Innovation collapses at the portfolio layer – long before the work starts
Let me start with a few scenes you’ve probably lived through:
- Approval for an innovation initiative requires a detailed feature list and effort estimation – before a single customer has been interviewed.
- A team is told to produce a 10-year ROI forecast with decimal precision for a market that doesn’t exist yet.
- In a steering committee, a team learns something important and wants to pivot – but they’re blocked because the new direction would invalidate the budget decision made three months ago.
- Teams discover what customers actually want – but can’t pursue it because “the plan” is now locked, so all their real insights get quietly dumped into the backlog.
- After two failed experiments, leadership starts asking whether the team is “reliable enough”, which is a polite way of saying: take fewer risks or this project will disappear.
If you work in product, innovation, engineering or agile coaching, these moments are painfully familiar.
And these aren’t just flaws in a process.
They’re symptoms of the wrong statistical worldview.
And the moment you see a different view, everything snaps into place.
The maths your company runs on – and the maths innovation actually follows
Most corporate systems implicitly assume that the world behaves like a normal distribution. It’s the statistical logic behind:
- budgeting
- forecasting
- risk management
- operations
- performance evaluations
- capacity planning
- project management
In a normal distribution, most outcomes cluster around the middle. Variance cancels out. Predictability matters. The average tells you nearly everything.
This worldview works extremely well for operations.
The problem is: innovation doesn’t follow a normal distribution.
Decades of research show that innovation, firm performance, startup outcomes and creative breakthroughs follow heavy-tailed or power-law distributions:
- A tiny number of ideas create massive value.
- Most ideas fail or produce only modest returns.
- The “average idea” is meaningless.
- Extreme outliers dominate the landscape.
This has been proven repeatedly:
- Pareto (1896) identified a power-law distribution in income and wealth.
- Zipf (1949) showed city sizes follow power laws – an early sign that social and economic systems skew heavily.
- Axtell (2001, Science) analyzed every U.S. firm and found that firm sizes follow a power law with remarkable consistency.
- Gabaix (2009, Annual Review of Economics) demonstrated that economic fluctuations and firm growth rates exhibit heavy-tailed behavior – a few extreme events dominate everything.
- Taleb (2007–2012) wrote extensively about how innovation and risk live in “Extremistan”, not “Mediocristan” – meaning outliers dominate results.
- VC return studies from the Kauffman Foundation, Horsley Bridge and others consistently show that a tiny fraction of startups generate most of the returns while the majority lose money.
The pattern is everywhere: innovation outcomes are naturally skewed.
This matters more than most leaders realize, because:
If the maths of innovation is heavy-tailed,
but your company governs it with normal-distribution logic,
your system will kill breakthroughs before they even begin.
And that’s exactly what’s happening.
Four portfolio-layer failures that quietly kill innovation
These four patterns show up in almost every company that struggles with innovation. Each one looks reasonable through a normal-distribution lens. Each one is lethal in a power-law domain.
Let’s break them down.
1. “We only fund ideas with a guaranteed ROI.”
This is the most common and the most destructive.
In heavy-tailed systems:
- The biggest wins look terrible on spreadsheets early on.
- They emerge from markets that don’t exist yet.
- They outperform all other ideas combined – but only in hindsight.
When leaders demand guaranteed ROI:
- You automatically kill blue-ocean ideas.
- You accidentally privilege small, predictable ideas.
- You filter out the long tail where breakthroughs live.
Axtell (2001) showed that firm success itself is distributed in a power law. Yet companies require innovators to pretend they live in a stable, Gaussian world.
You can’t get outsized returns when your funding logic forbids variance.
2. “Give us a detailed plan before you explore anything.”
In a normal-distribution world, planning before doing makes sense.
In innovation, it makes the plan fictional and the discovery illegal.
This looks like:
- fixed features
- fixed scope
- fixed timelines
- fixed cost estimates
…before any actual learning happens.
But innovation is by definition a discovery process. The first plan is always wrong – that’s the point.
Barabási’s preferential-attachment work (1999) shows that success snowballs only after signals emerge in the real world.
You cannot predict an outlier.
You can only discover it.
So when a company forces teams to commit to a detailed plan upfront, it locks them into the version of reality that existed before discovery.
No pivoting.
No course correction.
No integrating new information.
The team learns – but is not allowed to act on what it learns.
That is systemic failure.
3. “Failure tolerated once, questioned twice, punished at three.”
In operations, repeated failure is a problem.
In innovation, repeated failure is a precondition for success.
VC return data from the Kauffman Foundation and Horsley Bridge show:
- Most startups fail completely.
- A few return 10x, 50x, 100x.
- Those few wins carry the entire portfolio.
Innovation works the same way.
If you punish failure:
- Teams avoid variance.
- They avoid risk.
- They stop exploring.
- They sandbag.
- They “play it safe”.
- They deliver small, predictable outcomes.
Which looks good politically –
and kills any chance at breakthrough value.
Taleb’s work on Black Swans is clear on this point:
You must be exposed to small losses to position yourself for large gains.
But most companies design systems where even small failures are career-limiting.
Under that pressure, no one explores the long tail.
4. “Stay on track – changes invalidate our original decision.”
This one is especially heartbreaking.
A team learns something real.
Something validated.
Something from customer behavior.
Something that opens a door.
But they can’t pivot because:
“Budget was approved for this feature set.”
Innovation ends not with a bang but with a calendar invite:
“Please stay aligned with the baseline business case.”
Meanwhile:
- the real opportunity goes cold
- the discovery is shelved
- the team delivers the wrong thing, knowingly
- and everyone pretends the plan still makes sense
Gabaix (2009) pointed out that extreme events shape economic outcomes far more than “typical” cases. When an organization forbids pivots, it blocks the path to those extreme positive events.
And the breakthrough dies quietly in the corner.
Innovation doesn’t fail in execution.
It fails at the portfolio layer.
You can have the best:
- agile coaches
- developers
- product people
- user researchers
- designers
- experimentation frameworks
…and still fail consistently at innovation if the portfolio logic is wrong.
Because in a heavy-tailed world:
- variance is the point
- discovery is the strategy
- uncertainty is the environment
- pivots are progress
- most ideas die
- a few ideas explode
- averages are meaningless
- outliers matter most
Normal-distribution governance simply cannot cope with this.
It tries to flatten variance.
It forces predictability.
It rewards premature certainty.
It punishes course correction.
It destroys optionality.
In other words:
It kills breakthroughs before they begin.
So what’s the alternative?
Think like a VC, work like an agile team, govern risk through small bets.
This article is not the place for a full playbook, but the outline is clear.
From VC logic:
- Many small bets
- Expect lots of failures
- Double down only on traction
- Kill ideas quickly and kindly
- Evaluate the portfolio, not the project
From agile logic:
- Short cycles of learning
- High contact with customers
- Pivot based on evidence
- Plans that update with reality
- Discovery as a core activity
From sound risk governance:
- Limit exposure, not exploration
- Small initial budgets
- Fixed check-ins
- Clear exit criteria
- No “all-or-nothing” bets
This is how you protect the organization and protect innovation.
The shift that changes everything
When you look at innovation through the lens of power-law dynamics, the familiar struggles start to make sense. Teams weren’t underperforming. Ideas weren’t inherently weak. The problem wasn’t discipline, or motivation, or “lack of accountability”.
The system was optimized for the wrong domain.
Once you see that, many decisions that felt sensible suddenly reveal their unintended consequences. The insistence on predictable ROI. The pressure for detailed plans upfront. The discomfort with pivots. The scrutiny after early failures.
It all fits the logic of operations. It just doesn’t fit the maths of innovation.
And the moment that distinction becomes clear, a different set of options open up. Portfolio thinking starts to feel natural. Small bets stop looking reckless. Learning becomes a measurable asset. Variance becomes something to manage, not eliminate.
From there, redesigning the system feels less like an act of anarchy, and more like setting yourself up for that next breakthrough innovation.
If you want help with overhauling your portfolio and budgeting system, send me a message.