A venture studio builds twelve companies over four years. Three reach meaningful scale. The studio points to its process: the structured ideation, the stage gates, the shared services layer, the CEO matching. The narrative is clean. The studio’s system produced the outcomes.
But did it?
This is the attribution problem, and it sits at the centre of every claim the venture studio model makes about itself. Studios argue that their value lies in the system, that a repeatable process for creating companies produces better outcomes than the traditional model of backing independent founders. It’s a compelling argument. It may even be correct. But the evidence most studios offer in support of it doesn’t actually establish what they think it establishes.
What attribution requires
To attribute an outcome to a cause, you need more than correlation. You need a defensible account of why the cause produced the effect, and you need to rule out, or at least account for, the other plausible explanations.
In venture, the other plausible explanations are formidable. A company that succeeds after going through a studio process also had a specific founder (or founding team), entered a specific market at a specific moment, hired specific people, encountered specific competitive dynamics, and benefited from (or survived) specific macroeconomic conditions. The studio’s process is one variable among dozens, and it’s entangled with all of them.
The attribution claim isn’t “this company succeeded.” It’s “this company succeeded because of the studio’s process, and would not have succeeded, or would have succeeded less, without it.” That’s a counterfactual, and counterfactuals are notoriously difficult to evaluate in complex systems with small sample sizes.
The selection problem
There’s a layer underneath the attribution problem that makes it harder. Studios don’t just build companies. They select which companies to build. The ideation process, the screening, the stage gates: these are selection mechanisms as much as they are creation mechanisms.
If a studio’s real skill is selecting which ideas to pursue (and which to kill), then the outcomes of the surviving companies reflect selection quality, not process quality. A studio with excellent selection and mediocre execution might produce the same portfolio returns as a studio with mediocre selection and excellent execution. From the outside, the two are indistinguishable. Both point to their track records. Both claim their systems work.
This matters because studios typically attribute their outcomes to the full process: ideation through to scale. But if the value concentrates at the selection layer, then the execution infrastructure (the shared services, the playbooks, the operational support) might be less important than it appears. It might even be unnecessary, provided the selection is strong enough and the execution is handled by capable founders who would have succeeded in any context.
I’m not arguing that this is the case. I’m arguing that most studios have no way of knowing whether it’s the case, because they don’t measure selection quality and execution quality independently. They measure portfolio outcomes and attribute them to the bundle.
The comparison problem
Even if you could isolate the studio’s contribution to a specific company, you’d face a second question: compared to what?
The implicit comparison in most studio narratives is the traditional VC-backed startup. Studios claim better outcomes than the default path of an independent founder raising venture capital. But the comparison is structurally unfair in both directions.
Studios select from a different pool. The ideas they pursue, the founders they recruit, the markets they enter are all shaped by the studio’s thesis, capabilities, and constraints. Comparing studio outcomes to VC outcomes is comparing two different populations, not two different treatments applied to the same population.
The honest comparison would require something that doesn’t exist: a randomised trial where identical founding conditions are assigned to studio and non-studio paths, and outcomes are measured over a meaningful time horizon. Short of that, every comparison is confounded by selection effects, survivorship bias, and definitional inconsistency about what counts as a “studio” in the first place.
The venture studio industry is young enough that the data is thin and the definitions are unstable. Different studies use different criteria for what qualifies as a studio, which companies count as studio-originated, and what constitutes a successful outcome. Under those conditions, anyone claiming definitive comparative evidence is overreading the data.
Why this matters beyond methodology
The attribution problem isn’t an academic concern. It has practical consequences for how studios are built, funded, and evaluated.
If a studio’s outcomes are primarily driven by selection, then the rational investment is in better selection mechanisms, not better execution infrastructure. If outcomes are primarily driven by the quality of recruited CEOs, then the studio’s value proposition is really a talent network, not a company-building machine. If outcomes are driven by market timing, then the studio’s process is largely incidental to its returns.
Each of these has different implications for how a studio should allocate resources, how LPs should evaluate the model, and how honestly the industry should represent itself.
The uncomfortable reality is that most studios, including the ones with strong returns, cannot currently distinguish between these explanations. The sample sizes are too small, the feedback cycles are too long, and the incentive to construct a clean narrative is too strong. This is the same structural problem I examined in the context of venture capital more broadly: an industry where confidence accumulates faster than evidence.
The second-order problem
Everything above is about whether the studio’s process caused the outcome. There is a harder question underneath it: even if the process did cause the outcome, does the studio know that?
These are different problems. The first is causal. The second is epistemic. And they come apart in ways that matter.
Consider a studio that builds ten companies, three of which succeed. The studio believes its stage-gate process was the critical factor. The belief is justified: the stage gates filtered out weak concepts early, concentrated resources on the strongest bets, and provided structured support through the build phase. The belief is true: the stage-gate process genuinely did improve the odds for those three companies relative to the counterfactual.
Justified. True. But is it knowledge?
Gettier showed that justification and truth aren’t sufficient. The justification has to be connected to the truth in the right way. If the studio would have held exactly the same belief with exactly the same confidence regardless of whether the process was actually causative, then the connection between the justification and the truth is accidental. The studio arrived at a correct belief through a path that would have produced the same belief even if it were incorrect.
This is not a hypothetical concern. It describes the default epistemic condition of nearly every venture studio operating today. Studios begin with the conviction that their process creates value. They build companies. Some succeed. The successes are absorbed as confirmation. The failures are attributed to execution problems, market timing, or founder shortcomings, but not to the process itself.
Notice what’s happening: the belief in the process is unfalsifiable in practice. Not because it’s wrong, but because there is no outcome that the studio would accept as evidence against it. A successful portco confirms the process. A failed portco indicts the execution, the market, the founder, anything but the system. Karl Popper would recognise the structure immediately. A claim that is compatible with every possible observation isn’t functioning as knowledge. It’s functioning as ideology, regardless of whether it happens to be true. The studio isn’t testing its process. It’s defending it, and the defence looks identical whether the process works or not.
The mechanism by which this defence operates is worth understanding, because it isn’t cynical. It’s structural. Quine observed that our beliefs don’t face evidence in isolation. They form a web, and when an observation conflicts with the web, we have a choice about where to absorb the shock. We can revise the belief at the centre, or we can adjust something at the periphery and leave the core intact. In practice, we almost always protect the centre. For a studio, the process is the centre. When a portco fails, the studio adjusts its peripheral beliefs (the market wasn’t ready, the CEO wasn’t right, the timing was off) and leaves the core belief in the process untouched. This isn’t dishonesty. It’s how belief systems maintain coherence. But it means the process sits in a protected position where evidence flows around it rather than through it.
Strawson pushed this further. The issue isn’t just that studios protect their core beliefs. It’s that the process functions as an organising principle of the institution’s entire conceptual framework, the thing around which everything else is structured. Questioning whether the process is causative isn’t questioning a belief within the system. It’s questioning the system itself. Studios don’t resist that question out of stubbornness or bad faith. They resist it because the question threatens the coherence of the framework through which they understand their own activity. The conceptual scheme can accommodate a failed company. It cannot easily accommodate the possibility that the scheme itself is explanatorily empty.
This is, I think, the deepest version of the attribution problem. It escalates from a practical question (did this cause that?) through an epistemic one (can you know it even if it did?) to something closer to a structural one: whether the institution, as currently constituted, is capable of the kind of self-knowledge that would distinguish genuine insight from confident self-narration. The belief may be true. The justification may be sound. But the path from justification to truth runs through layers of self-reinforcing architecture so deeply embedded that the institution cannot see them as architecture. It experiences them as reality.
Living with the uncertainty
I don’t think the attribution problem is solvable in any definitive sense, at least not with the data and methods currently available. The sample sizes may never be large enough, the confounders may never be fully controllable, and the counterfactuals will always remain hypothetical.
But I think there’s a meaningful difference between a studio that acknowledges the attribution problem and builds accordingly, and one that doesn’t. The studio that takes attribution seriously designs its process to generate internal evidence: structured decision logs, explicit theses at each stage gate, pre-registered predictions about what will drive outcomes, post-mortems that interrogate the causal chain rather than just recording the result.
None of this produces certainty. It produces something more modest and more useful: a progressively less wrong understanding of what the studio’s process actually contributes. The willingness to pursue that, rather than settling for a narrative that the track record already supports, is itself a form of institutional seriousness.
The venture studio model may well be structurally superior to traditional venture capital. I think there are strong theoretical reasons to believe it is. But “I think there are strong reasons to believe it” is a different statement than “I know it is, and here’s my track record to prove it.” The distance between those two statements is where intellectual honesty lives, and where the real work of understanding this model remains to be done.