Truth is, most startups die.
— 9 out of 10 fail (according to Genome Project)
— 199 out of 200 (according to THNK & Deloitte Fast Ventures)
It’s the elephant in the room.

“The law of floatation was not discovered by contemplating
the sinking of things, but by contemplating the floating of things which floated naturally, and then intelligently asking why they did so.” (Thomas Troward)

TLDR;-) Most innovation frameworks organise by domain. The problem is that domain-based thinking hides the only question that actually matters in early-stage work: what needs to be true before this can work? Critical Path Layers reorders the familiar themes of startup growth and corporate innovation into a dependency sequence. Each layer gates the next. It doesn't tell you what to do. It tells you what to solve first.

Every coaching and advisory framework I've encountered makes the same structural error. Strategy in one column, operations in another, fundraising somewhere else. Neat. Logical. And almost entirely unhelpful for sequencing decisions.

Domain-based organisation tells you what to think about. It says nothing about when. And in early-stage work, when is everything.

Critical Path Layers takes the same familiar themes and reorders them into a dependency sequence. Each layer gates the next. You can work on anything you like at any time, of course. But effort spent on downstream themes before upstream prerequisites are resolved is the single most common pattern of wasted founder and corporate innovator effort. I see it constantly. Strategy before problem clarity. Pricing architecture before product-market fit. Hiring plans before unit economics.

The framework doesn't prescribe. It sequences.

TL;DR: Most conversations about sustainable performance start from the wrong premise—that the performance standards themselves are neutral. They're not. Before optimising for sustainability, ask: whose definition of "good" am I trying to meet? The answer might explain why it feels so hard.

You're exhausted. Not the kind of tired that sleep fixes—the kind that accumulates despite doing everything right. The productivity systems, the boundary-setting, the rest. You've tried it all.

The advice you get assumes the problem is execution. Work smarter. Delegate more. Manage your energy better.

But here's what that advice never questions: the performance standards themselves.

Most B2B proof-of-concept projects fail not because the technology doesn't work, but because no one validated whether the client was willing to bear the internal cost of solving the problem they just discovered.

Your POC worked perfectly. The technology performed. The data confirmed your hypothesis. The client nodded along in the final presentation.

And then nothing happened.

If this sounds familiar, you're not alone. After working with dozens of B2B startups navigating enterprise sales cycles, I've observed a pattern so consistent it deserves a name: the Successful Failure.

The POC technically succeeds. The commercial outcome fails. And founders are left wondering what went wrong.

Here's what went wrong: you validated the wrong thing.

Most feedback is useless.

Not because people lack good intentions. Not because organisations don't invest in training. But because we've been taught to give feedback in ways that trigger defensiveness, focus on personality rather than behaviour, and leave people with nowhere to go.

A colleague recently shared notes from a roundtable in Paris bringing together VCs, Operating Partners, and key players from the French startup ecosystem. The conversation, by all accounts, was sophisticated. Concrete ROI metrics. Honest acknowledgments that "capital alone isn't enough." Thoughtful discussion of AI's impact on productivity and talent.

Reading through the summary, I was struck not by what was discussed, but by what wasn't.

One question was conspicuously absent:

If I had to start a VC or Operating Partner function from scratch today, knowing what I know and accounting for the three-year trajectory, what would I fundamentally do differently?

Instead, the discussion centred on incremental improvements to existing models. Adding GPT wrappers to knowledge bases. Thinking about AI for network matching. Debating batch formats versus continuous intake.

This isn't a criticism of that particular conversation. It's representative of where most of the ecosystem stands: mature enough for sophisticated execution discussions, not yet ready for uncomfortable structural questions.

This article is an attempt to ask those questions. Consider it food for thought.

Most businesses chase growth the hard way. They obsess over customer loyalty, lifetime value, and retention while ignoring the fundamental laws that actually drive sustainable expansion.

The result? Wasted budgets, stalled growth, and missed opportunities.

A follow-up to "Corporate Innovation in the Age of AI: Navigating the Hype, the Hypertail, and the Hard Limits"

In my previous piece, I explored how corporate innovation leaders face four key scenarios in the age of AI: the "hypertail" overload of point solutions, the slow burn of transformation, regulatory compliance pressures, and talent bottlenecks.
While these strategic frameworks help navigate the landscape, they miss a more fundamental truth that's becoming increasingly apparent in boardrooms and innovation labs alike.

The real bottleneck isn't AI adoption—it's organisational readiness.

My previous article, "The AI-Native Paradox," explored how AI has created new challenges for both VCs and founders. But there's a deeper issue we need to address: the metrics we use to measure success are broken.

The same forces that make AI startups hard to evaluate and differentiate have also made traditional software metrics useless. ARR growth rates, churn calculations, and unit economics—the foundation of SaaS investing—don't work anymore.

This isn't just about tweaking formulas. We're witnessing a complete metrics revolution that demands new frameworks for measuring AI startup success. As we've explored in our work on corporate innovation in the AI age, what new metrics or evaluation frameworks are needed to assess the real potential of AI-native startups and solutions?