CPOs protect new bets by refusing to judge them on core-product logic. The Three Horizons model splits a portfolio into today's cash cow (Horizon 1), tomorrow's growth line (Horizon 2), and unproven experiments (Horizon 3) — each needs its own metrics, funding rules, and governance, or the mature business's stronger ROI will always win the argument and starve the bet.
Quick answer: Judge a Horizon 3 bet by Horizon 1 standards — revenue, margin, CAC payback — and it will lose every review, because a proven business always out-argues an unproven one on paper. Protect it by pre-agreeing different metrics, a ring-fenced budget, and separate governance before the pressure starts, not after.
The Three Horizons Model: Three Businesses Sharing One P&L
A CPO's portfolio is rarely one business. It is usually three, at different stages of maturity, competing for the same finance committee's attention. The Three Horizons framework, introduced by McKinsey's Mehrdad Baghai, Stephen Coley, and David White in The Alchemy of Growth (1999), names them plainly and gives each a different job.
- Horizon 1 — the core. Defend and extend what already makes money.
- Horizon 2 — emerging growth. Scale a business model that works but isn't yet mature.
- Horizon 3 — options. Seed and test ideas that may become tomorrow's Horizon 1.
The framework's real contribution isn't the taxonomy — it's the warning underneath it. Left alone, a company's planning process, incentives, and headcount gravitate almost entirely toward Horizon 1, because it is the only horizon with clean numbers to defend.
This isn't a one-time budget mistake — it's a recurring pattern. Every fiscal year, the same debate resurfaces: a proven business with real revenue asks for more investment, and a fragile new bet asks for patience. Without a standing rule for how horizons get funded and judged, the proven business wins that debate almost by default, every single time.
| Horizon | Time frame | Job to be done | Typical evidence of success |
|---|---|---|---|
| Horizon 1 (Core) | Now, this fiscal year | Defend market share, extract margin | Revenue, margin, retention, NPS |
| Horizon 2 (Adjacent) | 1–3 years | Scale a proven-but-immature model | Growth rate, unit economics trending positive |
| Horizon 3 (New bets) | 3+ years, uncertain | Validate a new problem/market/model | Learning velocity, validated assumptions |
Google's own accounts of its 70/20/10 resource-allocation model are the best-known attempt to force this split into a budget line rather than leave it to instinct — 70% core, 20% adjacent, 10% new bets. Google executives, including in How Google Works by Eric Schmidt and Jonathan Rosenberg, have acknowledged the actual split drifted well below 10% for the riskiest work in practice.
The ratio itself is directional, not a formula — plenty of well-run companies land somewhere else entirely. The discipline it's meant to enforce is the point: a pre-committed number, revisited on its own schedule, beats no number at all. A complete CPO playbook treats horizon allocation as a standing portfolio decision, reviewed on its own cadence, not an afterthought of the annual budget cycle.
Why Core-Business ROI Logic Kills Horizon 3
Applying Horizon 1 financial discipline to a Horizon 3 bet isn't rigor — it's a category error. The core's metrics assume a known customer, a known demand curve, and a repeatable cost structure. A new bet has none of those yet, so every number you plug into a standard ROI model is a guess dressed up as a forecast.
Clayton Christensen's research in The Innovator's Dilemma explains why this keeps happening even at well-run companies. Resource-allocation processes are built to reward proposals that serve known, profitable customers with clear near-term payback — which structurally starves anything serving a smaller or unproven market, no matter how strategic it looks on a roadmap.
Rita McGrath and Ian MacMillan's Discovery-Driven Planning (Harvard Business Review, 1995) names the specific failure: traditional NPV and payback-period models require you to already know the assumptions that a new bet exists to test. When a Horizon 3 team is forced to produce a five-year revenue model to get funded, they aren't being held accountable — they're being asked to fabricate false precision.
| What Horizon 1 assumes | Why it breaks for Horizon 3 |
|---|---|
| A known, addressable market size | The market may not exist yet — or exists differently than assumed |
| Stable CAC/LTV to model payback | No repeatable acquisition channel or price point yet |
| Quarter-over-quarter revenue growth | The right early metric is often adoption, retention of intent, or problem validation |
| A fixed discount rate on future cash flows | Cash flows are speculative; the real currency is validated learning |
This is also where pricing logic breaks. A core product's monetization and pricing strategy is built on known willingness-to-pay data. A Horizon 3 bet frequently doesn't have a price yet — because it doesn't have proof anyone will pay at all. Forcing a monetization model before that proof exists is how good bets get killed on a technicality: "it isn't revenue-positive," when revenue was never the question being tested.
There's an older, blunter way to see the same imbalance: the Boston Consulting Group's growth-share matrix, developed by Bruce Henderson in 1970, labeled a mature, cash-generating business a "cash cow." A cash cow's whole job is to fund question marks and stars — not to out-compete them for capital on the same ROI bar it was built to clear decades ago. When a CPO lets the core's cash-cow economics set the bar for a Horizon 3 "question mark," the comparison was rigged before the meeting started.
Ring-Fencing Innovation Funding: The Rules That Actually Protect a Bet
A ring-fenced budget is money set aside for Horizon 3 work before the fiscal year starts, governed by rules that make it structurally hard to quietly reabsorb into the core. Without that pre-commitment, a new bet's funding gets relitigated every quarter against a business that already has better numbers.
Geoffrey Moore's Zone to Win is the sharpest treatment of why this fight is so predictable. Moore describes recurring "custody battles" over people and budget between horizons, and argues the only durable fix is to wall Horizon 3 off structurally — a separate "incubation zone" evaluated on its own terms, not folded into the core business unit's quarterly reviews.
In practice, ring-fencing means committing to a small set of explicit rules, in writing, before pressure arrives:
- Fix the allocation as a percentage of revenue or profit, set annually — not derived bottom-up from each idea's projected ROI, which will always lose to a core-product feature with a cleaner business case.
- Fund in milestone-based tranches, not annual budget lines — release the next tranche when a learning milestone is hit (validated demand, a working prototype, a signed pilot), not when a fiscal quarter rolls over.
- Protect headcount separately from core sprint capacity — a Horizon 3 team whose engineers get "borrowed" for core-product firefighting during crunch effectively has no protected funding at all.
- Name an executive sponsor with real veto power — someone senior enough to say no when a budget review tries to reallocate the money, and accountable for defending that no.
This is portfolio-level capital allocation, not project-level prioritization — the two disciplines use different tools for a reason, which is why a dedicated capital-allocation framework for CPOs treats horizon funding as its own decision layer above individual roadmap trade-offs.
Governance: Different Metrics, Different Boards, Different Cadence
Funding rules only hold if the governance reviewing them asks different questions per horizon. A steering committee that reviews a Horizon 3 bet with the same revenue-and-retention dashboard it uses for the core product will, without meaning to, kill the bet within two quarters.
| Horizon | Primary metric | Review cadence | Who can kill it |
|---|---|---|---|
| Horizon 1 | Revenue, margin, churn | Monthly/quarterly, standard board reporting | Business unit leadership, standard budget process |
| Horizon 2 | Growth rate, unit economics trend | Quarterly, growth-stage board | GM + growth committee, against scaling milestones |
| Horizon 3 | Validated assumptions, learning velocity | Milestone-based, not calendar-based | Named exec sponsor only, against pre-agreed criteria |
For a Horizon 3 bet, the honest early signal usually isn't a revenue chart — it's whether the team can show evidence that a real, underserved job is getting done better than the alternatives people currently use. That's precisely the evaluation lens a Jobs-to-Be-Done framework is built for: opportunity scoring and progress-forces analysis instead of a P&L.
Governance also has to accept that a new bet is often serving an unfamiliar customer path, not optimizing the one your core product already knows well. Mapping how that new audience actually experiences the emerging offering — where the friction and drop-off really are — matters more early on than a customer journey benchmarked against your mature product's funnel, which doesn't yet exist for something this new.
This is exactly the audit a new CPO should run early. A first-90-days plan should include an honest horizon-mapping exercise: which initiatives are actually Horizon 1, 2, or 3, and which ones are Horizon 3 bets quietly being judged — and slowly killed — by Horizon 1 governance inherited from a predecessor.
Signs a Horizon 3 Bet Is Being Judged by the Wrong Yardstick
Watch for these tells in a quarterly business review — each one means Horizon 1 governance has quietly annexed a Horizon 3 bet:
- The first question in the review is about revenue or margin, not what was learned.
- The bet is on the same slide template, with the same columns, as core-product line items.
- "Kill" and "scale" decisions get made by the core business unit's leadership, not the named exec sponsor.
- Funding gets pulled mid-quarter to cover a core-product shortfall, with no formal reallocation decision.
A Separate Operating Model in Practice: IBM's Emerging Business Opportunities Program
The clearest documented example of a structurally separate Horizon 3 operating model is IBM's Emerging Business Opportunities (EBO) program, studied and written up by Charles O'Reilly and Michael Tushman in The Ambidextrous Organization (Harvard Business Review, 2004) and later in their book Lead and Disrupt.
Under EBO, championed by then-strategy executive Bruce Harreld, IBM identified promising but immature businesses and deliberately separated them from the core reporting structure:
- Their own small team, reporting to a senior executive rather than the relevant business-unit P&L owner.
- Different success metrics — strategic milestones and market-validation evidence, not the revenue and margin targets applied to mature IBM units.
- Explicit executive sponsorship, shielding the unit from being folded back into core reporting before it was ready.
O'Reilly and Tushman's research found this structural separation, paired with tight strategic alignment to the parent company (shared brand, shared customer relationships, shared distribution where useful), correlated with EBOs that survived long enough to mature — while EBOs left inside standard business-unit reporting were more likely to be judged by the wrong yardstick and starved early.
Amazon's early build-out of AWS inside a retail company followed a similar shape, if less formally documented: a new, capital-intensive infrastructure bet, evaluated early on adoption and developer traction rather than held to retail-level margin targets, before it eventually became large enough to report and govern as its own segment. The pattern in both cases is the same own team, own metrics move — not because the bet was mature enough to deserve autonomy, but because autonomy was the precondition for letting it get there.
What "Own Team, Own Metrics" Actually Requires
A separate operating model for a Horizon 3 bet is not a full spin-off. It typically means three concrete things:
- A dedicated, cross-functional team small enough to move fast, staffed with people not double-booked on core-product delivery.
- A written charter defining success criteria specific to that bet's stage — agreed with the exec sponsor before the work starts, not renegotiated at the first uncomfortable board review.
- A reporting line that bypasses the core business unit for at least the incubate/validate stages, so the team is evaluated against its own charter instead of a neighboring unit's dashboard.
Documenting the Rules Before the Pressure Hits
None of this holds under real organizational pressure unless the different success bar is written down at the moment a bet is funded — not reconstructed from memory six months later, under a CFO who's asking why the new thing "still isn't profitable."
This is a narrow, honest use case for Prodinja's Journals. A CPO can use the decision journal to log, at the moment a Horizon 3 bet is approved, the specific criteria it's actually being held to — learning milestones, validation thresholds, the funding tranche schedule — separate from the core product's KPIs.
When a later budget review reaches for the wrong yardstick, there's a timestamped record of what was actually agreed, instead of a fragile, relitigated memory. It won't make the funding decision for you — it's a place to make sure the decision, once made, stays made.
Key Takeaways
- The Three Horizons model (McKinsey: Baghai, Coley, White) splits a portfolio into Horizon 1 (core), Horizon 2 (scaling), and Horizon 3 (new bets) — each needs different metrics, funding, and governance.
- Applying Horizon 1 ROI logic — revenue, CAC payback, discounted cash flow — to a Horizon 3 bet forces false precision onto assumptions that don't exist yet, per Christensen and McGrath/MacMillan's research.
- Ring-fence Horizon 3 funding as a fixed percentage set in advance, released in milestone-based tranches, with protected headcount and a named executive sponsor who can veto reallocation.
- Governance must ask different questions per horizon: revenue and margin for Horizon 1, growth trend for Horizon 2, validated learning and assumption testing for Horizon 3.
- A separate operating model — its own team, its own metrics, its own reporting line — is what let IBM's EBO program and early AWS survive long enough to mature, per O'Reilly and Tushman's ambidextrous-organization research.
- Write the horizon-specific success criteria down when a bet is funded, not after the first hard budget review — a decision journal is one honest way to keep the record straight.
Frequently Asked Questions
What is the Three Horizons model in product management?
The Three Horizons model, from McKinsey's Baghai, Coley, and White, splits a company's portfolio into Horizon 1 (the mature core business), Horizon 2 (scaling growth bets), and Horizon 3 (early, unproven experiments). Each horizon needs its own metrics and funding logic — treating all three the same is what causes new bets to lose out to the core.
Why do good new product bets get killed too early?
New bets usually die because they're judged against core-business metrics — revenue, margin, CAC payback — before those numbers can honestly exist. Christensen's research on resource allocation shows this isn't a leadership failure so much as a structural one: standard funding processes are built to reward proposals with known, near-term payback, which a Horizon 3 bet can't yet produce.
How much budget should a CPO set aside for Horizon 3 innovation?
There's no universal number, and any figure you see (including Google's well-known 70/20/10 split) is directional, not a formula. What matters more than the exact percentage is that the allocation is fixed in advance as an explicit budget line, funded in milestone-based tranches, and protected by a named executive sponsor — not re-derived project-by-project against core-product ROI.
What metrics should a Horizon 3 product be measured on?
Early on, measure validated learning: evidence that a real job is getting done better than current alternatives, adoption among early users, and progress against the specific assumptions the bet exists to test — not revenue or margin. McGrath and MacMillan's discovery-driven planning approach was built exactly for this stage, replacing a forecast with a testable checklist of assumptions.
How do you structure a separate operating model for a new bet?
Give the bet a small dedicated team not double-booked on core-product work, a written charter with success criteria specific to its stage, and a reporting line that bypasses the core business unit's leadership during the incubate and validate stages. IBM's EBO program and Amazon's early build-out of AWS both followed this shape, evaluated on strategic and adoption milestones rather than core-business margin targets.