Bulletproof Problem Solving

Source: Charles Conn & Robert McLean, "Bulletproof Problem Solving: The One Skill That Changes Everything" (Wiley, 2019)

The McKinsey 7-step problem solving process. This skill codifies the entire framework into modular components you can pull individually. Each component is standalone -- use the one that fits your task.

Core principle: Problem solving is an iterative cycle, not a linear checklist. You "porpoise" between steps -- diving into data, surfacing to refine your hypothesis, diving again. The "one-day answer" (your best current synthesis) should be ready at any point.

The 7 Steps at a Glance

1. Define the Problem
2. Disaggregate (logic trees)
3. Prioritize (prune the tree)
4. Build a Workplan
5. Conduct Analysis
6. Synthesize Findings
7. Tell the Story

Each step feeds the next, but you iterate backwards constantly. A finding in Step 5 may send you back to redefine Step 1.

Step 1: Define the Problem

The most underrated step. A surprising number of failures originate in poor problem definition. Teams surge into data gathering without being clear about boundaries, criteria for success, time frame, or accuracy required.

Problem Statement Qualities (SMART+)

A good problem statement is:

1. Outcomes-focused -- expressed as outcomes, not activities or intermediate outputs
2. Specific and measurable wherever possible
3. Time-bound -- clear deadline
4. Decision-maker aligned -- explicitly addresses their values, boundaries, aspirations, accuracy needs
5. Sufficiently scoped for creativity -- too narrow kills novel solutions
6. Solved at the highest level possible -- for the whole org, not just one unit

Mnemonic: SMART = Specific, Measurable, Action-oriented, Relevant, Timely. But also add: outcomes focus + highest level.

Problem Definition Worksheet

Don't try to fit everything into one sentence. Use a worksheet:

Porpoising

Problem statements get better when facts are brought to bear. Set up a dialogue with questions, dive into some data, resurface to sharpen the statement. The Pacific salmon team iterated their problem statement 4+ times over years, each version sharper than the last.

Anti-pattern: "I've seen this before" -- rushing to a solution based on analogy without testing fit. This is how newspapers got disrupted (they defined the threat as "better editorial content" when the real attack came from classified ad erosion).

Testing Your Problem Statement

• Try a challenging antithesis or counterfactual to test robustness
• Write it from the CEO's perspective -- is it different?
• Write it from a competitor's perspective -- what would they see?
• Add elements of design thinking (empathize, define, ideate) -- particularly active empathy for the user

Step 2: Disaggregate the Problem (Logic Trees)

The most important step. Any problem of real consequence is too complicated to solve without breaking it into logical parts. The magic comes in seeing which type of tree makes an elegant solution path obvious -- like cutting a diamond the right way.

Types of Logic Trees

MECE Principle

Mutually Exclusive, Collectively Exhaustive.

• Mutually Exclusive: Branches don't overlap or contain partial elements of the same factor
• Collectively Exhaustive: Taken as a whole, the tree contains ALL elements of the problem

If you're having trouble getting clear hypotheses, check if your tree is really MECE. Government policy, for example, is often incorrectly shown as its own branch when it actually affects every other branch -- that's an overlap violation.

Tree Construction Process

1. Start with a simple component tree (days of research, not weeks)
2. Read literature, talk to experts -- just enough to generate a first-cut tree
3. Iterate: as understanding grows, transition from components to hypotheses
4. Try 2-3 alternative disaggregations to find which yields the most insight
5. Use Post-it notes on a whiteboard -- move branches around until hierarchy clicks
6. Aim to go 3-4 levels deep (trunks -> branches -> twigs -> leaves)

The ROIC Tree (Business Essential)

For almost any business problem, start with a Return on Invested Capital tree:

ROIC = Profit Margin x Capital Turnover

Profit Margin = (Revenue - Costs) / Revenue
  Revenue = Price x Volume x Market Share
  Costs = COGS + SGA + Overhead
    COGS = Materials + Labor + Distribution
    SGA = Sales + Marketing + Admin

Capital Turnover = Revenue / Invested Capital
  Working Capital = Receivables + Inventory - Payables
  Fixed Assets = PPE + Other

This tree makes "what-if" competitive scenario analysis easy. The Hechinger vs Home Depot case revealed that Home Depot compensated for 15-18% lower prices with lower distribution costs (-5-7%), cheaper sourcing (-5%), lower overheads, and much higher asset productivity (2x sales per store).

Step 3: Prioritize (Prune the Tree)

Good problem solving is as much about what you DON'T do as what you do. Before investing significant time, prune branches.

The Prioritization Matrix (2x2)

Plot each branch/lever on a 2x2 matrix:

HIGH  |  Consider    |  FOCUS HERE
      |  carefully   |  (big + moveable)
      |              |
Can   |--------------|--------------
we    |              |
move  |  Ignore      |  Acknowledge
it?   |  (small +    |  but don't
      |   immovable) |  invest
LOW   |              |  (big + immovable)
      +--------------+--------------
         LOW              HIGH
         
         Impact on the problem

Decision rules:

• High impact + High ability to move it = Focus here. This is your critical path.
• High impact + Low ability = Acknowledge. Example: climate's effect on salmon -- huge impact, but a foundation can't change it. Lop it off.
• Low impact + High ability = Consider if easy wins. Otherwise skip.
• Low impact + Low ability = Ignore completely.

When to Revisit Pruning

• After relaxing constraints in the problem statement
• When early analysis changes magnitude estimates
• When you're stuck and need to check if you pruned a promising branch

Step 4: Build a Workplan

A good workplan prevents the trap of "boiling the ocean" with unfocused analysis. It saves enormous time.

👉 For the full workplanning method, templates, knock-out analysis logic, "dummy the chart" discipline, chunky-plan + lean-Gantt structure, SCR one-day answers, anti-patterns, and the pre-launch checklist, use the dedicated /work-plan skill.

Quick reference (full detail in /work-plan)

The 5 core rules:

1. No analysis without a hypothesis -- never build a model without a question it answers
2. Dummy the chart -- sketch the output exhibit before doing the analysis
3. Knock-out order -- analyses that can eliminate later work go FIRST
4. Painfully explicit -- WHO does WHAT using WHAT source by WHAT date
5. Chunky + Lean -- 2-3 week detailed workplan + lean Gantt for milestones

The 6-column template:

The one-day answer (SCR):

• Situation: state of affairs at the outset
• Complication: what changed, what's the tension
• Resolution: your best current idea of what to do

Use one-day answers as a strawman to attack, an elevator-test answer, and a guide for remaining analysis.

Step 5: Conduct Analysis

Start simple. Escalate only when needed.

Heuristics & Rules of Thumb (Start Here)

Question-Based Problem Solving (Sherlock Holmes)

Ask who, what, where, when, how, why to paint a picture of the problem:

• Often you can get to rough-cut solutions in as few as 3 questions
• Heart attack patient classification: 3 questions (blood pressure > 91? Age > 62.5? Tachycardia present?) segregate high vs low risk
• The 5 Whys (Toyota): Keep asking "why?" 5 times to push through proximate causes to root causes. Visualize as a fishbone diagram.

When to Use Big Guns

Escalate to advanced analytics only after heuristics have sized the problem:

Decision rule: Are you trying to understand DRIVERS (left branch -- statistics, experiments) or PREDICT outcomes (right branch -- ML, game theory, simulations)?

Step 6: Synthesize Findings

Synthesis is where the magic happens. Individual branch-level findings combine into an overall picture, and cross-cutting insights emerge.

How to Synthesize

1. Draw pictures. Humans are visual learners. Represent each finding as a graphic highlighting the insight.
2. Map findings to your tree. Show each graphic as a branch on your revised tree structure. This illuminates the overall problem and lets insights emerge across branches.
3. Team synthesis sessions. Have each member present their analyses to the assembled group. Checkpoint for alternative hypotheses.
4. Look for the cross-cutting insight. Often the biggest insight wasn't obvious in branch-level work. BHP's team only saw the massive option value of nearby deposits when they synthesized findings from the transportation infrastructure analysis.

From One-Day Answers to Pyramid Structure

Early stage: reasoning toward implications (situation -> observation -> resolution)

Later stage: marshaling reasons to drive action ("You should do X because 1, 2, 3")

Step 7: Tell the Story

Done right, your conclusions are an engaging story that convinces someone to act. Done wrong, you get the Anxious Parade of Knowledge (APK) -- reams of data with no point of view.

The Pyramid Principle (Barbara Minto)

Structure your communication as a pyramid:

        [Governing Thought / Answer]
       /           |              \
  [Reason 1]   [Reason 2]    [Reason 3]
  /    \        /    \         /    \
[Data] [Data] [Data] [Data]  [Data] [Data]

• Top: Your answer + call to action (lead with this for most audiences)
• Second level: Major supporting arguments (3 is typical)
• Third level: Data, analyses, evidence backing each argument
• Two structures: Deductive (general rule -> specific application -> conclusion) or Inductive (observation -> observation -> general conclusion). Both work; mix them.

Situation-Complication-Resolution (SCR)

The governing thought at the top is usually your latest SCR one-day answer:

• Situation: Context everyone agrees on
• Complication: What changed, what's at stake, the tension
• Resolution: What to do about it (your recommendation)

For most audiences: Lead with the Resolution. "You should do X." Then provide the reasons.

For Difficult Audiences (Revealed Approach)

When your conclusion is unwelcome, use a decision tree structure instead of a pyramid. Provide evidence for each yes/no branch, slowly revealing the answer. Layer by layer, the audience reaches the conclusion themselves.

Storyboarding

1. Divide paper into 9 boxes
2. Cartoon/dummy each exhibit with a storyline lead across the top
3. Read across the leads -- check horizontal logic (does the narrative flow?)
4. Each slide's title should be an argument, not a boring label

Cleaving Frames (Advanced)

Expert problem solvers use existing frameworks to quickly and elegantly disaggregate problems. Try multiple frames -- different slices yield different insights.

Business Cleaving Frames

Societal Cleaving Frames

Personal Cleaving Frames

Using Cleaving Frames

1. Try a frame on your problem
2. Do quick back-of-envelope calculations to confirm it yields insight
3. If it reveals clear leverage points, use it
4. If not, try another frame
5. Often combine 2+ frames for richer analysis

Bias-Fighting Team Processes

The Big 5 Biases

Team Process Toolkit

1. Obligation to dissent -- McKinsey core value. Not just OK for juniors to disagree -- it's mandatory. Senior people who squash dissent can be fired.
2. Try multiple trees/cleaves -- Even when a frame seems perfect, try 2-3 alternatives in a 30-min session.
3. Add question marks to hypotheses -- Activates discovery mode vs. confirmation mode.
4. Role-playing -- Act out solutions from client/supplier/competitor perspective. Embarrassing at first. Remarkably clarifying.
5. Dialectic standard -- Every idea must be met with its antithesis before synthesis.
6. Perspective taking -- Model another's belief compellingly enough to argue it yourself.
7. "What would you have to believe?" -- Spell out all implicit assumptions. A dot-com stock justified at 40% compound growth for 5 years, then 2x GDP in perpetuity? Didn't happen.
8. Distributed voting -- 10 sticky notes per person, allocate to priority analyses. Senior person votes LAST.
9. Solicit outside views -- But be careful with experts (they often reinforce the mainstream view). Talk to customers, suppliers, adjacent industries.
10. Pre-mortem analysis -- Work through all implications of even unlikely failure cases. Model extreme downsides (Black Swan thinking).
11. Diversity in team members -- Different backgrounds = more creativity + less bias. Teams always outperform individuals (Tetlock). The best superforecasting teams even beat prediction markets.

Uncertainty Framework (5 Levels)

Actions by Uncertainty Level

Strategic Staircases (Growth Under Uncertainty)

For building new businesses in uncertain environments:

1. Work backward from planned outcome -- what capabilities, assets, learning are required?
2. Step-by-step strategic moves, expanding commitment as uncertainty reduces
3. Frame in time -- staircase steps, knowledge captured, capabilities built

Three considerations:

• Stretch: Distance of step-out from established platform. Balance absorption capacity vs. speed demands.
• Momentum: Positive effect of early success. Critical in winner-take-all markets.
• Flexibility: Maintain fluidity. Pick step-outs with greatest option value for follow-on moves. Avoid sunk assets that can be stranded.

Wicked Problems

Problems that seem to defy definition because they're embedded in complex systems. Multiple causes, values disagreements, unintended consequences, behavioral change required.

How to Attack Wicked Problems with 7 Steps

1. Don't put them in the too-hard basket. The same framework works -- you just need more iterations.
2. Set ambitious targets. MGI set a goal of 20% obesity reduction in 5 years in the UK.
3. Cleave along supply/demand with a cost curve. Rank interventions by cost per unit of impact. Start with the widest, cheapest bars.
4. Use a theory of change map for multi-year problems -- your strategic map for visualizing change at scale.
5. Build a portfolio of strategies matrix -- balance seed/cultivation/harvest stages and incremental/transformational aspirations.
6. Make externalities endogenous -- the California fisheries case solved overfishing by creating market-based property rights (catch shares + community quota funds) that internalized the commons problem.
7. Use multiple competing and complementary strategies -- some will succeed, some won't. Manage trade-offs as the world unfolds.

Warning Signs of a Wicked Problem

• Problem changes shape when you intervene (welfare dependency)
• No single right answer (nuclear energy)
• Values play a central role (gun control)
• Real problem is nested inside more apparent problems (homelessness: shelter is visible, but root causes are domestic violence, mental health, addiction)

Quick Reference: Applying the Method

For a Business Decision

1. Define: What decision, by when, with what accuracy?
2. Disaggregate: Start with ROIC tree or profit lever tree
3. Prioritize: Which levers have biggest impact and can you move them?
4. Workplan: Knock-out analysis first (does the economics work?)
5. Analyze: Heuristics -> order of magnitude -> detailed if needed
6. Synthesize: Map findings to tree, look for cross-cutting insights
7. Story: SCR + Pyramid. Lead with recommendation.

For a Personal Decision

1. Define: What am I choosing between? What matters to me? (Factors + weights)
2. Disaggregate: Weighted factor tree (like where-to-live: 20 variables, data for 12 options)
3. Prioritize: Which factors actually differentiate between options?
4. Workplan: Gather data online. Most data is available for free.
5. Analyze: Normalize to 1-100 scale, apply weights, simple arithmetic
6. Synthesize: What does the data say? Does it match my gut?
7. Story: Explain the rationale to family/stakeholders so they buy in

For a Policy/Societal Issue

1. Define: Problem statement as citizen. What's the real issue, not the popular framing?
2. Disaggregate: Try multiple cleaving frames (incidence/severity, supply/demand, mitigate/adapt)
3. Prioritize: What's the highest-impact lever that can actually be moved?
4. Workplan: A few hours of online research + phone calls
5. Analyze: Follow the evidence. Which proposed interventions actually address root causes?
6. Synthesize: Can you move from emotional position to reasoned one?
7. Story: Explain to others why you reached your conclusion

The 10 Commandments of Great Problem Solving

1. Take time upfront to understand your problem. The initial problem posed often isn't the right question.
2. Get started with nothing more than a problem statement. Whiteboard + pencil. Don't wait for data.
3. Try several cuts at the tree. Sticky notes. Move them around. Test 2-3 cleaving frames.
4. Use a team wherever you can. Diversity in thinking reduces bias and deepens creativity.
5. Make the right investment in a good workplan. Prune savagely. Be painfully explicit. Chunky plans, 2-4 weeks out.
6. Start analysis with heuristics. Summary statistics, rules of thumb, back-of-envelope. Often sufficient to close the book.
7. Don't be afraid of big guns when required. Regression, ML, game theory, experiments are more accessible than ever.
8. Put as much effort into synthesis and storytelling as analysis. The problem isn't solved at your Eureka moment. You have to persuade someone to act.
9. Treat the 7 steps like an accordion. Compress or expand depending on the issue. One-day answers at every stage.
10. Don't be intimidated by any problem. Even wicked ones. Systematic problem solving yields insight into almost any problem of consequence. "That is why bulletproof problem solving feels like a superpower when you get good at it."