A century-old enterprise found its capital controls were driving entrepreneurs away.

The Mulliez family built more than 130 businesses across five generations. Its portfolio now spans 90 companies in 80 countries, including Decathlon and Auchan.

Yet for almost 15 years, no family member launched a new venture within the family’s capital network.

The problem was not a shortage of ideas or ambition. It was the architecture surrounding them.

In 2002, the family launched CREADEV, a formally governed investment fund that required lengthy applications and majority ownership stakes. According to one shareholder reporting on the family, the vehicle struggled to attract the very entrepreneurs it was created to support.

The family then tried a different model. A leaner investment vehicle removed the majority-stake requirement and simplified the application process. More than 40 family entrepreneurs participated, and the vehicle exceeded typical venture-fund success rates.

The capital had not changed. The entrepreneurs had not changed. The threshold for experimentation had.

CREADEV tried to control risk before an idea had the opportunity to prove itself. Its successor allowed ideas to begin, then judged them through results.

Resilient organizations do not eliminate discipline. They move it to the point where evidence exists.

The Principle That Replaced Layers of Control

Joseph Tsai faced the same organizational temptation at a much larger scale.

As vice chair of Cathay Financial Holdings, Tsai helped guide a third-generation enterprise that now employs roughly 52,000 people. Growth at that scale normally invites additional committees, approval layers, and risk controls.

Tsai chose a simpler operating principle: protect customers and employees, even when doing so reduces short-term returns.

That principle shaped decisions across the organization. In a difficult year, the bank retains capital rather than prioritizing a dividend. When leaders need perspective, Tsai seeks out people willing to disagree rather than adding another formal review.

The results appear in operating scale. Cathay United Bank now runs Taiwan’s largest digital banking platform, with more than 6.8 million users and 47 million monthly logins, according to an interview with Tsai.

Total assets nearly doubled over the past decade, rising from $240 billion to $457 billion.

Cathay did not avoid discipline. It created a principle strong enough to guide thousands of decisions without requiring thousands of approvals.

That distinction matters. Rules prescribe what people may do before circumstances are known. Principles clarify what must remain true while people respond to those circumstances.

The organizations that accumulate controls fastest are not necessarily the ones managing risk best. Often, they are transferring judgment from the people closest to the work into processes too distant to recognize opportunity.

Why Small Experiments Outperform Large Commitments

Organizations often treat change as a choice between full authorization and complete rejection.

An executive proposes a new operating model. Leaders request a business case, multiple reviews, cross-functional alignment, and proof of an outcome that cannot exist until the idea has been tested. The organization demands certainty before permitting the activity capable of producing it.

The alternative is not ungoverned experimentation. It is a smaller unit of commitment.

This is a microexperiment: one limited, observable test of a different behavior.

The concept applies beyond individual development. Instead of approving a transformation, test one workflow. Instead of building a new governance layer, give one team temporary authority. Instead of debating whether a model will scale, define the smallest environment in which it can produce evidence.

Small experiments lower the cost of beginning without lowering the standard for continuing.

That is the architecture CREADEV lacked. Its controls operated before evidence existed. Its successor made participation easier, then allowed results to determine which ventures deserved more capital.

Resilient leaders separate two decisions that organizations routinely collapse into one:

• Is this idea safe enough to test?

• Has this test produced enough evidence to scale?

The first threshold should be low. The second should be demanding.

Five Disciplines for Governing Experimentation

1. The Minimum Viable Permission

Most internal ventures begin with an approval burden designed for a mature operation. Leaders demand projections, stakeholder alignment, and precise returns before allowing a small test.

That process selects for people willing to navigate bureaucracy, not necessarily those best equipped to build something valuable.

Implementation Architecture

Define the smallest version of an idea that can be tested without creating unacceptable risk. Set firm limits on its budget, duration, customer exposure, and operational scope.

Release additional permission only as evidence accumulates. Do not require evidence that only the experiment itself can produce.

2. The Evidence Checkpoint

Astellas Pharma keeps early research investment broad across multiple candidates. When a program reaches clinical proof of concept, the company significantly increases its resources.

The checkpoint prevents promising experiments from becoming permanent programs through tenure, momentum, or internal advocacy.

Implementation Architecture

Define one observable result that must be reached before the next round of funding, staffing, or executive attention is released.

Review the evidence on a fixed date. If the program misses the checkpoint, pause or close it regardless of how long it has operated or who supports it.

3. End-to-End Ownership

Astellas eliminated its functional axis in 2025, replacing therapeutic-area silos with cross-functional teams responsible for programs from discovery through delivery.

This structure prevents weak initiatives from hiding between departments while each function claims to have completed its portion.

Implementation Architecture

Build a small team around the program’s outcome, with one accountable leader and access to every function required at its current stage.

Rotate leadership as the program advances, but never transfer accountability during a handoff. The team owns the result until the initiative scales or closes.

4. The Precommitted Ending Rule

Experiments become expensive when organizations define how they begin but not how they end. By the time failure is undeniable, sunk costs and internal identities have become arguments for continuation.

Clear endings preserve more trust than vague reassurances.

Implementation Architecture

Before launch, document the metric, date, or condition that will trigger closure. Include the ending rule in the original approval so it cannot be quietly rewritten later.

When the condition is reached, communicate the decision, reason, constraint, and consequence. State what is ending, why it is ending, and what happens next.

5. The Replacement Portfolio

Astellas built five strategic brands with blockbuster potential partly to offset the eventual patent loss of its top-selling drug, which has benefited more than 1.5 million patients globally.

A single initiative becomes difficult to end when the organization has no alternative already producing evidence.

Implementation Architecture

Fund two or three bounded experiments around every strategically important outcome rather than concentrating all expectations in one flagship program.

Track them as a portfolio with shared checkpoints. When one experiment closes, redirect its resources toward the alternative demonstrating the strongest evidence.

A pipeline with one initiative is not a pipeline. It is a bet.

The 90-Day Experimentation Audit

The Mulliez family restored entrepreneurship by removing the terms preventing ideas from being tested, then letting evidence determine which deserved more capital.

Over the next 90 days, audit every active initiative against three questions:

• What is the smallest safe test?

• What evidence would justify further investment?

• What condition would require it to end?

Programs without clear answers are either being obstructed before they begin or protected after they should stop.

Resilient leaders do not demand certainty before experimentation. They demand evidence before continuation.