Westerveld et al. — The Business Model Portfolio as a Strategic Tool
Citation
Westerveld et al. (2023). The Business Model Portfolio as a Strategic Tool for Value Creation: Opportunities and Paradoxes. Journal of Strategy and Management. Developmental literature review (Templier & Paré 2015) + Gioia methodology coding over 38 empirical case studies, databases screened 2010–2021.
Ingested in the wiki as a recap document, not the full paper — the source file in raw/ is a structured 63-line summary of central thesis, evidence, critical evaluation, and methodological assessment.
Central claim
Operating multiple business models — a Business Model Portfolio (BMP) — has evolved from a defensive contingency strategy into a proactive strategic tool for value creation. Modern firms cannot rely on a single monolithic BM; they must harmonise diverse, often interdependent BMs to leverage demand-side and supply-side synergies. This strategic pluralism is inherently paradoxical — value opportunities and organisational tensions are co-produced and cannot be solved, only managed.
The paper’s principal contribution to this project is operational: it provides the prerequisite scoping rule — the BM, not the firm, is the unit of strategic analysis — that makes the weber-ai-startup-business-models classifier work for multi-BM AI startups. See business-model-portfolio for the concept page that grounds this rule.
Key arguments
1. The three strategic intents (chronological)
- Diversifying — autonomous BMs targeting distinct market segments (legacy airline + low-cost carrier). Historically dominant.
- Sensing — BMs as experimental probes; failed probes are discarded without disrupting the core. Aligns with the “thousand flowers bloom” phase in wef-ai-in-action-2025 and the early stages of enterprise-adoption-ladder.
- Complementing — tightly integrated BMs where one reinforces another via shared data and functional architecture (Amazon Prime ↔ retail ↔ streaming; LAN Airlines passenger ↔ cargo). The most advanced contemporary intent.
Synthesis observation in the recap: the Complementing intent strongly aligns with Huang’s “extreme co-design” in karpathy-software-3, where a computing platform’s survival is dictated by its integrated install base across applications.
2. Digital technology as connective tissue
Digitalisation is the primary catalyst enabling modern BMPs. Customer data captured in one BM can power the value proposition of another; shared functional architecture lets supply-side synergies cross BM boundaries. The recap cites GoMore (car-sharing data matching car-leasing supply) and omnichannel retailers (online data informs offline value proposition).
The recap explicitly extends this to AI: the Deloitte AI Dossier’s “multi-agent AI ecosystems” (deloitte-ai-dossier-eri) are the 2026 mechanism through which the functional-architecture-as-bridge claim is currently being realised.
3. The four BMP paradoxes (permanent, not solvable)
- Horizon paradox (exploit vs explore) — anchored in March (1991) ambidexterity / exploitation-exploration framework.
- Cannibalization paradox — new BM captures revenue from the incumbent. In 2026 AI: Steinberger’s GUI-vs-agent-API thesis in karpathy-software-3 is the canonical instance.
- Digital paradox — physical vs digital asset tension; omnichannel complexity vs customer’s desire for seamlessness.
- Dogmatic paradox — organisational confirmation bias toward the incumbent. Echoed by Sutskever’s “human reward hacking” in eval-real-world-gap: researchers (and managers) over-optimise current recipes rather than seek contradictory evidence.
4. Theoretical scaffolding
- Business Model concept (Massa et al. 2017; Teece 2010) — Value Proposition / Value Architecture / Functional Architecture / Financial Sustainability — used as the structural taxonomy.
- Paradox Theory (Smith & Lewis) — tensions as “contradictory yet interrelated elements” requiring dynamic equilibrium, not resolution.
- Ambidexterity / Exploitation-Exploration — underpins the Horizon paradox.
Method
- Developmental literature review (Templier & Paré 2015) + Gioia methodology (1st-order / 2nd-order coding, abductive transition).
- Databases screened 2010–2021.
- Final sample: 38 highly relevant empirical case studies.
Critical evaluation
- Retrospective constraint — the 38 cases were not originally designed to investigate BMPs or paradoxes; opportunities and tensions are extracted through post-hoc interpretive coding. Risk of uneven focus.
- Industry and geographic bounding — sample concentrated in airlines, media, retail; lacks deep coverage of heavy / asset-intensive sectors (deep-tech manufacturing, nuclear energy), so hardware / regulatory paradox classes may be missing.
- Pre-LLM cutoff — the literature window closes in 2021. The recap’s AI-era extensions (Karpathy, Sutskever, Steinberger, Huang, WEF, Deloitte) are interpretive bridges added by the recap author, not findings of the original paper.
Hypothesis touches
- H1_L0-L7-ladder — neutral. The BMP frame operates orthogonally to the ladder; a startup can sell BMs at different ladder positions.
- H2_u-curve-of-value — modest reinforcement. A single firm can run one BM at the top of the U (a distribution-moat-backed product) and another at the shoulder (services, analytics) — explains why empirical firm-count data is messier than the value-capture claim.
- H3_orthogonal-axes-under-priced — direct reinforcement. The BMP frame is the structural reason a single firm occupies multiple positions on every analytical axis, which is exactly the situation H3 names as under-priced.
- H4_rl-specialization-value-pocket — neutral. H4 is a per-BM claim; the BMP frame clarifies that an H4 BM can coexist with non-RL BMs in the same firm.
Open questions surfaced
- Does the Complementing intent dominate the Diversifying / Sensing intents in 2026 AI specifically, or is the AI-startup mix still Sensing-heavy (per wef-ai-in-action-2025 “thousand flowers”)?
- Of the four paradoxes, which is most salient for AI startups specifically? The recap argues Cannibalization (Steinberger thesis) and Dogmatic (Sutskever thesis); evidence in the wiki is thin.
- Does the BMP frame transfer cleanly to small / early-stage firms, or is it primarily an incumbent / scale-up phenomenon? Westerveld’s 38 cases are biased toward incumbents.
See also
- business-model-portfolio — the concept page grounded in this source; carries the unit-of-analysis claim into the wiki.
- weber-ai-startup-business-models — the classifier this source is prerequisite to.
- ai-startup-business-archetypes-weber — the operational classifier; applied per-BM in light of the BMP frame.
- weber-taxonomy-2026-gaps — the diagnostic page that now opens with the BMP prerequisite framing.
- H2_u-curve-of-value · H3_orthogonal-axes-under-priced — the hypotheses most affected by the BMP frame.
- karpathy-software-3 — Huang’s extreme co-design ↔ Complementing intent; Steinberger ↔ Cannibalization paradox.
- eval-real-world-gap · sutskever-age-of-research — Sutskever’s “human reward hacking” ↔ Dogmatic paradox.
- wef-ai-in-action-2025 · enterprise-adoption-ladder — thousand-flowers-to-enterprise-reinvention arc ↔ Sensing-to-Complementing arc.
- deloitte-ai-dossier-eri — multi-agent ecosystems as the 2026 enactment of functional-architecture-as-connective-tissue.
- digital-core — buyer-side mirror of the connective-tissue claim.