learn-strategic-thinking-with-phoebe / Session 8 of 8
Learn Strategic Thinking with Phoebe · Session 8 of 8

The strategy stack: which model when, and the board pitch

No new heavy frameworks today - you own fifty already. The capstone teaches the two skills that make them dangerous: SELECTING the right stack for the question in front of you, and ASSEMBLING seven sessions of Himalaya artifacts into one board-ready recommendation. Then we deliver it, the way it would be delivered in the room.

🔴 Capstone Managers & execs Bring all your Himalaya artifacts 45 min live + self-study
0-3 · Welcome 3-20 · Concepts 20-42 · Apply-along 42-45 · Send-off
Part 0

Where seven sessions land

You carved the stall (S1), decided among three bets (S2), dug the root cause (S3), read the battlefield (S4), redesigned the offer (S5), placed the growth horizons (S6), and found the customer leaks (S7). The frameworks did their jobs. Today they stop being a toolkit and become a strategy - and you learn the meta-skill this whole course was secretly about: knowing which model to reach for, and how their outputs snap together.

Live - presented in session Self-study - read after class ★ Apply-along on Himalaya The case: Himalaya, B2B2C SaaS
★ What you walk out with today A one-page decision tree that routes any business question to the right framework stack, Rumelt's kernel as the assembly frame for everything you have built, and Himalaya's final board pitch delivered live - plus YOUR kernel and 90-second pyramid pitch for the real problem you have carried since Session 1.
Part 1 · the meta-skill

Which framework when 7 min live

Fifty frameworks are useless if you reach for the wrong one - or for all of them. Selection starts with a single move: name the KIND of question you are holding. Every question in this course routes to one of six stacks, and each stack has a natural first tool.

What kind of question are you holding? Why is X happening? Which option? Where do we play? How do we make money? Where's growth next? Why aren't customers growing? Investigation stack · S3 is/is-not → fishbone → 5 whys Decision stack · S2 doors → matrix → pre-mortem External stack · S4 SWOT-TOWS · Five Forces · PESTEL Model stack · S5 BMC · Lean Canvas · VPC Growth stack · S6 lifecycle · Three Horizons · Ansoff Customer stack · S7 lifecycle · AARRR · LTV/CAC Route by the question, never by the framework you feel like using. Each stack's output becomes the next stack's input.
🔍 Click to zoom - the selection tree: six question types, six stacks, one route each
LiveThe three selection rules4 min
  • Start from the QUESTION, not the framework. "Let's do a SWOT" is a framework looking for a problem. "Why did the Q2 cohort churn?" is a question - and it routes itself: that is investigation work, and is/is-not goes first.
  • One framework per question - stacks, not soups. Within a stack the tools sequence (is/is-not scopes, fishbone widens, 5 whys deepens). Across stacks, pick one lane and finish it. A page with a SWOT, a canvas, and a fishbone on it answered no question at all.
  • Every framework's output feeds another's input. S3's root cause became S7's lifecycle fix; S4's TOWS actions became S6's horizon bets. If an analysis output feeds nothing downstream, ask why you did it.
Real world

The strategy offsite that produced eleven frameworks and zero decisions: a leadership team once spent two days filling every template in the facilitator's binder - and the CEO's closing summary was "great energy". Nobody had written the question on the wall first. The following quarter they ran ONE question ("where does the margin erosion come from?") through one stack, and cut the answer to the board in a week.

LiveThe 60-second triage3 min

Before any meeting, triage the question into one of four verbs - then the stack picks itself:

VerbSounds likeStack
Diagnose"Why is X happening?" · "What broke?"Investigation (S3), or Customer (S7) if the X is customer numbers
Decide"Which option?" · "Should we?"Decision (S2) - doors first to size the stakes
Design"How do we make money on this?" · "What's the offer?"Model (S5), with the External stack (S4) as the reality check
Grow"Where next?" · "What's the story for the raise?"Growth (S6), fed by Customer (S7) numbers
The exec move Say the triage out loud before the debate starts: "This is a diagnose question, so before anyone proposes fixes, let's spend ten minutes on what is and is not happening." You just chose the meeting's operating system in one sentence.
Part 2 · assembly

Stacking into one narrative 6 min live

Richard Rumelt's kernel (from Good Strategy Bad Strategy) is the assembly frame: a strategy is a diagnosis of what is really going on, a guiding policy for dealing with it, and a set of coherent actions that execute the policy. Deliver it pyramid-first (Minto, Session 1): answer, reasons, data underneath. Watch seven sessions of Himalaya work slot in.

Sessions 1 + 3 issue tree · root cause Sessions 4 + 6 outside read · horizons Sessions 2 + 5 + 7 verdict · starter tier · leak fixes 1 · DIAGNOSIS "The stall is an activation problem, not a market problem." 2 · GUIDING POLICY "Win on depth and the consumer side, not on price." 3 · COHERENT ACTIONS H1 · activation playbook + starter-tier experiment H2 · AI copilot, funded to a gate H3 · consumer-side network effect Each session's artifact slots into Rumelt's kernel - then the whole thing is told pyramid-first: answer, reasons, data underneath.
🔍 Click to zoom - the strategy stack assembled: seven sessions, one kernel
LiveWhat makes strategy BAD (Rumelt)3 min

Rumelt is more useful on bad strategy than on good, because bad strategy is what boards actually see. His three tells:

  • Fluff: abstraction dressed as insight. "We will leverage synergies to become the customer-centric platform of choice" says nothing a competitor would disagree with.
  • Goals mistaken for strategy: "grow 25%" is a wish. A strategy says what obstacle stands in the way and how you will get past it. Himalaya's board deck last year said "return to 30% growth" - and that was the whole slide.
  • The dog's-dinner action list: twelve initiatives, one per department, no policy connecting them. Everyone got their pet project funded; nothing compounds.
Real world

The tell you can spot in ten seconds: a strategy deck where deleting the company name and pasting in a rival's changes nothing. One exec team ran that exact test on their own annual plan - it survived the swap perfectly, which was the most useful finding of the offsite. A real diagnosis (like "our churn is a partner-onboarding gap") never survives it.

LiveThe coherence test3 min

Line up your actions and ask: could you delete one without weakening the others? If yes, it is a list, not a strategy. Coherent actions reinforce each other because they all execute the same policy.

  • Run it on Himalaya: the activation playbook (H1) creates multi-module habits → which is exactly the usage data the copilot (H2) learns from → which is what makes the consumer-side network effect (H3) more than a slogan. Delete H1 and H2 starves; the chain snaps.
  • Run it on the rejected path: "fix activation + push enterprise + cut price" fails instantly - enterprise pushes AWAY from the depth policy, and the price cut fights the starter tier's cleaner answer to the same need.
The exec move In any strategy review, pick the weakest-looking initiative and ask "what breaks if we cut this?" A confident, specific answer means coherence. A defensive silence means you found budget wearing a strategy costume.
Part 3 · the room

The board pitch 4 min live

Here is the whole course cashed in: Himalaya's recommendation, delivered exactly as it would be with the chairman across the table. Pyramid form, about 120 words, every number earned in a previous session.

LiveThe pitch, verbatim2 min

"We recommend: fix the leak, fund the copilot, explore the network effect. Three reasons. One - the stall is an activation problem, not a market problem: churn is concentrated in single-module clients onboarded by partner resellers with no activation playbook, while multi-module NRR is still healthy. Two - our right to win is depth and the consumer side, not price: two AI-native rivals own cheap-and-fast, and we will not outrun them there. Three - the economics are visible: closing the activation leak recovers most of the $8M annual NRR swing, and the starter-tier experiment validated a path to new logos. The ask: approve the 50-client activation pilot, fund the copilot to a Horizon 2 gate, and put one metric on the board pack - activation rate."

Why it lands Answer in sentence one. Reasons MECE (problem, right to win, economics). Actions coherent - each feeds the next. The ask is concrete, gated, and measurable. Nothing in it requires the deck; the deck is just the data layer waiting underneath.
LiveHandling the chairman's five questions2 min

Every board pitch meets the same five questions. Each one is answered in a single sentence, pulled straight from an earlier session's artifact:

QuestionOne-sentence answer (source)
Why now?NRR has slid from 118% to 98% over six quarters, and with ~30 months of runway the next raise needs a credible growth story, not a hope. (S1, S6)
Why us?Multi-module clients still expand healthily, and we hold the workflow depth plus 8M consumer touchpoints the AI-natives do not. (S4)
What kills this?The pre-mortem's top risk is that the pilot fixes setup but not habit - so the pilot measures weekly active usage, not onboarding completion. (S2)
What does it cost?The pilot redeploys existing CS capacity; copilot funding is staged and released only if H1 metrics hold at the gate. (S2, S6)
How do we know it's working?One board metric: activation rate - new clients live on a second module within 90 days. (S7)
Real world

The one-sentence discipline is the point: a CFO once told a strategy lead "you answered my kill question in nine words, so I stopped trying to kill it." Boards probe for whether YOU have already probed. Pre-answered questions read as de-risked strategy; improvised answers read as optimism.

Part 4 · the deep end

After the room: anti-patterns and the execution bridge

Self-studyThe anti-patterns gallery4 min read
  • Framework theater: decks full of beautifully filled templates and no decision. The test from Part 1 applies - if an output feeds nothing downstream, it was performance, not analysis.
  • The SWOT graveyard: a SWOT that never became TOWS actions (Session 4's whole point). Strengths lists are where strategy goes to feel busy.
  • Analysis as procrastination: the fifth layer of the issue tree, the third market study - rigor that exists to delay a decision someone is afraid to make. The doors test (S2) is the antidote: if it is reversible, decide today.
  • The McKinsey 7S name-check: dropped into decks to sound thorough, almost never actually used. Honest note: 7S IS useful - for org alignment AFTER a strategy change, checking that structure, systems, staff, skills, style, and shared values line up behind the new direction. Himalaya should run it in six months, not now.
Self-studyOKR: the execution bridge3 min read

Strategy without an execution bridge decays into the dog's dinner within two quarters. Andy Grove's OKRs (evangelized by John Doerr in Measure What Matters) are the simplest bridge: the strategy's H1 becomes one objective with three measurable key results. For Himalaya's activation fix:

Himalaya H1, as an OKR
ObjectiveNew clients build a multi-module habit in their first 90 days
KR1Activation rate (second module live within 90 days) from ~30% to 55%
KR2100% of partner-reseller onboardings run the activation playbook (from 0%)
KR3Single-module cohort churn run-rate cut by a third by the H2 funding gate

Note what the OKR is NOT: it is not the strategy. It is the strategy's first 90 days, made countable.

Self-studyBalanced scorecard: keeping the horizons honest2 min read

Kaplan & Norton's balanced scorecard (HBR, 1992) tracks four perspectives - financial, customer, internal process, learning & growth - precisely so the financial view cannot strangle the others. For Himalaya it is the guard rail on the three-horizon plan: H1's activation economics live in the financial and customer rows, but the copilot (H2) and network effect (H3) are learning bets - measured on validated learning, not revenue, until their gates. A board that scores all three horizons on this quarter's revenue will kill its own future and call it discipline.

Apply-along 1 of 3

The triage drill ★ 7 min · tables route

Facilitator reads six real-sounding questions, one at a time: "our best engineers keep leaving" · "should we open a Singapore office" · "the new tier isn't selling" · "margin is eroding 2 points a year" · "which of these 3 vendors" · "the board wants an AI story".

For each: tables get 45 seconds to route it through the tree - name the question TYPE, the stack, and the FIRST framework they would reach for.

Compare answers out loud. Expected routes: engineers leaving → diagnose → is/is-not first. Singapore → decide → doors first (it is a one-way-ish door, so slow down). New tier not selling → diagnose BEFORE redesigning the offer - is/is-not, not a new canvas. Margin erosion → diagnose → is/is-not, then fishbone. Three vendors → decide → weighted matrix. AI story → grow → Three Horizons first.

Debrief the traps: two of the six tempt you into the wrong stack (the tier question begs for a canvas, the AI question begs for a tech deep-dive). The question type, not the topic, picks the stack.

Apply-along 2 of 3

Write YOUR kernel ★ 8 min · everyone writes

Take the real problem you have carried through the homework since Session 1 - the issue tree from week one, whatever artifacts you built since.

Write the three kernel lines: DIAGNOSIS (what is really going on, one sentence, "X not Y" form), GUIDING POLICY (how you will deal with it - a direction that says no to something), COHERENT ACTIONS (2-4 moves that reinforce each other).

Run the two tests on your own draft: the rival-name swap on the diagnosis (would it survive being pasted into a competitor's deck?), the delete-one test on the actions.

Swap with a neighbor: each reads the other's kernel cold and says the diagnosis back in their own words. If it comes back as fluff, the diagnosis was fluff.

Apply-along 3 of 3

Pitch it: 90 seconds, answer first ★ 7 min · 3 volunteers

Three volunteers deliver their pyramid pitch to the room, 90 seconds each: answer sentence, reasons, the ask. Standing up. No slides.

The room scores each on three checks, thumbs up or down: answer first? (did sentence one carry the recommendation) · reasons MECE? (no overlaps, no obvious gap) · actions coherent? (would deleting one weaken the rest).

Thirty seconds of feedforward per pitch (one improvement, not a review). Then the closer, back at Session 1's promise: compare this pitch to the three-sentence pyramid you wrote that day. That distance is the course.

Real world

The 90-second constraint is not theater - it is the actual size of the slot. Board agendas slip, the chairman's flight boards at six, and the corridor version IS the pitch more often than the scheduled one. Strategists who can only deliver the 30-minute version do not get to deliver at all.

After the session - the send-off

This is now yours ◐ the rest of your career

Check yourself

Three questions before you go 🎯 ◐ 90 seconds

1 · A colleague opens a working session with "let's start with a SWOT". The first selection rule says...

Frameworks answer question types. Until the question is named (diagnose? decide? design? grow?), any framework is a soup ingredient looking for a soup.

2 · Rumelt's kernel of a good strategy has exactly three parts. They are...

Diagnosis (what is really going on), guiding policy (the direction that says no to something), coherent actions (moves that reinforce each other). Everything else is packaging.

3 · The coherence test on an action list asks...

Owners and deadlines are execution hygiene, and one-per-department is exactly Rumelt's dog's dinner. Coherent actions compound - remove one and the chain visibly weakens.

Source material

Frameworks covered & their origins

The capstone teaches selection and assembly rather than new machinery - the sources below are where the assembly ideas come from, and where to go deeper.

Framework-selection decision tree ("which model when")this course - Part 1 + Apply-along 1
Stacking frameworks into one board narrativeMinto pyramid + Rumelt kernel - Part 2 + Apply-along 2
Himalaya final recommendation walkthroughrunning case - Part 3 + Apply-along 3
Anti-patterns: framework theater, SWOT graveyard (Rumelt, Good Strategy Bad Strategy)self-study card, Part 4
McKinsey 7Sone-row cameo in the anti-patterns card - useful for org alignment AFTER a strategy change
OKR as execution bridge (Andy Grove; John Doerr)self-study card, Part 4
Balanced scorecard (Kaplan & Norton, HBR 1992)self-study card, Part 4

Session 8 cheat sheet · pin this

The triageName the verb first: diagnose, decide, design, or grow. The verb picks the stack; the stack picks the first tool.
Selection rulesQuestion before framework · one stack per question, never soups · every output must feed a downstream input.
Rumelt's kernelDiagnosis → guiding policy → coherent actions. A policy that says no to nothing is fluff; a goal is not a strategy.
Coherence testDelete one action - do the others weaken? No = it is a list. Also run the rival-name swap on the diagnosis.
The board pitchAnswer in sentence one, three MECE reasons with numbers, a concrete gated ask, ONE board metric. 120 words, 90 seconds.
The whole course in one lineCarve it, map it, decide it, verify it, model it, time it, price it, tell it top-down.