Case Studies: Good Strategy Bad Strategy in Practice skill

Case Studies: Good Strategy Bad Strategy in Practice

Table of Contents

• Case Study 1: Auditing a SaaS Annual Plan
• Case Study 2: A Startup Chooses Where to Concentrate
• Case Study 3: Rewriting a Vision Deck into a Kernel
• Key Takeaways

Case Study 1: Auditing a SaaS Annual Plan

Context

A 60-person B2B SaaS company sells compliance-training software to mid-market HR teams. $9M ARR, growth slowing from 40% to 12% year over year. The leadership team produced a 34-slide annual plan titled "Strategy FY27" and asked for a review before presenting it to the board.

The Document

The deck contained: a vision slide ("To be the most trusted partner in workforce compliance"), three values slides, a market-size slide ($4.2B TAM), targets (25% ARR growth, NRR from 103% to 112%, logo churn under 8%), and fourteen initiatives spanning every department — AI course generation, a manager dashboard, SOC 2 Type II, a partner program, two new verticals, a website refresh, and eight more.

Applying the Skill

Step 1 — Inventory and classification. Every slide was classified as diagnosis, policy, action, goal, evidence, or fluff. Result: 0 diagnosis slides, 0 policy slides, 19 goal/initiative slides, 6 fluff slides, 9 evidence slides. The headline finding wrote itself: the plan contained targets and activities but no statement of what stands in the way.

Step 2 — Hallmark scan. Fluff: "trusted partner in workforce compliance" failed the negation test (no rival aspires to be untrusted). Goals for strategy: the verb audit found grow, increase, deliver, expand, accelerate — not one choice verb. Bad objectives: fourteen initiatives for nine engineers was a dog's dinner; nothing conflicted with anything, the signature of an assembled rather than chosen list.

Step 3 — Hunting the missing diagnosis. The audit reconstructed the challenge from evidence the deck itself contained but never confronted: 70% of historical leads came from SEO content; organic traffic was down 45% in three quarters as AI answers absorbed compliance questions; win rates were stable. The company did not have a growth problem — it had a demand-generation collapse the plan never mentioned, plus a quieter fact buried in the churn appendix: accounts using the audit-trail feature churned at one-third the rate of the rest.

Step 4 — Kernel rewrite. Working session with the exec team produced one page:

• Diagnosis: Our acquisition engine (SEO → trial) is structurally broken by AI search, and we are funding fourteen initiatives as if it still worked. Meanwhile our stickiest value — audit-proof training records — is treated as a feature, not the product.
• Guiding policy: Reposition from "training content" to "audit protection" — the system of record HR shows the regulator — and rebuild demand on channels AI answers cannot absorb: compliance-auditor partnerships and integration marketplaces. Therefore we will not: chase the two new verticals, build AI course generation this year, or bid on broken SEO terms.
• Coherent actions: (1) Audit-trail becomes the lead product and demo (product, Q1); (2) partner program rebuilt exclusively for compliance auditors and HRIS marketplaces (one senior hire, Q1-Q2); (3) pricing anchored to audit risk, not seat count (Q2); (4) kill eight initiatives, redeploying six engineers (immediately).

Outcome

Aspect	Before	After
Plan length	34 slides	1 kernel page + 6 evidence slides
Initiatives	14	4 coordinated actions + kill list
Challenge named	Nowhere	First paragraph
Board reaction	(prior year) "ambitious"	"First time we understood the business"
Two quarters later	—	Partner-sourced pipeline 0% → 31%; NRR 103 → 109

Lessons

1. The decisive audit move was classification — counting zero diagnosis slides ended the debate about whether the plan was a strategy.
2. The diagnosis was already in the appendix data; bad strategy is usually evidence avoidance, not evidence absence.
3. The kill list freed more capacity than any hiring plan could have.

Case Study 2: A Startup Chooses Where to Concentrate

Context

A seed-stage startup (9 people, $1.8M raised, 13 months of runway) sells an AI agent that turns sales calls into CRM updates. $21K MRR spread across three customer types: SMB sales teams, recruiting agencies, and management consultants. Each segment uses the product differently; the roadmap is a queue of per-segment requests. Growth is 4% a month and the founders disagree about where to push.

Applying the Skill

Step 1 — Name the pattern. The diagnosis workshop recognized a threshold shortfall: three motions, none past the visibility threshold. Nine people cannot hold three ICPs above threshold; the strategy question was not "how do we grow" but "where do we concentrate."

Step 2 — Anticipation scan. The team wrote the "announced futures" register. Decisive entry: the two dominant CRMs had both announced native call-summary features shipping within a year. Generic "calls → CRM notes" was a dying wedge — whatever segment they chose had to be one where native features would still lose.

Step 3 — Score the segments against sources of power. Each segment was scored 1-5 on: pull (inbound, usage depth), asymmetry (why incumbents/native features can't serve it), isolating-mechanism potential (what compounds), and threshold reachability within runway.

Criterion	SMB sales	Recruiting agencies	Consultants
Pull today	3	4	2
Asymmetry vs. native CRM features	1 — head-on	4 — ATS fragmentation, candidate + client double-entry	3 — but bespoke workflows
Isolating-mechanism potential	1	4 — placement-outcome data network	2
Threshold within 13 months	2 — broad market	4 — dense niche, 3 conferences, 2 communities	2
Total	7	16	9

Recruiting agencies won on the only criterion that mattered most: the CRM vendors' announced features (anticipation) would commoditize the sales use case first, while recruiting ran on fragmented ATSes the natives would not prioritize — a rival's-roadmap asymmetry plus a reachable, dense niche.

Step 4 — Set the proximate objective and the no-list. High ambiguity ruled out a revenue target. The proximate objective: in 8 weeks, 20 recruiting agencies live, with intake-call → ATS + client-update workflow complete end-to-end, and weekly usage retention above 60%. The no-list: no new SMB sales features, no consultant customizations, sunset both with 90-day migration help, decline non-recruiting inbound.

Step 5 — Create-destroy before commit. One founder spent two days building the case against recruiting (small TAM, ATS API risk, agency churn). The attack surfaced a real rabbit hole — two ATS APIs covered only 60% of target agencies — which became action item one (integration coverage) instead of a Q3 surprise. The bet survived its own destruction and shipped.

Outcome

Six months later: $58K MRR, 96% from recruiting; growth 11% monthly; the placement-outcome dataset (which intake answers predict successful placements) became the demo's closing slide — an isolating mechanism no horizontal note-taker had. The CRM-native features shipped as predicted and erased three of the startup's former competitors. One founder's note: "Choosing felt like killing two-thirds of the company. It was the first day the company existed."

Lessons

1. Anticipation — reading rivals' announced roadmaps — eliminated an option that pull alone would have kept alive.
2. The scoring table did not make the choice; it forced the disagreement into criteria, where it could be resolved.
3. Create-destroy converted the riskiest unknown into the first action item.

Case Study 3: Rewriting a Vision Deck into a Kernel

Context

A 200-person payments scale-up prepared for its Series C with an offsite that produced a 30-slide "Vision 2030" deck: mission, five values, five strategic pillars (Growth, Innovation, Customers, People, Excellence), and 23 initiatives distributed under the pillars. A board member's only comment: "This is lovely. What's the strategy?" The CEO brought the deck to a working group with this skill.

Applying the Skill

Step 1 — Honest classification. The group labeled every slide. Pillars were categories, not choices — each one passed the negation test into absurdity ("we will not pursue Excellence"). The 23 initiatives mapped 1:1 to org-chart departments: the deck was the org chart wearing a costume, the signature of unwillingness to choose.

Step 2 — Force candidate diagnoses. Instead of debating pillars, each of five leaders wrote a one-paragraph diagnosis of the single critical challenge. Three candidates emerged: (a) enterprise deals stall in security review; (b) unit economics depend on interchange fees that regulators in two core markets have signaled they will cap; (c) product breadth has outrun reliability.

Step 3 — Create-destroy with a virtual panel. Each candidate was attacked through three personas: a skeptical CFO ("show me the number"), a churned customer, and the rival's head of product ("which diagnosis would I hope they pick?"). The interchange-cap diagnosis (b) survived strongest: the CFO persona found 61% of gross margin exposed to announced regulatory intent — a wave, with a date range. Candidates (a) and (c) were real but downstream: both became easier with the margin problem named, failing the unlock test in reverse.

Step 4 — Write the kernel.

• Diagnosis: 61% of gross margin rides on interchange fees that two regulators have signaled they will cap within ~three years. Our "vision" assumes an economic engine that is scheduled to shrink; every initiative priced against it is built on sand.
• Guiding policy: Shift the revenue base from payment rails to the software workflow around money movement — reconciliation, spend controls, forecasting — where willingness-to-pay survives fee caps, concentrating on the 400 mid-market customers who already use two or more workflow features. Therefore we will not: expand to new payment geographies this cycle, compete on rail pricing, or fund initiatives that scale fee-dependent volume.
• Coherent actions: (1) Software-revenue line target with its own owner (CRO, immediate); (2) workflow suite packaged and priced standalone (product, two quarters); (3) the 400-account expansion motion staffed by reassigning the geo-expansion team (Q1); (4) initiative review: each of the 23 initiatives re-justified against the policy.
• Review trigger: any regulatory ruling, or software revenue share below 25% by year-end.

Step 5 — The cull. Against the policy, 14 of 23 initiatives were killed or parked, including two sacred ones (a consumer app pilot and a sponsorship). The remaining nine were re-sequenced so each fed the software-revenue shift. The five pillars were retired; the values slides moved to the employee handbook, where they belonged.

Outcome

Aspect	Vision deck	Kernel
Length	30 slides	1 page + appendix
Initiatives	23 unranked	9, sequenced, each policy-traced
Margin exposure named	No	First paragraph, quantified
Series C diligence	—	Kernel reused verbatim in the data room
One year later	—	Software revenue 14% → 33% of gross margin

Lessons

1. "What's the strategy?" is answered by a diagnosis, never by pillars — categories are where choices go to hide.
2. The virtual panel converted a political fight (whose diagnosis wins) into an analytical one (which diagnosis survives attack).
3. Retiring the vision deck cost nothing: mission and values survived intact in their proper home, and the strategy finally existed.

Key Takeaways

1. Classification beats argument. Counting diagnosis slides (usually zero) settles "is this a strategy?" faster than any debate.
2. The diagnosis is usually avoidable evidence, already in hand. Appendix churn tables, announced regulations, rivals' public roadmaps — bad strategy is the art of not looking.
3. Choice creates losers, and that is the point. Every case required killing real work with real sponsors; the relief came after, never before.
4. Proximate objectives make strategy executable under ambiguity. Eight-week, owner-named, done-testable targets moved teams that grand targets had stalled.
5. Destroy your strategy before the market does. Create-destroy and the virtual panel found the rabbit holes and weak diagnoses while they were still cheap.