Why this course exists
Most managers collect frameworks like souvenirs - a SWOT here, a canvas there - and still walk into the board meeting with opinions instead of structure. This course teaches 50+ thinking models the way strategists actually use them: as a stack, applied to one real business. Meet Himalaya, a $40M ARR B2B2C SaaS platform whose growth has stalled. Over 8 sessions you will diagnose it, decide for it, and build its board-ready strategy - framework by framework.
MECE: carve the problem cleanly 5 min live
Himalaya's board asks "why is growth stalling?" and gets three answers: the CPO says product, the CRO says market, the CS lead says churn. All plausible, all overlapping, none complete. MECE - Mutually Exclusive, Collectively Exhaustive - is the discipline that turns that noise into a clean carve.
LiveMutually exclusive, collectively exhaustive3 min▶
MECE is a quality bar for any list you make: categories that do not overlap (mutually exclusive) and that together cover everything (collectively exhaustive). It is the single most-used idea in consulting for a reason - a MECE carve means nothing gets counted twice and nothing gets missed.
- Cheap MECE cuts that almost always work: new vs existing · internal vs external · people/process/technology · before/during/after · a formula (revenue = clients × price × usage).
- The formula trick is the strongest: decompose a metric into its arithmetic parts and the carve is MECE by construction. Growth = new logos + expansion - churn. Done.
- MECE is for analysis, not communication: you carve MECE in private, then present only the branches that matter (that's 80/20, Part 4).
The double-count that burned a real board: a leadership team listed "churn", "product gaps", and "onboarding issues" as three separate problems and funded three separate fixes - but two of them were the same clients counted twice. A MECE carve by revenue line would have caught it in five minutes and saved a quarter of duplicated effort.
Self-study5W2H: the completeness checklist2 min read▶
When you cannot find a clean carve, brute-force exhaustiveness with 5W2H - What, Why, Where, When, Who, How, How much. It comes from Toyota's factory floor and works anywhere: run the seven questions over a problem and the gaps announce themselves.
| Question | On Himalaya's stall |
|---|---|
| What exactly stalled? | ARR growth rate: 45% → 12% in 3 years |
| When did it start? | NRR began sliding 6 quarters ago |
| Where (segment)? | Concentrated in single-module clients |
| Who is affected? | 1,200 business clients; 8M end consumers untouched |
| How much? | 98% NRR vs 118% two years ago ≈ $8M/yr swing |
Notice what happened: three questions in, "growth is stalling" became "single-module clients started churning six quarters ago". That is already half a diagnosis.
Issue trees: the question, exploded 5 min live
Stack MECE carves two or three layers deep and you get an issue tree - the complete map of where an answer could possibly live. It is how strategists guarantee they looked everywhere before claiming they found the cause.
LiveGrowing a tree that actually works3 min▶
- Start with the question, not the answer. The root is always a question ("why has X stalled?" / "how might we double Y?"). Diagnosis trees ask why; solution trees ask how.
- Each split is MECE. The formula trick shines here: revenue math gives you layer one for free.
- Stop at testable leaves. A leaf earns its place when you can name the data that would confirm or kill it. Then you prioritize which leaves to test first - that is hypothesis-driven work (self-study below).
- Two or three layers is plenty. Five-layer trees are procrastination wearing a suit.
Self-studyHypothesis-driven problem solving: test, don't boil3 min read▶
An issue tree maps everywhere the answer COULD be; hypothesis-driven work is choosing where to look FIRST. Instead of analyzing every branch (boiling the ocean), you bet: "I believe the stall is mostly existing-revenue leakage, and I can check NRR by cohort by Friday."
- Form the hypothesis from the strongest early signal (Himalaya: churn concentrated in single-module clients).
- Define the kill test: what data would prove you wrong? If you cannot name it, it is a belief, not a hypothesis.
- Timebox it. One week per hypothesis round. Wrong fast beats right slow - the tree holds your other options warm.
McKinsey teams call this "day-one answer": on day one of a study, they commit a best-guess answer and spend the project trying to break it. It feels arrogant; it is actually humility with a deadline - every analysis has to earn its place by testing the answer.
Self-studyFirst-principles thinking: when the tree needs new roots2 min read▶
Issue trees decompose the problem as framed. First-principles thinking questions the frame itself: strip the situation to physical facts and rebuild. Himalaya framed as "how do we stop churn?" assumes clients should stay; first principles asks "what would make a business PAY for a consumer platform in 2026, from scratch?" - which is where the AI-native competitors started, and why they onboard in days.
- Use sparingly: re-deriving everything from atoms is slow. Reach for it when incumbents (including you) all share an assumption a newcomer could ignore.
- The pairing: first principles to choose the right question, issue tree to explore it, hypotheses to test it. Session 4 returns to this when we read Himalaya's competitors.
The Pyramid Principle: answer first 5 min live
You carved, you mapped, you found something. Now you have four minutes of the chairman's attention. Barbara Minto's Pyramid Principle is the single highest-leverage communication idea ever taught to consultants: lead with the answer, support with grouped reasons, keep the data underneath.
LiveAnswer → reasons → data3 min▶
- Lead with the answer. Not background, not method, not suspense. Executives listen top-down; every minute before the answer is a minute they spend guessing it.
- Group reasons in threes-ish, and make each group MECE-clean. Under each reason: the data, shown only if asked. The pyramid is your issue tree, inverted and pruned.
- The situation-complication-question opener: one sentence of agreed context ("we grew 45%, now 12%"), one twist ("the board meets in 6 weeks"), then the question your answer resolves. Thirty seconds, then the pyramid.
The email test: Minto's rule survives because of moments like this - a VP sent a 14-paragraph analysis; the CEO read two lines and replied "so what do you want?". The rewrite led with "I recommend we pause the enterprise push; three reasons below." Approved in 11 minutes. Same analysis, different architecture.
80/20: where the leverage lives 3 min live
Pareto's observation - most effects come from few causes - is the strategist's permission slip to ignore things. Your issue tree has eight leaves; the data will point hard at two. 80/20 says: test those two first and say so out loud.
LivePareto as a scoping weapon3 min▶
- On problems: Himalaya's churn is 14% overall, but single-module clients churn at 3× the rate of multi-module ones. One segment, most of the leak. That is where Session 3's investigation goes.
- On effort: before any analysis, ask "which 20% of this work produces 80% of the decision?" A cohort table you can pull today usually beats a perfect model you get next month.
- On portfolios: expect revenue, bugs, complaints, and feature usage to be Pareto-shaped. If your top 5 clients are NOT 40%+ of revenue, that is itself a finding.
- The honest caveat: 80/20 prioritizes, it does not conclude. You still verify the tail is not hiding a landmine (Session 2's pre-mortem catches those).
Carve Himalaya's stall, MECE ★ 7 min · everyone carves
The brief, read aloud: Himalaya, $40M ARR B2B2C platform, growth 45% → 12%, NRR 98% and falling, churn 14% concentrated in single-module clients, win rate 31% → 22%, two AI-native competitors priced 40% lower.
Two minutes solo: carve "why has growth stalled?" into a first MECE layer. Hint if stuck: what is the revenue formula?
Compare with a neighbor: run the two tests on each other's carve - can one fact land in both buckets? Can you name a fact that lands in neither?
Converge on the room's layer one: new revenue down vs existing revenue leaking. Notice several different carves passed the test - MECE is a bar, not a single right answer.
Grow the issue tree ★ 9 min · everyone maps
Split each layer-one branch once more: new revenue → fewer at-bats vs losing more often; existing revenue → churn up vs expansion down. Check each split is MECE.
For every leaf, write the data that would test it (pipeline count, win rate by segment, churn by cohort, upgrades per client per year). A leaf with no testable data gets rewritten.
80/20 pass: circle the two leaves the brief's numbers already point at (win rate collapse; single-module churn). Say the sentence out loud: "We will test these two first, and we know what we are NOT testing yet."
Photograph your tree. It is Session 3's starting point - the investigation session digs the circled leaves to root cause.
The sentence in step 3 is the whole game. "Here is everywhere the answer could be, here is where we are looking first, here is what we are consciously deferring" - said to a chairman, that sentence buys more trust than any deck. It sounds like control, because it is.
Pyramid it in three sentences ★ 6 min · everyone writes
Pretend the circled leaves confirmed (they will - Session 3). Write ONE answer sentence for Himalaya's board: your best "the stall is X, not Y" formulation.
Under it, three reason bullets - each one a branch of your tree, each backed by a number from the brief.
Read a volunteer's aloud. Room checks: did the answer come first? Could the chairman stop listening after sentence one and still know what you think?
This week ◐ 30 min total
- Pick one real problem from YOUR desk and build its issue tree - root question, two MECE layers, testable leaves, two circles. Ten minutes, pen and paper.
- Rewrite one email pyramid-style: take something you sent this month that buried the ask, and rebuild it answer-first. Compare lengths.
- Read the self-study cards above: 5W2H, hypothesis-driven problem solving, first principles - each pairs with a live tool.
- Optional deep end: Barbara Minto's The Pyramid Principle chapters 1-3, or Conn & McLean's Bulletproof Problem Solving chapter 1 (the seven-step loop this whole course walks).
Three questions before you go 🎯 ◐ 90 seconds
1 · A carve of churn causes reads: "product gaps, onboarding issues, unhappy clients". What is wrong with it?
"Unhappy clients" contains the other two, and nothing covers pricing or competitors. Run both tests: overlap check, gap check.
2 · What earns a branch the right to be a LEAF on an issue tree?
Testability is the bar. "Win rate fell" points at data; "morale is low" points at a feeling. Depth is whatever gets you to testable - usually 2-3 layers.
3 · Four minutes with the chairman. The Pyramid Principle says you open with...
Answer first, always. One sentence of situation-complication is allowed as a runway - then the answer. Executives listen top-down.
Frameworks covered & their origins
This course teaches the working 80% of each framework and cites the canon - the original books and papers stay the deep end for anyone who wants it.