Discovery-Driven Planning¶

A high-uncertainty venture plan works backward from required success, lists its make-or-break assumptions, and stages commitment around tests that turn assumptions into knowledge.

Version: v2 · 2026-10-03 · History
Domain-specific #: 13152
Domain group: Professional & Organizational Practice
Origin domain: Organizational & Management Science
Subdomains: Innovation Management, Strategic Planning → Organizational & Management Science
Aliases: DDP

Core Idea¶

Discovery-driven planning (DDP), introduced by Rita Gunther McGrath and Ian C. MacMillan in 1995, is a way to plan a venture when too much of a conventional forecast would be assumption rather than established knowledge.^[1]^[2] It starts by defining the result that would make the venture worthwhile, works backward to the economics and operating conditions that result requires, names the uncertain assumptions, and sequences tests at checkpoints before larger resource commitments. McGrath later summarizes five linked practices: define success and plan backward, benchmark externally, specify operations, document assumptions, and reassess assumptions and goals at checkpoints.^[3]

The method is still planning. It does not say “act without a model” or “experiment forever.” It makes the model's uncertainty visible so the next commitment can depend on what was learned. An authorial technical note contrasts the conventional goal of delivering close to forecast numbers with DDP's goal of useful learning at limited expenditure in highly uncertain settings.^[4] Learning is instrumental: evidence should lead to continuation, revision or exit decisions, not merely another report.

Structural Signature¶

Success condition: an explicit required outcome gives the plan a backward starting point.
Reverse economics: required revenue, margins, volume or operating capability follow from the target.
External and operational checks: benchmarks and capability specifications test plausibility.
Assumption register: make-or-break unknowns are stated separately from facts.
Learning milestones: checkpoints seek evidence before escalating commitment.
Revision/exit rule: results can change the venture's design or stop it.

Sig role-phrases: required success result; backward economic and operating requirements; external benchmarks; explicit unknowns; staged learning milestones; conditional continuation, revision or exit.

What It Is Not¶

DDP is not a conventional forecast with optimistic numbers or a fixed schedule whose milestones only mark elapsed time. A reverse income statement alone is insufficient if its implicit demand, cost and capability assumptions are never tested. It is not identical to scenario planning, which compares possible futures, or to generic lean experimentation with no stated threshold of worthwhile venture success. Nor does the 1995 article's list of costly failed ventures prove those ventures actually used or rejected DDP; the examples motivate the authors' problem, not a measured causal evaluation of their method.^[1]

Scope of Application¶

The method was developed for new ventures and strategic moves into unfamiliar territory, where a firm lacks a reliable basis for extrapolating past sales or costs. The authors' original article uses Kao Corporation's entry into floppy disks as a worked example, according to the Columbia Business School publication abstract.^[2] McGrath's own later description says the same discipline can apply where an established business faces unusually high uncertainty, not only to start-ups.^[3] The source-accessible material does not expose Kao's detailed numeric tables, so this draft does not manufacture them or claim the company executed each step exactly as a live operating policy.

Clarity¶

Separate requirements, assumptions and Evidence. “We need one million dollars in revenue” can be a requirement derived from a target and margin; “customers will buy at our planned price” is an assumption; observed paid conversion in a small pilot is evidence relevant to it. A milestone is discovery-driven only if passing or failing it can change a resource decision. A calendar date for launch with no test is schedule management, not the method's learning gate. External benchmarks and an operating specification ask whether the required numbers fit the market and the firm's capabilities rather than treating a spreadsheet as self-validating.

Manages Complexity¶

DDP breaks a high-uncertainty venture into claims that can be investigated in an order. Each projected number can be traced to an assumption, each assumption to a possible test, and each test to a commitment decision. This can contain losses from weak premises, but testing is not free. Small pilots consume time and money, and over-testing may postpone a valuable entry. The point is not to minimize all spending; it is to spend enough to learn about the decisive unknowns before committing far more.^[4]

Abstract Reasoning¶

Consider a constructed subscription venture aiming at a $100{,}000$ annual surplus, with $300{,}000$ fixed cost and a 40% contribution margin. Its reverse income requirement is $(300{,}000+100{,}000)/0.40=\$1{,}000{,}000$ annual revenue. That result is arithmetic, not a prediction. It creates questions: how many paying customers at what price, what retention, what service capacity, and what acquisition cost would make $1 million feasible? The planner can test paid conversion on a small segment, compare external benchmarks, and pause before a broad rollout if the implied volume is implausible. The exact numbers are illustrative, not from McGrath and MacMillan's Kao case.

Knowledge Transfer¶

The principle transfers from start-ups to any initiative with a high assumptions-to-knowledge ratio, as McGrath explicitly notes.^[3] The conserved structure is prospective commitment made conditional on learning about required success. But the named method is management practice: financial thresholds, operational capability, customers, budgets and decision rights matter. A software team's test plan or a scientist's experiment may resemble part of it without becoming DDP unless it also links learning to a venture's success requirements and staged investment.

Examples¶

Kao's floppy-disk entry as an author-used worked setting. Columbia Business School's record of the 1995 article says McGrath and MacMillan use Kao Corporation's entry into floppy disks to present a step-by-step approach to planning under uncertainty.^[2] Mapped back: Kao's unfamiliar product/market entry is the venture context; the authors use it to expose potentially dangerous implicit assumptions and show a learning-oriented plan. The accessible abstract does not provide the case's required profit, manufacturing benchmarks, assumption list or observed milestone results. This is source-attested as a worked teaching case, not independently verified as a real-time implementation of every DDP step.

A constructed reverse-income gate. In the subscription venture above, target surplus $100{,}000$ plus fixed cost $300{,}000$, divided by a 0.40 contribution fraction, sets a $1 million annual revenue hurdle. At $50 annual revenue per paying customer-equivalent, that implies 20,000 customers; it is a derived requirement, not evidence that demand exists. A low-cost paid pilot can test conversion and retention before funding a full sales team. Mapped back: the target defines success, arithmetic exposes required volume, buyer uptake and retention are assumptions, and the pilot is a milestone whose result changes the go/revise/stop choice. This numerical case is our constructed diagnostic, not a reported company outcome.^[3]^[4]

Structural Tensions¶

The method manages an actual speed/commitment versus learning/loss tradeoff. Spending heavily and launching immediately may gain time in a competitive market, but exposes the organization to untested make-or-break assumptions. Staging pilots costs time and smaller resources and can delay entry, while reducing the risk of scaling a false premise. The right balance depends on cost of delay and reversibility of the next spend. Diagnostic: which uncertain assumption would reverse the next investment decision, and what is the cheapest timely test that would make the decision materially better? A milestone that cannot change action provides little discovery value.

Structural–Framed Character¶

DDP is mixed but strongly practice-bound. Its structure—backward requirement, explicit unknowns, evidence gates and conditional commitment—is portable across ventures, yet what counts as “worthwhile” success is a managerial value judgment and institutional budget decision. McGrath and MacMillan's business-school/HBR origin explains the vocabulary of income statements, benchmarking and milestones. The method travels from start-up settings to uncertain established businesses, but recognition requires more than using the word “learning”: a plan must expose assumptions and let checkpoint evidence revise commitment.^[3] Importing the label onto a fixed forecast that merely records deviations is not the same relation. Its character: a disciplined high-uncertainty planning practice whose value lies in making prospective investment conditional on evidence.

Structural Core vs. Domain Accent¶

The skeleton is a goal-to-action plan revised through tests of its uncertain premises before escalating commitment. The domain-bound mechanism is managerial: required financial/operational outcomes, external market benchmarks, resource gates and authority to continue or stop. Kao floppy disks and the constructed subscription model are different carriers, not required ingredients. This named method fails the prime bar because the broad prospective structure already belongs to the live Planning prime; DDP adds a specific business method for high assumption-to-knowledge situations, not a new substrate-independent skeleton. The strict Planning subtype edge is approved in this staged bundle only; scenario planning is related but not a necessary genus.

This entry is a kind of Planning.

DDP builds a revisable, forward-looking action sequence to meet a stated future objective, with dependencies and evidence-based revision points, which is what Planning requires. Placing it under Planning says nothing about whether it improves venture outcomes.

Scenario Planning and business-systems planning are related approaches, but neither is a broader kind that DDP falls under.

Relationships to Other Abstractions¶

Current abstraction Discovery-Driven Planning Domain-specific

Parents (1) — more general patterns this builds on

Discovery-Driven Planning is a kind of Planning Prime

Discovery-driven planning is a venture-specific prospective goal-to-action plan with assumption tests and conditional commitments.

Hierarchy path (1) — routes to 1 parentless root

Discovery-Driven Planning → Planning → Foresight

Neighborhood in Abstraction Space¶

Discovery-Driven Planning sits in a sparse region of the domain-specific corpus (88^th percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.

Family — Unclustered & Miscellaneous (2551 abstractions)

Nearest neighbors

Computed from structural-signature embeddings · 2026-10-08

Not to Be Confused With¶

Conventional extrapolative budgeting: treats estimates as forecasts to meet rather than assumptions to test.
Backward arithmetic alone: necessary requirements without tests and checkpoints are not the full method.
An ordinary deadline: a milestone must change a decision based on evidence.
Aimless experimentation: DDP begins with a defined worthwhile result and tests toward it.
A guarantee of venture success: testing reduces some avoidable uncertainty but cannot eliminate competition or all risk.

References¶

[1] Rita Gunther McGrath and Ian C. MacMillan, “Discovery-Driven Planning,” Harvard Business Review 73 (July–August 1995), original article landing page and accessible opening; full body inaccessible. https://hbr.org/1995/07/discovery-driven-planning registry ↩a ↩b

[2] Rita Gunther McGrath and Ian C. MacMillan, “Discovery-Driven Planning”, Harvard Business Review 73 (1995), Columbia Business School faculty abstract of the same article, including the Kao case. The faculty page's single-author and January 1 metadata conflict with the publisher's two-author July–August record; the publisher metadata governs the work identity. registry ↩a ↩b ↩c

[3] Rita Gunther McGrath, author presentation material, “Discovery Driven Planning,” p. 7, five interdependent practices and high-uncertainty scope. https://ritamcgrath.com/pdf/ritamcgrath_topics.pdf registry ↩a ↩b ↩c ↩d ↩e

[4] Rita Gunther McGrath, Ian C. MacMillan and Robert Cooper, “Technical Note: Putting Discovery-Driven Planning to Work,” 2008, authorial product description. https://store.hbr.org/product/technical-note-putting-discovery-driven-planning-to-work/KEL355 registry ↩a ↩b ↩c