The Three-Year Picture: An Innovation Portfolio You Can Steer

The middle horizon is where strategy documents go to be vague. One-year plans have budgets to keep them honest; ten-year visions are allowed to be aspirational; but the three-to-five-year layer — where innovation portfolios actually live or die — usually reads like the annual plan with bigger numbers, which is the tell that no thinking happened there at all. Traction’s Vision/Traction Organizer has an underrated answer for this layer, and it’s a formatting decision: the three-year picture is explicitly not a plan. Gino Wickman’s instruction is to paint a picture — what the business looks like on a specific date three years out: revenue, measurables, and a bulleted description vivid enough that everyone can see the same destination — and to refuse to plan the route, because the route will be re-derived every year by the next one-year plan. Describe the destination precisely; hold the path loosely.

That’s not small-company folk wisdom; it’s the same epistemics the innovation literature arrived at from the other direction. Rita McGrath’s options-reasoning work (Academy of Management Review, 1997) argues that under real uncertainty, upfront commitment to a path is exactly wrong — you hold a portfolio of options on possible futures, kept cheap until uncertainty resolves, and a failed option isn’t waste if the option was priced as one. A three-year picture plus a portfolio of options is an innovation strategy at the middle horizon. This post is about building both halves: the picture, and the portfolio that steers toward it.

Paint the picture: the strategy test hiding in a formatting rule

Writing a three-year picture for innovation forces the choices most portfolio decks dodge, because description is harder to fake than planning. “We will invest in AI-adjacent opportunities” is a plan-shaped sentence that commits to nothing. “In three years, a fifth of revenue comes from products that don’t exist today, we operate in two segments we’re not in now, and our cost of running an experiment has dropped tenfold” is a picture — dated, measurable, arguable. If your leadership team can’t write that paragraph and mean it, you don’t have a portfolio problem; you have a strategy problem, and Roger Martin’s Playing to Win cascade is the repair kit: the picture is a where-to-play, how-to-win claim about the near future, and every bet in the portfolio should trace to it. A bet that serves no line of the picture is a hobby with a budget code.

The best public example of a picture-plus-sequence is also the shortest. Tesla’s 2006 “Secret Master Plan” — still live on their site — is four lines: build a sports car; use that money to build a more affordable car; use that money to build an even more affordable car; provide solar power. It named no dates, promised no features, and specified only the destination and the funding logic between stages. Then it executed: Roadster 2008, Model S 2012, Model 3 2017. Netflix ran the same three-act structure without publishing it — DVDs (1998) funded streaming (launched 2007), streaming’s subscriber base justified originals (House of Cards, February 2013, reportedly around $100M for two seasons, financed in part with hundreds of millions in junk bonds SEC filings document). The transferable pattern in both: each bet’s payoff is the next bet’s funding, which is what “steerable” means — you’re never more than one stage committed, and every stage-gate is a real decision point with fresh information.

Size the portfolio: the golden ratio and its fine print

For the allocation across bet types, the reference point is Nagji and Tuff’s “Managing Your Innovation Portfolio” (HBR, 2012): firms that outperformed tended to allocate roughly 70% to core innovation (improving what exists for existing customers), 20% to adjacent (expanding into new-to-the-company territory), and 10% to transformational (new things for markets that don’t yet exist) — with the returns roughly inverted: the transformational tenth generating the large majority of the long-run payoff. Handle with the fine print the authors themselves attached: the ratio is a descriptive average of winners, not a prescription — the right split varies by industry, competitive position, and how far behind you already are. Its real function in a three-year conversation is as an audit: most incumbent portfolios, honestly categorized, come out near 95/5/0, while their three-year picture quietly assumes 70/20/10 results. Making that gap visible is worth more than any particular ratio.

Two structural notes on where these bets should live, both of which this series’ neighbors cover in depth so I’ll only place them: the horizons vocabulary and its modern failure modes are in the Three Horizons and ambidexterity post — including Steve Blank’s warning that horizons no longer map to time, so a “horizon three” attacker can arrive on commodity tech inside your three-year window — and the institutional home for transformational bets (separate unit, executive integration) is the ambidexterity result I also leaned on in the innovation lab post.

The existence proof: IBM’s EBO program

The strongest evidence that a big company can run a steerable middle-horizon portfolio — not anecdote, a documented program — is IBM’s Emerging Business Opportunities system, born from a famously blunt prompt: Lou Gerstner asking, after IBM missed yet another wave it had itself invented, why the company kept killing its own futures. The design, run by Bruce Harreld from 2000 and documented in the Harvard case literature: EBOs were managed as a separate class of business with different metrics (milestones and learning, not quarterly P&L), senior one-line-of-sight sponsorship, dedicated A-team leaders — signaling mattered; these were prestige postings, not exile — and graduation criteria for rejoining the core. The results over the first five years: roughly 25 EBOs launched, three shut down (and celebrated as correct kills), and the survivors — Linux, pervasive computing, life sciences among them — compounding to about $15 billion in annual revenue, growing faster than the core. For scale: that’s more middle-horizon growth than most companies’ M&A programs produced in the same period, from a program whose unit of management was the portfolio, not the project.

The EBO program is also the cleanest illustration of what the three-year picture does organizationally: IBM didn’t fund “innovation” — it funded named future businesses, each with a described three-years-out state and staged funding against it. When people ask what a middle-horizon innovation strategy looks like inside a real incumbent, the answer is: like that.

Steering: the yearly re-paint

The last mechanism is the one Traction gets exactly right and most corporate portfolios skip: the three-year picture is re-painted every year. It’s not a plan you execute for three years and then review; it’s a rolling destination, revisited at each annual planning session with everything the year’s experiments settled — which is the hinge to the one-year plan post: the quarterly rocks and kill decisions generate the evidence; the yearly re-paint is where the evidence gets to move the destination. A picture that survives three consecutive re-paints unchanged is either remarkably prescient or — far more likely — not being tested by the bets underneath it.

One external calibration worth adding to the yearly session: the corporate venturing market itself. Per PitchBook data, corporate venture capital has participated in deals representing well over 40% of total VC deal value across the last decade — which means your adjacent and transformational bets are competing with, and can sometimes be replaced by, positions in outside startups. Build-versus-back is a real portfolio choice at this horizon, and the series this post lives beside covers the outside game — spin-outs and backing startups — for exactly that reason.

The three-year picture on a page

The artifact, assembled: a dated, one-paragraph picture of the business three years out — revenue mix, markets, capabilities — each claim traceable to a where-to-play/how-to-win choice. A bet table — every material initiative classified core/adjacent/transformational, with the actual allocation totaled next to your intended ratio, gap explained or fixed. A sequencing logic — which bets fund or unlock which, Tesla-style, so no bet needs more than one stage of commitment at a time. Option pricing — transformational bets funded as options (small, staged, killable), never as scaled-down core projects. A standing yearly re-paint, fed by the one-year plan’s evidence.

What the three-year picture borrows from the future, though, it has to borrow from something — a destination worth steering toward across a decade, held through the years when the market calls it a mistake. That’s the ten-year target, the top of the V/TO and the horizon this series starts from.