Trial, Error, Feedback: The Four Conditions for Innovation
Audit any place, company, or domain for the four conditions innovation actually requires.
- Difficulty
- Advanced
- Time to result
- ~months to results
- Steps
- 5
- Confidence
- 82%
Naval compresses Ridley's argument into a four-part mechanism. Innovation is evolution, and evolution needs trial, error, and feedback. The inputs are a body of innovators you can borrow from, so each attempt starts one step further along. The process is volume of tries, which requires venture capital, cheap company formation, and a friendly environment for starting a business. The selection pressure is error: attempts must be allowed to fail visibly, because covering the downside risk also cuts off the upside. The output is a feedback loop from a large customer base that tells you which variants survive. Remove any one condition and the system stalls. This is why Naval is optimistic about digital and crypto domains, where all four conditions hold, and pessimistic about energy, transport, and biotech, where tries are slow, errors are punished, and regulation lengthens the loop.
Origin
Naval synthesises the argument of Matt Ridley's How Innovation Works into four conditions during the conversation, tying Ridley's historical evidence to what he sees as an investor in Silicon Valley and crypto.
Core principles
- 01Innovation is an evolutionary process, not a breakthrough event.
- 02Evolution needs variation, selection, and feedback; innovation needs the same three.
- 03A dense cluster of other innovators is an input, not a nice-to-have.
- 04Capping the downside of failure also caps the upside of success.
- 05Without a large early-adopter customer base, the feedback loop never closes.
How to run it
- 1
Locate the body of innovators
Find the cluster you can build on top of, whether a physical city or an online community. You need neighbours whose work becomes your starting point rather than something you must rebuild.
Pro tip Ask which nearby companies would sell you components or code so you skip a year of work.
Watch out A talent pool with no shared gathering place is a list of names, not an ecosystem.
- 2
Engineer the ability to take many tries
Volume of attempts is the engine. That requires funding willing to back many shots, cheap and fast company formation, and an environment that does not punish starting.
Pro tip Budget for attempt count, not for a single plan; the plan is one sample from the distribution.
Watch out One well-funded attempt is not the same as many cheap ones, and behaves worse.
- 3
Permit error, including visible error
Selection only works if failures are allowed to happen and be seen. Insurance schemes, guarantees, and precautionary rules that remove downside also remove the upside they were meant to protect.
Pro tip Write down what the organisation does when an attempt fails; that policy is your real innovation rate.
Watch out Do not confuse tolerating error with tolerating negligence; the point is unknown outcomes, not sloppy execution.
- 4
Close the feedback loop with real customers
Feedback comes from the environment, which in practice means a customer base large enough to produce a usable signal. Small markets produce prototypes that never get deployed at volume.
Pro tip Prefer a market where your first customers are themselves innovators; they give faster, sharper feedback.
Watch out A prototype validated only by insiders has not been selected against anything.
- 5
Score the domain and choose accordingly
Rate your domain on all four conditions. Where every condition holds, expect compounding progress; where tries are slow or error is punished, expect stagnation regardless of how important the problem is.
Pro tip Importance of the problem and rate of progress are independent; check both before committing years.
Watch out The most valuable problems often sit in the worst-scoring domains, which is exactly why they stay unsolved.
In the wild
Naval applies the test to crypto and finds every condition present. The body of innovators is global and connected rather than confined to one valley, with more than half of his own crypto investments outside the Bay Area. Tries are cheap and constant, and funding comes from issuing tokens in public rather than a single road of venture firms. Error is tolerated to the point that a full hype cycle collapsed without ending the field. Feedback arrives from users worldwide. His conclusion is that the field kept building through the quiet years and will deploy widely over the following five to ten years.
→ A domain scoring well on all four conditions keeps compounding even after a public bust.
Naval points at speed as the diagnostic case: we cannot travel any faster than we used to, and he attributes that mostly to regulatory reasons rather than physics or engineering ability. Nuclear fusion at scale, hypersonic transport, and parts of biotech all need physical infrastructure, large markets, and relatively deregulated environments. Each attempt is slow and expensive, error is heavily penalised, and the feedback loop from real users can take a decade. The four conditions are not present, so progress stalls even though the problems are among the most valuable humanity has.
→ High-value physical domains stagnate because attempts are rare, failure is punished, and feedback is slow.
Common mistakes
Buying talent without building the loop
Governments and corporates recruit researchers and fund labs, then skip the tolerance for failure and the early-adopter customer base. Variation with no selection pressure produces papers, not innovations.
Insuring away the downside
Removing the possibility of visible failure feels like risk management but deletes the selection step. The same policies that cover the downside cut off the upside.
Prototyping in a market too small to deploy in
A friendly micro-jurisdiction can host the build, but without volume adoption there is no feedback from the environment and the prototype never evolves.
Is it for you?
Best for
Anyone choosing where to build, which domain to enter, or how to design an innovation policy.
Not ideal for
Solo creative work or fields where a single correct answer exists and iteration adds nothing.
From the transcript
“Like any process of evolution, it requires trial and error.”
“There has to be the ability to take lots of tries. You need venture capital. You need startups.”
“And then error. We don't like people making error anymore. So, we try to cover the downside risk, but by doing that, we also cut…”
From the episode
Matt Ridley: How Innovation Works, Part 1
Matt Ridley