F · Unit economicsExternally proven

Discovery-Driven Planning / Reverse Income Statement

A method by Rita Gunther McGrath and Ian C. MacMillan for ventures with many unknowns. Instead of planning forward, you compute backward: the required result comes first, from it the revenue needed and the cost allowed. Every unproven number goes onto an assumption list, and the budget is released only up to the next checkpoint.

When you need this method

You plan a new product line, a new market, or an AI initiative using the tools of the running business. The plan looks solid because it is written in numbers, but most of those numbers are assumptions nobody has marked as such. The budget is released in one block, and the first honest reconciliation with reality arrives once the money is largely spent.

Approach

1Define success first and compute backward: which result justifies the venture, which revenue that requires, and which total cost is therefore allowed.
2Break the arithmetic down to unit level, price per unit and units required per period, and hold that volume against market comparisons instead of simply asserting it.
3Specify the operation: which activities, volumes and people are needed to deliver those units, and which costs follow from that.
4Put every number that is not evidenced onto an assumption list, with a number, an owner, a planned test and a date.
5Plan checkpoints: for each milestone, fix which assumption is tested there, at what cost, and what result means continue, redirect or stop.
6Release funding only up to the next checkpoint, then redo the arithmetic with the numbers that have been tested.
7Mark refuted assumptions on the list as refuted rather than quietly replacing them with new ones.

Typical application

A typical B2B SaaS in the HR field wants to build an add-on module with AI features and sets a revenue target for year three. The reverse arithmetic inverts the order: from the required contribution and the cost it allows, a realistic price per seat implies a number of new customers per month that exceeds what the sales team has ever closed in a full quarter. The team does not drop the venture. It writes down the three numbers everything hangs on, namely price acceptance, time to first productive use, and effort per implementation. Each of them gets a cheap test before the first line of production code. After the second checkpoint, price acceptance holds, but implementation effort per customer turns out to be roughly twice what was assumed. The venture continues with a narrower scope and with a budget that reaches only to the next checkpoint.

Limits and counter-indications

The method presumes that a required result can be named at all. Where market and goal do not yet exist, the prediction logic behind it stops working; Sarasvathy shows for exactly that case that reasoning from available means can beat reasoning from a preset goal. Without solid reference values for price, volume and effort, the backward calculation also produces nothing but a more precise-looking invention. Milestone funding needs an organization that can tolerate stopping; where budgets are granted annually and status depends on project size, checkpoints turn into ritual. For the running core business with a reliable history, ordinary forward planning is the better tool. And the effect itself is unmeasured: the sources carry the procedure and its rationale, not a controlled demonstration of impact.

How to measure impact

Track two numbers per venture: the share of originally listed assumptions actually tested by the respective checkpoint, and the cost per refuted assumption. At every checkpoint, record the recalculated allowable cost line alongside them.

Related methods

Tools for this

Lovable

Sources

1.McGrath, MacMillan: Discovery-Driven Planning, Harvard Business Review 73(4), 1995, S. 44-54 (opens in a new tab) · Harvard Business Review · 1995-07 · practitioner source · supports the underlying mechanismEstablishes authorship, year of publication, and the occasion for the method: large companies lost substantial sums in new fields because they planned ventures full of unknowns with the control tools of the running business. Limit: the full text is behind a paywall. Freely verifiable are title, authors, issue, and the framing of the problem including three named failures; the claim that better planning tools could have contained those losses is the authors' thesis, not a measurement.
2.Discovery-driven Growth: The Only Plan Is to Learn as You Go (Interview mit McGrath und MacMillan) (opens in a new tab) · Knowledge at Wharton, University of Pennsylvania · 2009-04-29 · practitioner source · describes the methodCarries the procedure as the originators themselves present it: set a target, break the required scope down into volumes, write assumptions down and test them at deliberate checkpoints, and invest incrementally so that failure stays affordable. The reverse income statement is explicitly described as the order from required profit through allowable cost to necessary revenue. Limit: this is a book-launch interview, not a reviewed study, and the application example given is a single case.
3.McGrath: Who Learns Fastest, Wins. Lean Startup and Discovery Driven Growth, Journal of Management 50(8), S. 3162-3182 (opens in a new tab) · Journal of Management (SAGE), geprüft über den frei zugänglichen DOI-Nachweis · 2023 · academic and scholarly literature · provides the contextPlaces the method in its field: it consists of five interacting design steps, among them defining success, checking for realism, and documenting assumptions, and it grew out of research on innovation inside established companies, whereas lean startup comes from entrepreneurship research and aims more narrowly at product-market fit. Limit: what was verified is the freely available abstract; the publisher's full text is paywalled. The paper compares methods conceptually and measures no effect.
4.Sarasvathy: Causation and Effectuation, Academy of Management Review 26(2), 2001, S. 243-263 (opens in a new tab) · Academy of Management Review, frei zugängliche Volltextfassung · 2001 · academic and scholarly literature · limits the methodMarks the boundary of the method. The paper distinguishes a goal-driven logic that selects among means from a means-driven logic that selects among ends. To the extent the future is unpredictable, goal-driven reasoning does not hold; it fits where markets exist and goals can be named. Discovery-driven planning, with its backward computation toward a required result, belongs to the first group. Limit: the paper is conceptual and does not test discovery-driven planning itself.

Origin: McGrath / MacMillan

← Back to the catalog