§13
Phaal/Farrukh/Probert · IfM Cambridge
Fast-start roadmapping workshops — S-Plan / T-Plan
When you need alignment fast, the process is the deliverable — roadmaps are "dirty mirrors": the exercise reflects back the hidden gaps in current strategy. Treat the problems it surfaces as the point, not a flaw. Run it as a workshop, not a doc; customise first — no single format fits.
S-Plan (strategic landscape): one big wall-chart to build the 'landscape' and spot 'landmarks', then small-group deep-dives (~1½ days) — the fuzzy front end. T-Plan (fast-start, ~1–2 days): four linked workshops — market → product → technology → charting — connected by linkage grids.
Guard the consensus: Motorola required a 'minority report' on every consensus roadmap; pair with a written-first pre-mortem. Staff three roles: owner/customer (content, budget, decisions) · facilitator (runs the process) · functional experts (own the layers). Success = confidence + consensus on the decisions each iteration, not forecast accuracy. HIGH.
§14
Christensen, Verlinden & Westerman, Ind. & Corp. Change 11(5), 2002
Overshoot-driven integrate-vs-modularize & value migration
The dynamic engine under Porter's static map. The cycle: performance improves faster than a tier can absorb, so an underserved tier becomes overserved (overshoot). Then the basis of competition shifts performance → speed/customization, architecture shifts interdependent → modular, and structure shifts vertically-integrated → horizontal specialists.
Decision 1 — integrate or modularize? Read satisfaction. Underserved → interdependent/integrated, own the stack. Overserved → modular, disintegrate. An interface can go modular only when all three hold: the customer can specify the attributes, metrics measure them, and the customer understands the interdependencies. Any one missing → keep it inside the firm.
Decision 2 — where will profit sit? Follow the not-good-enough. Attractive profit accrues to whatever stage is still performance-limiting; as a stage overshoots and modularizes, scale economics flatten and differentiation collapses — margin migrates to the subsystem that now gates system performance. Law of Conservation of Modularity: interfaces alternate interdependent↔modular, so the profit pool never stops moving.
The sharpened outsourcing rule & the disk-drive evidence
Replace "outsource everything non-core" with outsource only the overserved (modular, specifiable) interfaces — never the one still performance-limiting, because that's where profit is heading. (The PC assembler that outsourced both microprocessor and OS.)
Worked (paper figures): over 1988–94 non-integrated disk-drive assemblers returned ~11%/yr to shareholders while the interdependent head/disk makers Komag and Read-Rite returned ~38%/yr; computer systems makers held ~80% of industry profit in 1986 and ~20% by 1991 as PCs modularized and profit moved to Intel/Microsoft/Applied Materials. HIGH; blind spot: "underserved vs overserved" is a judgment call — misread it and both decisions invert.
§15
Keeley, Pikkel, Quinn & Walters, 2013 · Doblin
Ten Types of Innovation — beyond product features
Doblin's empirical taxonomy (~2,000 innovations). Not a sequence — no hierarchy; start anywhere. Ten types in three categories:
Configuration — Profit Model · Network · Structure · Process. Offering — Product Performance · Product System. Experience — Service · Channel · Brand · Customer Engagement.
The load-bearing finding: product/tech innovation is the easiest to copy, so a Product-Performance-only roadmap is the default trap. Top innovators integrate ~2× as many types as average ones (directional). Method: diagnose (map yourself + 2–3 competitors → find errors of omission) → pick where to play → construct a play (combine multiple types — each leg copyable, the combination not).
The ambition dial & the "don't max it" caution
Ambition dial: Core (few types; rarely wins a beachhead) · Adjacent (typically 3–4) · Transformational (5+, orchestrated with care). New-market entry needs adjacent-or-above.
⚠️ Don't max the dial: more types = more defensible AND more integration cost, larger teams, more ways to fail — add a type only when it earns its complexity. Overlaps: Product System ≈ §6 platform; Profit Model ≈ §Pricing; Structure ≈ Team Topologies; Customer Engagement ≈ HEART/Hook. HIGH on the taxonomy; MODERATE on the quantified premium.
§16
Wheelwright & Clark, HBR 1992 · Revolutionizing Product Development
Aggregate Project Plan — the project-type mix & capacity math
Use when you own a portfolio and the symptom is over-commitment or drift (the opening case: PreQuip's budget rose while completed projects fell — 30 projects against ~2–3× the capacity to deliver them). "No single project defines a company's future; the set does." Classify every project on product change × process change into five types:
| Type | What it is | Intensity |
|---|
| Derivative | cost-reduce/enhance existing; incremental | lowest — bounded, light mgmt |
|---|
| Platform | more change, no untried new tech; built for easy derivation | high — heavy up-front planning |
|---|
| Breakthrough | fundamentally new core + process → new category | highest — latitude for new processes |
|---|
| R&D | the know-how precursor; competes for the same engineers | separate funding & expectations |
|---|
| Alliance/partnership | any of the above with a partner | varies — still staff in-house to capture knowledge |
|---|
Two levers other tools miss: a capacity cushion (PreQuip ran 75 of 80 engineers, reserving slack — over-commitment produced the delays) and a multi-year sequencing strategy.
The PreQuip turnaround & the sequencing strategies
Worked: PreQuip (<20% of its projects were platforms) cut 30 projects → 11 (3 platform, 1 breakthrough, 3 derivative, 1 partnership, 3 R&D); the ~50% platform / 20% derivative / 10% breakthrough / 10% partnership figures are the capacity allocation across its 8 commercial projects (R&D funded separately though it competes for the same engineers) — and productivity ~tripled (case figures, not a universal prescription).
Sequencing — a senior call: steady stream (a new platform every other year + spaced derivatives) · secondary wave (ship a platform, move the team, refocus a derivative wave) · compressed wave (Kodak: platform → immediate derivative burst → next-gen). Mix follows maturity: mature → weight platforms; growth → weight breakthroughs; revisit every 6–12 months. Platform-refresh cadence is a competitive benchmark — late-1980s autos: EU re-platformed ~every 12yr, US ~8, Japan ~4 (illustrative). HIGH.
§17
Carr, "IT Doesn't Matter," HBR 2003
Infrastructural-tech commoditization — offense → defense
A technology's strategic value tracks scarcity, not ubiquity. Proprietary tech can be owned and, while protected, is durable advantage. Infrastructural tech is worth far more shared than used in isolation, so broad sharing is inevitable and it ends a commodity. The trap: in early buildout an infrastructural tech masquerades as proprietary, so early movers win — and assume the window stays open. It doesn't.
Three-phase lifecycle: proprietary (advantage to forward-lookers) → buildout (capital floods in, prices down, standards universal) → commodity (the only edge left is cost). The pivot — offense to defense once a resource is "essential to competition but inconsequential to strategy": (1) spend less (Alinean 2002: the 25 highest-return of 7,500 firms spent ~0.8% of revenue on IT vs 3.7% typical) · (2) follow, don't lead · (3) focus on vulnerabilities, not opportunities.
The buildout sprints & the AI trap
Buildout sprints (sourced): world rail trackage 17,424 → 309,641 km (1846–76); US central generating stations 468 → 4,364 (1889–1917); processing power $480/MIPS (1978) → $50 (1985) → $4 (1995).
⚠️ The AI trap: the railroad/electricity advantage came largely in the second half of buildout, as firms re-invented practices — the differentiation lives in the business-practice innovation the tech enables, not the technology itself (McKinsey: an IT-productivity link in only 6 of 59 industries). Apply "follow, don't lead" to the commoditized layer only — for a still-generative tech (frontier AI) the proprietary window is the practice/insight you build on top, not the model you rent. HIGH.
§18
Feitzinger & Lee, "The Power of Postponement," HBR 1997
Postponement / delayed differentiation — where & when to commit a unit
Modularity decides what to share; postponement decides where & when to commit a unit to its final variant. The rule: push the differentiating step to the latest cost-effective point — carry generic stock, customize on order → variety, speed AND lower cost together. Three coupled pillars (all three, or it leaks):
1Modular product
common parts early, differentiators last. HP LaserJet universal power supply → total cost −5%/yr
2Modular process
postpone (paint mixed at counter) · resequence (Benetton dyed after knitting) · standardize (apparel cut-and-sew within 48 hours)
3Agile supply network
HP DeskJet moved customization Singapore factory → Stuttgart DC → total cost −25%, generic inventory −50%
The standardization tradeoff & the software translation
The tradeoff: a common component costs more in materials; it pays only when the inventory/transport savings beat the premium — and those savings rise with demand uncertainty, lead time, short life cycle, stock-out cost. Build-to-order vs build-to-stock falls out of this: modular product + modular process + on-order customization ⇒ BTO feasible and SKU count collapses; no postponement ⇒ locked into build-to-stock and eat the write-offs.
Software: the same lever is late binding / config-on-deploy — ship one generic build, differentiate at the latest point (feature flags, tenant/runtime config, edge personalization) instead of forking a build per customer. HIGH.
§19
Fricke & Schulz 2005 · Suh, de Weck & Chang 2007
Design for changeability & flexible platforms — price it as a real option
Platforming cost-optimizes predetermined variants; design-for-changeability architects for foreseen AND unforeseen change. The old default — front-load to prevent change because each later phase is ~10× costlier ("Rule of Ten") — fails in fast markets; build the right amount of change-ability in on purpose.
Name the aspect (Fricke & Schulz) — separated by who acts and how fast: Robustness (delivers unchanged — nobody acts) · Adaptability (self-adjusts) · Flexibility (changed easily by an external actor) · Agility (changed rapidly). Nine enabling principles (3 basic serve all four; 6 extending serve some) conflict — independence ✗ redundancy (Ariane 501: identical redundant controllers failed the same way at once) — so surface the harmful pairs up front.
Size the window — neither max nor min is optimal. Upfront cost-of-changeability trades against lifecycle cost-of-changes; total is U-shaped, minimized in a "window of opportunity" set by market velocity and technology half-life. Skip changeability for short-life expedient systems, precedented systems in slow markets, or ultra-high-performance markets with no performance-loss allowable.
Value flexibility as a real option (CPI) & the body-in-white case
When the driver is uncertainty, map uncertainty → attributes → design variables → components, use the Change Propagation Index (CPI = changes-sent − changes-received) to find the "multiplier" components (CPI > 0 — their change cascades), embed flexibility there, and value it by expected NPV under Monte-Carlo scenarios.
Worked (automotive body-in-white, paper figures): 10 of 21 components made flexible cost ~34% more upfront but cut switching cost 31.9 → 5.4 and payback 2.5 → 0.5 yrs for a representative future change. Under near-zero uncertainty the rigid design wins — don't pre-abstract. Software: the multiplier is the high-fan-out module (a core schema, a shared contract, an identity/payments layer); paying upfront for a versioned interface there is the real option — justified only when expected avoided rework beats the cost. HIGH; blind spot: flexibility only insures against the uncertainties you named.
The multiplier / high-fan-out module ↔ Design-for-Variety CI-S · pro §metrics §10 renewal timing