Traditional reporting is blind to a startup’s progress: business-plan milestones and financial metrics presuppose a predictable business model (managing the predictable), while at an early stage revenue is near zero and uninformative. The principal asset being produced—validated knowledge—is absent from the balance sheet altogether. So accounting must be redesigned, not abolished: accountability remains; the unit changes.
There has already been a precedent for such a redesign: the BSC (Balanced Scorecard) rebuilt accounting “around strategy”—metrics are meaningful only when they are tied to objectives and interconnected (Svyaz’ SSP i KPI.ztk). Ries’s innovation accounting is the next step in the same move: accounting “around uncertainty,” where metrics are tied to hypotheses and their testing.
The negative pole is vanity metrics: numbers that rise without an increase in knowledge (downloads, total registrations). The mechanism of corruption is described by Goodhart’s law: a metric that becomes a target ceases to be reliable. Innovation accounting therefore measures the outcomes of experiments—confirmation or refutation of hypotheses—rather than activity: activity is easy to simulate; there is no point in simulating a refutation of one’s own hypothesis. Cf. the trap of “substituting the telos with a metric” in Телос.
There are two layers to blindness to progress-as-knowledge. The institutional layer—described above—is that reporting does not record knowledge. The psychological layer—deeper—is the habit of measuring progress in tangible units (features, releases, revenue), which impedes the very perception of learning as progress because knowledge is intangible. The layers are independent: reporting can be redesigned and still yield a team that does not “feel” that it has worked. It is this mechanism that makes the prediction of team resistance during the introduction of learning milestones come true—the familiar unit has been taken away, while the new one does not feel like work (see The theory of managing uncertainty is falsifiable at the process level, not by outcomes).
[ET] Personal level: the working maxim “prefer visible evidence of completion” is the same pull toward tangible units, merely put in one’s own service. The practical implication is not to fight the habit, but to feed it the right unit—to make knowledge tangible (a written hypothesis, a recorded experimental outcome, an entry in the learning ledger), so that the visible evidence of completion is an artifact of learning rather than an artifact of busyness.
Accountability is a language between the innovator and the person who holds them to account (“…and the people who hold them accountable”): the principle is addressed to sponsors as well. Without a shared accounting system, a sponsor measures what is familiar—and either kills viable experiments because they have zero revenue or funds a theater of activity.
A bridge from observability: the [ET] clarification in the note on the goal of a startup—that verifiable learning is an observable reduction in uncertainty—and the new accounting is precisely what institutionalizes that observability: principle 5 is principle 3 turned into reporting.
Assumes that A startup exists to learn how to build a sustainable business
The demand for new accounting rests on a teleology: only if progress is knowledge is traditional reporting incomplete. Reject the teleology, and principle 5 collapses into ordinary management accounting.