Venture Studio Characteristics Associated with New Venture Equity Exits

Startups → Investing → Research

The BVSR24 dataset contains information about venture studios, the people who run them, the ventures they build, and the products created by those ventures. Where in that stack are the associations with equity exits?

The Dataset

BVSR24

Source

Big Venture Studio Research, 2024

Sample

117 respondent venture studios

Geography

7 regions

Design

Cross-sectional survey

Voluntary survey of active studios. Survivorship bias and self-selection bias acknowledged.

The Search Space

What the dataset captures

The studio

Studio age (years)

Portfolio size (ventures created)

Team size (FTEs)

Equity stake (ordinal)

Fund structure (categorical)

Venture throughput (per year)

Region (7 categories)

Has exits (binary)

Its people

Founder exit experience (ordinal)

Talent deployment by lifecycle stage:

Generation, Ideation, Validation, Creation, Growth

Its ventures

Vertical focus (binary)

Niche focus (binary)

Business model focus (binary)

Its products

MVP cost (ord. 1-3)

MVP complexity (ord. 1-4)

MVP build time (ord. 1-3)

Three constructed variables

TotalStages count of lifecycle stages deployed at (1-5)

Specialisation count count of focus areas: vertical, niche, biz model (0-3)

MVP composite cost + complexity + build time (3-10)

The Outcome

Why equity exits?

I needed an outcome variable. Not "success," which means different things to different studios. Something operationalisable.

●

An equity exit is a market-validated signal that the studio created something of economic value to at least one external party.

●

It is binary and observable: the studio either has at least one exit, or it doesn't.

●

It is not a measure of quality. A single small exit counts the same as an IPO. The bar is "at least one."

38 of 117 studios (32.5%) have at least one equity exit.

First Look

The bivariate picture

Every variable tested against exit status. No controls. What shows up on its own?

Shows something

Studio age

p < .001

Strong

Portfolio size

p < .001

Strong

Founder exit experience

p = .039

Significant

Team size (FTEs)

p = .063

Marginal

Shows nothing

Lifecycle stages

Creation .018 (1 of 5); rest > .10

Marginal

MVP characteristics

All > .43

Specialisation

Niche .072; rest > .16

Marginal

Everything else

All > .29

Composites

TotalStages

p = .168

Specialisation count

p = .902

MVP composite

p = .819

Age and portfolio size dominate. None of the three composites show anything either.

The Choice

Two correlated dominant variables (r = .584)

Studios with exits are nearly twice as old (5.5 vs 3.0 years). Age and portfolio size both associate with exits at p < .001.

Model fit (AIC)

StudioAge: 119.9

TotalNewCos: 132.3

Deviance reduction

StudioAge: 31.6

TotalNewCos: 19.2

Causal direction

Age is upstream (R² = .34)

Portfolio size is downstream

Age wins on every criterion: better fit, stronger confounder, and it's the upstream variable. Portfolio size is partly a consequence of age, not an alternative explanation.

The Approach

Controlling for age

There were 38 exit events, nearly 20 variables, and a methodology question.

The confounder

Studio age is the strongest bivariate predictor and the most obvious confounder. More time means more opportunity for exits. With 38 events, parsimony matters.

The method

Hosmer & Lemeshow's purposeful selection: control for the obvious confounder, then test every other variable independently against that baseline. Each variable gets a fair shot.

The composites

The three composites we constructed (TotalStages, specialisation count, MVP composite) showed nothing bivariately. But do any of them emerge once age is accounted for?

The question

Does any variable significantly improve on a model that already accounts for studio age?

Screening

Every variable against the age baseline

Portfolio size

p = .096

Team size

p = .419

Throughput

p = .198

Founders

p = .636

Equity stake

p = .379

Fund structure

p = .664

Vertical

p = .464

Niche

p = .578

Biz model

p = .793

Region

p = .516

Founder exits

p = .069

Generation

p = .175

Ideation

p = .702

Validation

p = .337

Creation

p = .085

Growth

p = .226

MVP cost

p = .743

MVP complexity

p = .364

MVP time

p = .513

Specialisation

p = .406

MVP composite

p = .777

TotalStages

OR 1.44

p = .044

Twenty-two tests. One crosses the threshold: TotalStages (p = .044). Roughly what you'd expect by chance. But the one that crosses is deployment breadth, and it only appears once age is accounted for.

Unpacking

Breadth, not any single stage

Each individual stage was also tested separately alongside studio age.

Generation

OR 1.95

p = .175

Ideation

OR 1.20

p = .702

Validation

OR 1.60

p = .337

Creation

OR 2.29

p = .085

Growth

OR 2.02

p = .226

TotalStages (breadth)

OR 1.44

p = .044*

The association is with the pattern of deployment across stages, not with deployment at any particular stage.

Forward Selection

What else improves the model?

TotalStages entered the model. Following Hosmer & Lemeshow: test every remaining variable against the updated baseline (StudioAge + TotalStages). Does anything else add?

Nothing crosses the significance threshold.

Closest candidate

Founder exit experience

OR 1.69, p = .088

The observation

The variable that entered (TotalStages) concerns deployment breadth. The runner-up (FounderExits) concerns founder experience. Both relate to how the studio deploys entrepreneurial talent.

Forward selection stops here. But the pattern raised a question: these variables describe different types of information about studios. Does any type, taken as a group, add to the age baseline?

Following the Lead

Four questions about studios

Scale

Portfolio size

Team size

Throughput

Studio specialisation

Vertical

Niche

Business model

Entrepreneurial talent deployment

TotalStages (deployment breadth)

Founder exit experience

Product

MVP cost

MVP complexity

MVP build time

How big is it? How is it focused? How does it deploy talent? What does it build? Four questions. Do any of them, taken as a group, add to the age baseline?

Domain Tests

One domain adds to the picture

Scale

Portfolio size, team size, throughput

p = .392

Not significant

Studio specialisation

Vertical, niche, business model focus

p = .188

Not significant

Entrepreneurial talent deployment

TotalStages (OR 1.41) + founder exits (OR 1.69)

Joint p = .031

Significant

Product

MVP cost, complexity, build time

p = .643

Not significant

One domain is associated with exits after controlling for age: how broadly the studio deploys entrepreneurial talent and whether the deployer has done this before.

Interpretation

What we observe

Of four domains tested, only entrepreneurial talent deployment is associated with exits after controlling for age. Not scale. Not product characteristics. Not specific stage choices. How the studio deploys talent and experience.

Within that domain, the pattern is in breadth of deployment combined with founder experience. Neither variable crosses conventional significance alone. Jointly, they do. The domain captures something neither variable captures individually.

The association strengthens in mature studios. Consistent with some kind of accumulation effect, though the cross-section cannot confirm the mechanism.

Limitations

Let's keep in mind.

Cross-sectional: associations, not effects.
TotalStages not significant bivariately; emerges only after controlling for age.
Within the entrepreneurial talent deployment cluster, neither variable is individually significant. The domain-level association should be interpreted as a joint association.
22 individual screening tests, 1 crossing the conventional threshold. Roughly what you'd expect by chance at α = .05.
Voluntary survey, active studios only, survivorship and self-selection bias.
Survey instrument captures a broad construct ("founders or operators").
Exploratory: data not designed to test specific theoretical frameworks.

Open Questions

Where to next?

What does broad deployment look like in practice?

The variable counts stages. It doesn't describe what studios actually do. Qualitative work could help describe the mechanism.

How do scale, specialisation, and product matter?

These three domains showed no association with equity exits. These are the things practitioners obsess over and intuitively they must matter, but how? The data says they don't differentiate.

The next survey can be sharper.

The survey instrument has a number of design complications that require adaptation for a longitudinal study. I have Max and Maxim's encouragement to take the study forward longitudinally so I welcome conversations on how to make that happen.