AI-Wiki

Strategy Theory Using Analogy: Rationale, Tools and Examples

4 min read

Carroll & Sørensen — Strategy Theory Using Analogy

TL;DR

Glenn R. Carroll and Jesper B. Sørensen (both Stanford GSB) examine why analogical reasoning is widely used in strategy formulation despite logicians regarding it as a weak inductive form. Rather than reject analogies, they propose disciplined tools for building and evaluating strategy analogies — particularly to support the theory-based view (TBV) of strategy (Felin and Zenger 2009, 2017).

Worked example: an extended Tripadvisor → Glassdoor analogy illustrating how to decompose, evaluate, and refine an analogical argument.

Key claims

Why analogies are pervasive in strategy practice

Analogies do two distinct kinds of work (Gentner 1982):

RoleUse
Rhetorical / persuasiveVivid metaphor for stakeholders (“Glassdoor is Tripadvisor for jobs”)
Generative / problem-solvingSurface candidate causal mechanisms; develop firm-specific theory of value

Six reasons analogies dominate strategy discourse despite weak logical foundations:

  1. Efficient communication — a one-phrase analogy carries vast detail.
  2. Concrete and memorable — material analogies vivid; abstract theories aren’t.
  3. Predictive even when conceptual understanding is limited — used like sound-as-water-waves heuristics in teaching science.
  4. Generative problem-solving tool — best solutions often generalize from familiar puzzles (Bartha 2015).
  5. Intermediate-level abstraction — neither too specific nor too abstract; matches how busy executives reason.
  6. Success-story anchoring — the source firm worked, providing “existence proof” for the strategy.

The structure of a predictive analogy

Source A (Tripadvisor)Target B (Glassdoor)
a₁: experiential-good info (hotels)b₁: experiential-good info (jobs)
a₂: free + open accessb₂: free + open access
a₃: data self-input by usersb₃: data self-input by users
a₄: aggregate ratings + detailed reviewsb₄: aggregate ratings + detailed reviews
a₅: collects data → reports + trend predictionb₅: same
a₆: ad/referral revenueb₆: ad/referral revenue
a_c: market successb_c: ?? plausibly similar

Tools for building / evaluating strategy analogies (the practical contribution)

  1. Decompose the global analogy into component premises (atomic features).
  2. Add negative analogies (where source and target differ) — protects against confirmation bias.
  3. Distinguish horizontal vs. vertical relations (Hesse 1966):
    • Horizontal = surface similarity between source and target features.
    • Vertical = causal relationships within source that produced its outcome.
  4. Test the vertical relations — does the source’s success theory plausibly transfer? Example: Glassdoor differs from Tripadvisor in whether users register, in event frequency, in user-employer power dynamics — these matter for whether the success theory transfers.
  5. Apply to the theory-based view — analogies aid in theory discovery (especially early), in involving wider teams in theorizing, in identifying what’s unique about a firm’s value theory, and in communicating the theory.

Step-by-step process for building strategy analogies (§6)

  1. Start with the target (your firm/initiative) and what conclusion you want to support.
  2. Select a source company carefully — features should produce the same conclusion in the source as you want in the target.
  3. Identify positive premises (similarities).
  4. Identify negative premises (differences) — don’t let the analogy stack the deck.
  5. Articulate the source’s causal/vertical theory (not just feature mapping).
  6. Translate vertical theory to target — does it still hold given the differences?
  7. Add firm-specific premises if the analogy is incomplete.
  8. Develop multiple analogies for the same problem in parallel (different teams) — increases predictive ability (Green & Armstrong 2007).

Cited example (Patel & Ahmad 2023, leaked Google memo) of an analogy in AI strategy discourse:

“If GPT-4 is the Walmart you go to for apples, what happens when a fruit stand opens in the parking lot?”

(Open-source AI ↔ fruit stand; OpenAI/GPT-4 ↔ Walmart.) Used as a small worked example of how analogies are deployed in fast-moving tech debate.

Methodology notes

  • Theoretical/methodological paper — no original empirical data.
  • Builds on Gavetti & Rivkin’s work (2005; HBR “How strategists really think”) and on Sørensen & Carroll (2021a, b) on first-order deductive logic in strategy.
  • Connects to philosophy of science (Hesse 1966; Bartha 2015, 2019) and cognitive science (Holyoak & Thagard; Gentner; Hofstadter).
  • Conference origin: Theory-Driven Strategy Conference, Bocconi University, October 2023.

Quotes worth saving

“Strategists must analyze chains of cause and effect” (Gavetti & Rivkin 2005, p. 8) — paraphrased framing of the generative use of analogy.

“The use of analogical argument presupposes a stronger causal relation than mere co-occurrence, … but it does not presuppose that the actual causal relation is known.” (Hesse 1966)

“We aimed here to demonstrate how to build strong theory-based analogies. … Analogies can lead to new ideas and stimulate fresh insights, thereby offering a way to develop and advance theory about strategy. Analogies may also be a good way to ‘pressure test’ a theory.”

Graph View

Backlinks