Innovation Accounting: Funding What Doesn't Fit a P&L

Every structure I’ve written about in this series has the same cause of death listed on the certificate. The separate exploratory unit, the lab-as-school, the Kickbox program and the internal venture — build any of them well and they will still die at budget season, the moment someone measures them with the instruments built for the core business. A nine-month-old venture asked for its ROI has exactly one honest answer — there isn’t one yet — and in an annual budget process, that answer loses to any core project with a confident spreadsheet attached.

This isn’t a story about villains. The finance team applying NPV to your venture portfolio is doing precisely what they were hired to do. The problem is that the corporate immune system I described in part one doesn’t only manifest as middle managers protecting turf. Its most durable form is measurement: ROI, NPV, and the annual budget cycle are the immune system wearing a spreadsheet, and they kill more new bets than any hostile executive ever will — politely, procedurally, and with everyone agreeing the math checked out.

So this closing piece is about the machinery on the other side: how you measure and fund work whose defining characteristic is that you don’t yet know if it’s worth anything.

Why the core’s instruments kill what they measure

Start with the sharpest version of the charge, because it comes from inside the house. Clayton Christensen, Stephen Kaufman, and Willy Shih published “Innovation Killers” in HBR in 2008, and its subtitle says everything: how financial tools destroy your capacity to do new things. Their argument isn’t that discounted cash flow is wrong math — it’s that it’s routinely misapplied. The standard DCF comparison weighs the investment against a do-nothing baseline that assumes the present continues indefinitely. But the present doesn’t continue; in a market where someone will eventually build the thing, the honest baseline is decline. Measured against a falsely flat status quo, almost any exploratory investment looks like a bad deal, and the company rationally, repeatedly, chooses erosion.

Layer the annual budget on top and the trap closes. An annual allocation forces a venture to predict twelve months of spend against a plan that will be invalidated by the first three customer conversations — and then holds the team accountable to the fiction. Worse, an annual grant is an entitlement: once awarded, it flows regardless of what the team learns, which means there is no natural moment at which anyone must decide whether the bet still deserves the money. The projects that should die don’t, the projects that should accelerate can’t, and the whole portfolio moves at the speed of the fiscal calendar rather than the speed of the evidence.

NPV, ROI, and annual budgets are fine instruments for the business you have. They are actively destructive instruments for the business you’re looking for. Different phase of knowledge, different instruments — that’s the whole discipline, and everything below is a version of it.

Innovation accounting: progress measured in learning

Eric Ries coined innovation accounting in The Lean Startup (2011) and built it out in The Startup Way (2017), and the core move is a redefinition: for a venture operating under uncertainty, the unit of progress isn’t revenue — it’s validated learning, a demonstrated change in what you know about customers, backed by data from real behavior. A team that spent a quarter discovering that its target segment won’t pay has made more progress than a team that shipped four features into a void. The accounting problem is that conventional instruments score it the other way around.

Ries structures the alternative in three levels, and the laddering matters more than the labels. Level one is a dashboard of metrics the team itself believes in — activation, retention, revenue per customer, whatever maps to the venture’s actual engine of growth — tracked per cohort, so you can see whether the people who arrived this month behave better than the people who arrived last month. Level two turns those metrics into learning milestones: for each leap-of-faith assumption, what did the dashboard show before the experiment, what does it show after, and did the needle move? Level three is the translation layer for the CFO — as evidence accumulates, you re-run the venture’s long-term model with measured inputs replacing guessed ones, and watch the net present value estimate converge from fantasy toward something a finance team can respect. The venture earns its way onto the corporate instruments; it doesn’t start on them.

The famous distinction underneath all this — vanity metrics versus actionable metrics — is the part most companies quote and least apply. Cumulative registered users, total downloads, gross page views: these go up and to the right almost regardless of what you do, which is precisely why dashboards love them and why they teach you nothing. An actionable metric ties a specific, repeatable action to an observed result — this cohort, exposed to this change, retained at this rate. My own test is blunt: if the number can’t go down, it’s not a metric, it’s a decoration. In every corporate innovation review I’ve sat through, the deck’s proudest slide was a decoration.

Metered funding and the growth board

Measurement only matters if money responds to it, which brings us to the funding mechanism: fund like a VC, not like a budget office. A venture fund doesn’t wire a startup five years of runway on day one. It releases capital in tranches, each round sized to answer the next set of questions, each release conditional on evidence from the last. Metered funding imports that logic inside the firm: small initial allocations — small enough to be non-events, which is what makes saying yes cheap and killing painless — with follow-on money released by demonstrated learning, not by the calendar. Under entitlement funding, teams fight to protect budget. Under metered funding, teams fight to produce evidence, because evidence is the only thing that buys the next tranche. The incentive flip is the entire point.

The governance piece has a name and a track record: the growth board, which Ries developed with GE during the FastWorks program in the mid-2010s and documented in The Startup Way. A growth board is a small standing group of senior executives that meets on a regular cadence — monthly or quarterly — to review a portfolio of internal ventures the way a VC partnership reviews its companies, and to make one of three calls per venture: persevere (release the next tranche), pivot (the strategy changed, re-fund against the new plan), or kill (retire the assumptions, recycle the team, celebrate the memo). Between meetings, the board’s other job is protection: the ventures answer to the board, not to whichever functions their staff came from.

Understand what this replaces, because it connects straight back to part one. In the default corporation, a new idea needs a yes from every layer it passes through — any single manager can veto it by declining to sponsor, staff, or fund it, usually by saying nothing at all. The growth board inverts the topology: one explicit forum where the decision actually lives, senior enough to make its yes stick, on a cadence fast enough to matter. The middle-management pocket veto — the immune system’s favorite weapon — doesn’t survive the existence of a room where the question must be answered out loud, on the record, on a date.

Discovery-driven planning: budgeting for what you don’t know

The oldest and, I’d argue, still the best tool in this kit predates the lean-startup vocabulary by fifteen years. Rita McGrath and Ian MacMillan published “Discovery-Driven Planning” in HBR in 1995, and Steve Blank has been candid that it’s an intellectual ancestor of the Lean Startup. Their diagnosis: conventional planning is a fine technology for extending a known business, because its central pretense — that you can predict outcomes — is roughly true there. Applied to a new venture, the same pretense is lethal: the plan’s assumptions quietly harden into facts, the venture gets managed to the plan rather than to reality, and by the time the gap is undeniable, the sunk costs have their own gravity. Their exhibit was Disney’s launch of Euro Disney — a meticulously planned venture whose plan embedded unexamined assumptions (about how long European visitors would stay, and what they’d spend inside the park) that turned out to be imported from the wrong continent.

Discovery-driven planning inverts the machinery, and the inversion is delightfully concrete. Start with the reverse income statement: instead of projecting revenues downward and hoping profit emerges, you begin from the profit the venture must eventually deliver to be worth anyone’s time, then work backward — what revenues does that require, what unit economics, what volumes, what price points? Every number you write down goes onto an explicit assumptions checklist, each one owned, each one testable. Then you plan not by phases of execution but by milestones that test assumptions — and at each milestone, you formally re-plan against what you just learned, revising the reverse income statement with measured numbers. Funding is released milestone by milestone, sized to the cost of the next test.

Notice how neatly this answers the CFO’s legitimate objection to all this learning talk. Discovery-driven planning doesn’t abandon financial discipline — it relocates it. The discipline shifts from “hit the projected numbers” (a fiction under uncertainty) to “retire the riskiest assumptions per dollar spent” (a ledger you can audit). A venture funded this way can walk into a budget review with something better than a hockey-stick chart: a documented trail of what was believed, what was tested, what it cost, and what’s now known. In my experience that ledger survives contact with finance far better than any demo ever has.

Real options, and the portfolio math of cheap deaths

There’s a name for the financial logic underneath metered funding, and it’s worth knowing because it reframes what a kill means. Stewart Myers coined the term real options in 1977: certain investments are best understood not as commitments but as options — a small payment that buys the right, not the obligation, to invest more later, once uncertainty resolves. A tranche of venture funding is exactly that. You’re not buying the business; you’re buying the right to see the next card, priced accordingly.

Take the logic seriously and two conclusions fall out that most corporations find emotionally intolerable. First, most options should expire worthless — by design. In finance nobody mourns an unexercised option; the small premium was the full, agreed price of resolving uncertainty. Rita McGrath’s academic work on real-options reasoning presses the same point for ventures: the goal isn’t to avoid failure, it’s to contain its cost while preserving the upside of the rare bet that pays for everything. Venture capital runs openly on this arithmetic — a fund’s returns come from a small handful of its portfolio, and the discipline is keeping every other position cheap. A corporate portfolio of new bets obeys the same skewed distribution whether or not the corporation admits it.

Second, killing must be cheap, fast, and normal — because the moment a kill is expensive or shameful, your options stop being options and become obligations, and the arithmetic collapses. This is where culture and finance meet: X, Alphabet’s moonshot factory, pays bonuses to teams that kill their own projects on the evidence, and the mechanism travels — reward the well-argued shutdown memo and zombie projects lose their food supply. A kill in a metered system isn’t a failure; it’s an option expiring at its agreed price, with the learning banked and the people recycled into the next bet.

On portfolio balance, one honest note about the heuristic your executives will quote: 70/20/10 — seventy percent of resources to the core, twenty to adjacencies, ten to transformational bets. Its provenance is thinner than its fame: Eric Schmidt described it around 2005 as how Google managed engineering time, it rhymed with later consulting research on allocation patterns, and it hardened into pseudo-law from there — even as Google itself visibly moved on (“more wood behind fewer arrows,” 2011, was the sound of the ten percent being pruned). Use it as a starting posture and a conversation-forcer — do we even know our current split? is usually a revelation — not as a target to manage to. The right ratio is a function of your industry’s clock speed and your core’s health, and no one else’s number knows either.

The failure modes that survive good machinery

Three ways this all goes wrong, each one I’ve watched from close range.

Theater metrics. Ideas generated, workshops run, hackathons hosted, patents filed, “engagement” with the program — activity counts wearing the costume of results. These are the corporate cousin of vanity metrics, and they’re seductive for the same reason: they only go up. A program measured on ideas generated will generate ideas the way a call center measured on call volume generates calls. If the number can’t go down, it’s not accountability, it’s set dressing.

Zombie projects. The inverse problem: ventures that produce no compelling evidence but never die, because entitlement funding keeps flowing and no forum owns the kill. Zombies are rarely defended loudly — they survive on silence, on the fact that killing them is somebody’s awkward conversation and nobody’s job. This is precisely the gap the growth board closes: a standing body whose explicit mandate includes kill, on a cadence, with the portfolio in front of it. No board, no kills; no kills, and your ten percent is a hospice.

The pilot that succeeds and still dies. The cruelest one. The venture does everything right — evidence, traction, a real customer signal — and then dies at handoff, because the receiving business unit never budgeted for it, isn’t measured on it, and experiences it as an unfunded mandate arriving mid-year with someone else’s fingerprints on it. This is the absorption problem from the lab piece restated in financial terms: graduation is a budget event, and if the money and the metric-relief for the receiving unit aren’t negotiated at the start — in writing, as a condition of the venture’s funding — then every success is merely a deferred kill. The immune system’s last antibody lives at the handoff, and only money moved in advance neutralizes it.

The inside game, complete

Five pieces so far, one claim. Good companies kill good ideas not out of stupidity but out of health — the immune system doing its job — so innovation from within is never a talent problem; it’s a design problem, solved with structure. The structure that works is separation with integration: exploratory units protected from the core’s processes but wired to its leadership, because separation alone builds brilliant orphanages. A lab earns its budget as a school — compounding capability in the people who rotate through it — or it’s a showroom on a shutdown clock. Intrapreneurs are made by systems — permission, funded time, an on-ramp anyone can take — not discovered as rebels, because a program that depends on heroes has a hero-shaped single point of failure. And all of it lives or dies on this final piece: funding that meters money against evidence, measures progress in validated learning, and makes killing cheap enough to be routine.

If you take one thing from these five parts, take this: the companies that innovate from within aren’t the ones with the best ideas or the bravest mavericks. They’re the ones that built the boring machinery — the sponsor, the forum, the tranche, the assumptions ledger, the handoff clause — so that a good idea’s survival stopped depending on luck and started depending on evidence. Ideas are abundant. Machinery is rare. Build the machinery.

And sometimes the honest output of that machinery is the conclusion that a bet can’t live inside your walls at any level of separation — the values conflict is terminal, or the capability simply isn’t yours to build. That’s not the machinery failing; that’s the machinery telling you to change vehicles. The second half of this series is about those vehicles: spinning the bet out as a sister company, investing in someone else’s version of it, learning to pick which startups to back, and the two operating disciplines — pricing and runway — that decide whether any of them survive contact with reality.