The One-Year Innovation Plan: Rocks, Metered Money, and a Kill Cadence

The most honest thing ever said about annual planning came from McKinsey’s own survey of executives: only 45% were satisfied with their strategic planning process, and just 23% said their major decisions were actually made inside it. That survey is from 2006, and I’ve never met a planning cycle since that would move the numbers. The ritual persists anyway — and for innovation work it’s actively corrosive, because an annual plan demands twelve months of projected outcomes for work whose defining property is that the outcomes can’t be projected. You end up with one of two lies: a plan full of confident fiction, or an “innovation” line item vague enough to survive review and be first against the wall when the budget tightens — which, as McKinsey has documented across downturns, is exactly where innovation spending goes first.

This post is about the alternative, and it starts from an unlikely marriage: Gino Wickman’s Traction — the Entrepreneurial Operating System book beloved by small-company operators and mostly ignored by innovation people — and the discovery-driven planning tradition that runs from Rita McGrath to the Lean Startup. Traction supplies the cadence; the innovation literature supplies the funding logic. Neither is sufficient alone. Together they produce a one-year plan that is mostly a machine for revising itself.

What Traction gets right: the year is four quarters, not one plan

EOS’s Vision/Traction Organizer — the two-page V/TO that the whole system hangs off — treats the one-year plan as the hinge between vision and execution: a handful of goals for the year, then rocks, three to seven priorities per person per quarter, reviewed and reset every ninety days. Wickman’s line is that vision without traction is hallucination, and the mechanism behind the slogan is the deliberately short list: if everything is a priority, nothing is, so EOS forces the brutal triage every ninety days instead of once a year.

Two properties of rocks matter enormously for innovation work, and they’re easy to miss under the folksy vocabulary. First, the quarter is the largest unit anyone commits to. The year has goals; only the quarter has commitments. That’s precisely the right epistemic humility for exploratory work — a twelve-month experiment plan is fiction by month three, but a ninety-day commitment to answer a named question is keepable. Second, rocks are reset, not rolled over. The quarterly session re-derives priorities from what was just learned; a rock that no longer makes sense dies without ceremony. Most corporate innovation portfolios have no such reset — projects continue because they were in the plan, and the plan is the artifact everyone negotiated hardest over.

So the Traction-shaped skeleton for an innovation year: a one-year plan stated as three to five questions the business needs answered (not features shipped), decomposed into quarterly rocks, each rock owned by one named person. What Traction doesn’t supply is how to fund and measure work like this — that’s where the innovation tradition takes over.

Fund assumptions, not roadmaps

The founding document here is McGrath and MacMillan’s discovery-driven planning (HBR, 1995), which I leaned on for governance in the innovation lab post and will compress to its one-year form: conventional plans fund a projection; discovery-driven plans fund the test of the assumptions the projection depends on. Practically, the annual innovation budget becomes a pool released in tranches — the internal version of venture rounds. A team gets enough money to retire its next tranche of assumptions (in practice a few months of runway), and re-pitches at each gate with evidence rather than progress-against-plan. (The full financial machinery — why a P&L is the wrong instrument for a venture, and what to report instead — gets its own treatment in the innovation accounting post.) The quarterly rock review and the funding gate are the same meeting — that’s the marriage. EOS gives you a ninety-day ritual everyone already respects; discovery-driven planning gives that ritual teeth by making money move on the outcome.

David Binetti’s “innovation options” framing supplies the sentence to say to your CFO: NPV is the wrong instrument for early innovation because it forces the biggest decision at the moment of least information. An early-stage project isn’t an investment with a return; it’s an option — a small price paid for the right, not the obligation, to invest more once uncertainty resolves. A CFO who would laugh at a $2M speculative roadmap will often happily buy twelve $50K options with quarterly expiry, because that’s an instrument they recognize.

And the ratio question — how much of the year’s capacity goes to this at all — has a serviceable, if over-quoted, answer in Google’s 70/20/10: seventy percent of engineering effort on the core, twenty on adjacent extensions, ten on speculative bets, the split Eric Schmidt institutionalized and from whose 10% Gmail and Google News famously emerged. Coca-Cola later ported the same ratio to its content budget (70% safe, 20% iterating on what works, 10% genuinely new) — evidence the ratio is a portable starting posture, not a law. (One disambiguation, because the collision causes real confusion: this is a resource allocation rule and has nothing to do with the “70-20-10 learning model” from the L&D world, whose evidentiary base is famously thin.)

The plan’s shape: now, next, later — and a bets board

For the artifact itself, drop the Gantt chart and steal two things. The first is the now/next/later roadmap — invented in 2012 by Janna Bastow and Simon Cast of ProdPad, and the correct format for exploratory work precisely because its horizons are confidence bands, not dates: now is committed rocks, next is candidates for the coming quarter, later is direction without commitment. The format refuses, structurally, to promise what you can’t know — which is the entire failure of the innovation Gantt chart.

The second is Spotify’s DIBB and bets board, documented first-hand by Henrik Kniberg: every significant initiative written as a chain of Data → Insight → Belief → Bet, held on a company-level board grouped — no coincidence — now/next/later. DIBB’s value is that it makes the reasoning inspectable: when a bet fails, you can see which link broke — wrong data, wrong inference, or right belief but wrong bet — and that post-mortem legibility is what turns a failed project into tuition rather than pure loss, the Kolb reflection loop from the lab post applied at portfolio level.

A warning about the fashionable alternative: OKRs. They’re excellent for the 70% — known direction, measurable progress. For discovery work they misfire, as a series of practitioner critiques has argued: a key result implies you know what result to pursue, which is the very thing discovery hasn’t discovered yet. Innovation OKRs drift toward theater (“launch 3 pilots”) or fiction (“new bet reaches $1M ARR”). Where you must use the corporate OKR machinery, write learning-shaped key results — assumptions retired, decisions reached — and accept that this is DIBB wearing an OKR costume.

The weekly pulse: a scorecard that counts learning

EOS’s other underrated export is the scorecard — a short list of weekly measurables reviewed in a standing weekly meeting, on the theory that annual goals without weekly numbers are wishes. The innovation version just changes what’s counted. Not revenue (there isn’t any yet) and not activity (workshops held, ideas collected — the vanity metrics Eric Ries’s innovation accounting was coined to kill), but learning velocity: experiments started and concluded this week, assumptions retired, cost and cycle time per validated learning, decisions taken on evidence.

The companies that industrialized this are explicit that volume is the metric. Booking.com reports running on the order of a thousand concurrent experiments — around 25,000 a year — and treats the experiment count, not any individual result, as the engine. Jeff Bezos made the same claim at whole-company scale, in a line Stefan Thomke uses as the epigraph for the experimentation-works argument: Amazon’s success is a function of how many experiments it runs per year, per month, per week, per day. You will not hit Booking numbers, and don’t need to; the point transfers at any scale. If your weekly scorecard can’t say how many experiments concluded and what they settled, you don’t have an innovation plan — you have an innovation intention.

The kill cadence

The piece almost every annual plan omits entirely: a standing, scheduled, socially safe way to stop things. Portfolios rot from zombie projects — too alive to cancel, too dead to matter — because killing a project mid-year requires someone to volunteer for an awkward conversation the plan never scheduled. The quarterly rock reset is the fix, if you run it with the incentive X uses and I described in the lab post: reward evidence-backed shutdowns, in public, at the same session where next quarter’s money moves. A kill at the gate should feel like an option expiring worthless — a small, planned cost of running the portfolio — not like a failure being confessed.

The one-year plan on a page

Assembled, the whole thing fits on one page, which is itself the test:

Three to five one-year goals stated as questions the business needs answered, each tied to the three-year picture (next post). A capacity split — 70/20/10 or your argued-for variant — declared once so the 10% stops being renegotiated monthly. A now/next/later board of bets, each written as a DIBB chain with a named owner. Quarterly rocks — the now column, three to seven per team — funded by tranche, re-decided every ninety days on evidence, with kills rewarded. A weekly scorecard counting experiments concluded and assumptions retired. That’s the entire artifact. Everything else — the deck, the Gantt chart, the twelve-month projections — is decoration for a negotiation, not a plan for a year.

What the one-year plan cannot do is tell you which bets belong on the board in the first place. That’s a three-year question — the picture of the business you’re trying to become — and it’s the job of the three-year picture, one level up the V/TO.