deck unchanged, it is a platitude, not a policy
- Coherent actions reinforce one another — each step makes the others easier — and every one carries an owner, resources, and a date
- A kernel needs no mission, vision, or values preamble; it fits on one page
- Most failed "strategies" skip the diagnosis entirely — prescribing before examining

**Applications:**

| Context | Application | Example |
|---------|-------------|---------|
| Annual planning | Kernel before targets | Diagnosis: week-one churn; policy: fastest time-to-value in segment; actions: onboarding rebuild + roadmap cuts |
| Strategy review | Trace each action to the policy | Initiative serving no policy → cut or re-justify |
| Pitch deck | Kernel slide, not goals slide | "The obstacle, our approach, three coordinated moves" |

**Ethical boundary:** An honest diagnosis names internal causes too — never soften it to protect egos or settle politics.

See [references/kernel.md](references/kernel.md) when you actually draft a kernel — diagnosis craft, guiding-policy formulation, coherent-action design, a fill-in template, and two worked examples with owners and done-tests.

### 2. Detecting Bad Strategy

**Core concept:** Bad strategy is not the absence of strategy — it is its own species with four hallmarks: **fluff** (gibberish masquerading as strategic concepts), **failure to face the challenge**, **mistaking goals for strategy**, and **bad strategic objectives** (dog's-dinner laundry lists or blue-sky impracticalities).

**Why it works:** Naming the hallmarks turns a vague sense that "this deck says nothing" into specific, fixable findings. Bad strategy persists for identifiable reasons — choice is painful, templates are easy, and positive thinking feels like leadership — so detection must hunt for substitutes for choice, not just bad writing.

**Key insights:**
- Fluff test: restate the sentence in plain words — "our fundamental strategy is customer-centric intermediation" collapses to "we are a bank," which says nothing
- If the document never names the obstacle, the strategy cannot be evaluated or improved — International Harvester's 1979 plan never mentioned its toxic labor relations, the actual problem
- "20% growth, 20% margin" is a goal; exhortation to push harder is motivation, not a lever — strategy is the lever
- Dog's dinner: a city plan with 47 "strategies" and 178 action items has no strategy; blue-sky: "become the leading platform" restates the end state and skips the how
- Bad strategy has causes: unwillingness to choose (every real choice creates losers — DEC's consensus produced mush), template-style vision-mission-values planning, and New Thought culture (belief that visualizing success produces it)
- The negation test: if the opposite of a statement is absurd ("we will *not* be customer focused"), the statement carries no information

**Applications:**

| Context | Application | Example |
|---------|-------------|---------|
| Strategy deck audit | Score sections against the four hallmarks | "Vision" slide flagged as fluff; no obstacle named anywhere |
| OKR review | Separate ambitions from mechanisms | "Double signups" kept as goal, paired with an explicit how |
| Board update | Demand the challenge slide | "What we're up against" before "what we'll achieve" |

See [references/bad-strategy.md](references/bad-strategy.md) when auditing a deck or plan — per-hallmark detection checklists, before/after rewrites, why bad strategy proliferates, and a step-by-step deck-audit procedure with a report format.

### 3. Sources of Power

**Core concept:** Good strategy applies strength where it has the greatest effect, drawing on recurring sources of power: **leverage** (anticipation, pivot points, concentration), **proximate objectives** (targets close enough to actually hit), **chain-link systems** (quality matched across links), **design** (premeditated, coordinated configuration), **focus**, and **using advantage** (asymmetries protected by isolating mechanisms).

**Why it works:** Resources are always scarce relative to ambitions. Power comes from asymmetry — knowing something rivals don't, pressing where effort is amplified, or concentrating where they are spread thin. A strategy that names no source of power is hoping effort alone will win, which is matching strength against strength.

**Key insights:**
- Leverage = anticipation × pivot point × concentration: anticipate predictable behavior, find the point where effort is amplified, then commit past the threshold where results become visible
- A proximate objective is one the team can see how to hit; under high ambiguity, choose closer targets — a JPL engineer made Moon-lander design feasible by simply *deciding* a lunar soil model others could build against
- In chain-link systems, performance is capped by the weakest link — investing in strong links is wasted until the weak one is fixed, which is why such systems stay stuck
- A fully matched chain is also the deepest moat: IKEA's in-house design, flat-pack logistics, and warehouse showrooms each fit the others, so copying one link gains a rival nothing
- Design-type strategy — tight, premeditated coordination of parts — pays when stakes are high and resources scarce; integration buys performance at the cost of flexibility
- An advantage matters only at the point of contention: deepen it, broaden it, or strengthen isolating mechanisms (network effects, brand, patents, tacit know-how) that block imitation

**Applications:**

| Context | Application | Example |
|---------|-------------|---------|
| Startup wedge choice | Concentrate past the threshold | One vertical owned end-to-end, not five touched |
| Stalled growth | Chain-link diagnosis | Fix activation (weakest link) before scaling paid acquisition |
| Ambiguous roadmap | Set a proximate objective | "Ten fintech design partners live" not "be the leader" |

**Ethical boundary:** Build isolating mechanisms on delivered value — lock-in engineered purely to trap users eventually isolates you from them.

See [references/sources-of-power.md](references/sources-of-power.md) when choosing where to apply strength — leverage, proximate objectives, chain-link systems, design, focus, and advantage, each with a when-to-use test (unlock test, addressability test, coherence check).

### 4. Riding Dynamics and Fighting Inertia

**Core concept:** Waves of change — technology shifts, deregulation, demographic change — are the attacker's best friend: they redistribute advantage and reset rules the incumbents had mastered. Incumbents are held back by three kinds of inertia (routine, culture, proxy) and by entropy — the unmanaged drift into blur and waste.

**Why it works:** In stable periods incumbents win on scale and accumulated advantage; in transitions their strengths become anchors — they defend legacy margins, rerun obsolete playbooks, and answer to cultures built for the old world. You don't need to predict the future, only to recognize that the present has already changed and act on it before those who can't.

**Key insights:**
- Guideposts for sensing waves: rising fixed costs (force consolidation), deregulation or rule changes, predictable biases (people extrapolate the present), incumbent response (watch them protect old margins), and attractor states (where the industry "should" land given the technology)
- An attractor state disciplines hype: ask "in the end state, who does the work and who gets paid?" — "all data transport becomes IP" correctly guided Cisco's rise
- Inertia by routine yields to new metrics and outside hires; inertia by culture requires simplification and breaking insulated units; inertia by proxy means the incumbent profits from its *customers'* inertia — banks kept paying low deposit rates because depositors were slow to move
- A rival's inertia is an exploitable asymmetry: attack where responding would force them to break their own economics
- Entropy shows up as blurred product lines, drifting prices, and accidental cross-subsidies — weeding it is real strategy work even with no competitor in sight

**Applications:**

| Context | Application | Example |
|---------|-------------|---------|
| Platform shift | Read the guideposts | Model training costs consolidate; value migrates to workflow owners |
| Pricing attack | Exploit margin defense | Usage-based pricing a seat-license incumbent can't match |
| Mature product | Entropy audit | Three overlapping plans collapsed into one clean ladder |

**Ethical boundary:** Ride waves by serving the new need better — never by manufacturing fear about the old one.

From wondelai/skills, MIT.

More from wondelai/skills