Darwinian Innovation Portfolio
Launch many bounded experiments and let evidence select the survivors
- Difficulty
- Moderate
- Time to result
- ~ongoing to results
- Steps
- 6
- Confidence
- 92%
The Darwinian Innovation Portfolio treats uncertain product development as a selection process rather than a single perfect forecast. Organizations launch multiple bounded, differentiated experiments because customers often reveal demand only after an offering exists. Each launch has predetermined limits, adoption signals, and closure criteria. Most experiments may remain small or fail, but those misses are the cost of discovering products that could not have been confidently predicted in advance. Weak offerings are retired, while survivors receive more distribution and investment. The approach explains why apparent market saturation should not end experimentation: new structures, channels, or customer needs can still produce valuable categories. Its discipline comes from containing downside and using evidence—not from eliminating failure.
Origin
Extracted from Masters in Business during Matt Hogan's defense of the ETF industry's launch-and-closure process.
Core principles
- 01Innovation requires tolerated misses
- 02Outcomes reveal demand better than confident forecasts
- 03Failures should be bounded rather than prohibited
- 04Successful experiments can create new categories
- 05Premature declarations of market completeness suppress opportunity
How to run it
- 1
Define the search space
Identify an unmet need, new channel, or differentiated exposure worth testing.
Pro tip Require each experiment to differ meaningfully from existing products.
Watch out A copycat launch without a distribution advantage is not useful exploration.
- 2
Bound the downside
Set explicit budgets, timelines, compliance limits, and shutdown conditions before launch.
Pro tip Make the cost of several failures survivable.
Watch out Tolerance for experimentation is not permission for unlimited losses.
- 3
Define selection signals
Choose measurable evidence such as adoption, retention, liquidity, revenue, or customer outcomes.
Pro tip Use leading and lagging indicators.
Watch out Do not redefine success after seeing disappointing results.
- 4
Launch the portfolio
Put several qualified experiments in front of real users rather than debating indefinitely.
Pro tip Stagger launches when operational capacity is limited.
Watch out Do not expose customers to products they cannot understand or use safely.
- 5
Select and close
Scale experiments with durable evidence and close those that fail their criteria.
Pro tip Capture lessons from closures for the next launch cycle.
Watch out Keeping failed products alive can consume resources needed by winners.
- 6
Repeat the cycle
Use observed customer behavior to generate the next set of experiments.
Pro tip Revisit assumptions that the market has reached the end of innovation.
Watch out Success in one cycle does not eliminate future uncertainty.
In the wild
ETF issuers launched many products without knowing which structures investors would adopt. Numerous funds closed, but the experimentation also surfaced exposures such as factor and low-volatility ETFs that later became established offerings.
→ A portfolio of launches produced important survivors despite numerous misses.
A software company releases three narrowly scoped workflow tools with fixed development budgets and six-month adoption targets. It retires two low-retention tools and invests further in the one customers repeatedly use.
→ Real behavior identifies the strongest product without risking the company on one forecast.
Common mistakes
Demanding zero failures
Eliminating misses also eliminates many experiments capable of revealing unexpected demand.
Launching undifferentiated copies
A product without unique value or distribution adds cost without meaningfully expanding the search space.
Refusing to close losers
Experimentation becomes wasteful when evidence cannot trigger discontinuation.
Is it for you?
Best for
Organizations able to run multiple bounded product experiments and discontinue weak performers.
Not ideal for
Safety-critical or irreversible domains where failed experiments can cause unacceptable harm.
From the transcript
“So you can't get innovation without some misses.”
“I'm fine with them being launched and closed.”
“we don't know what's gonna work, but we won't know until it until we you know, put up the flag, pas, until we give it…”
From the episode
Matt Hougan Discusses Cryptocurrency and ETF Indexes (Podcast)