There Is No Such Thing As a Long Feedback Loop
Feedback loops are as long as you choose — shorten them with necessary-but-not-sufficient milestones
- Difficulty
- Advanced
- Time to result
- ~months to results
- Steps
- 4
- Confidence
- 93%
People in long-horizon businesses (venture, biotech, enterprise strategy) excuse themselves from decision hygiene by claiming their feedback loop is a decade. Duke's rebuttal: you don't invest and then sleep like Rip Van Winkle — a huge number of observable things happen in between. If you identify the intermediate outcomes that are necessary (even if not sufficient) for the eventual result, or merely correlated with it, and forecast them explicitly at decision time, you can close the loop in months instead of years and start learning immediately.
Origin
Annie Duke, developed while pitching venture firms after Thinking in Bets (2018) and put into practice at First Round Capital (with Josh Kopelman) and Renegade Partners.
Core principles
- 01The length of your feedback loop is a choice, not a property of your industry
- 02A necessary-but-not-sufficient milestone is a legitimate learning signal even if it isn't the goal
- 03At the decision point you are already making an implicit prediction about how the world unfolds — write it down so it can be scored
- 04Long feedback loops offer psychological safety: they protect your reputation from ever being tested
- 05Humans reliably trade long-run improvement for short-run comfort — a short loop risks finding out you were wrong, which is exactly why it teaches you
How to run it
- 1
State the long-horizon outcome you actually care about
Name the terminal result (a >$1B exit, a category-defining product, a drug approval) and be honest about its true horizon.
- 2
List the necessary-but-not-sufficient waypoints
Ask: what must happen for the terminal outcome to be possible? For a seed investor: has any company ever exited above $1B without funding a Series A? No — so a Series A is necessary, and therefore a valid early signal.
Pro tip Also list merely-correlated signals: product-market fit, net new ARR, churn, ability to retain top talent.
Watch out Necessary is not sufficient — don't treat hitting the waypoint as proof the decision was good.
- 3
Forecast the waypoints explicitly at the decision point
When you make the decision, record a probability for each near-term waypoint (e.g. 'probability this company funds a Series A'), with a rationale.
Pro tip You are making this forecast implicitly anyway — writing it down costs nothing and makes it scorable.
- 4
Track the waypoints back to the decision and score yourself
As the waypoints resolve (16 months, not 10 years), compare outcomes to forecasts and feed the accuracy back to each decision-maker.
Pro tip Ask directly: is 16 months shorter than 10 years? If yes, you have no excuse.
Watch out Expect resistance — the tighter the loop, the more often you find out you were wrong, and that is genuinely unpleasant.
In the wild
Venture firms told Duke her methods didn't apply because their feedback loop was a decade. She pointed out that the simplest intermediate signal — does the company fund a Series A? — is necessary for any billion-dollar exit, resolves in roughly 16 months (six months in 2021), and is more correlated with a good exit than the seed check itself. First Round now forecasts Series A funding probability explicitly at the decision point.
→ First Round can now score hundreds of companies against partners' explicit forecasts and tell each partner whether their predictions are better than random.
Common mistakes
Hiding behind the horizon
Claiming a decade-long loop is often psychological self-protection: if you were early into a breakout winner, staying un-measured lets everyone keep believing you're good without ever testing whether it was insight or luck.
Confusing outcome quality with decision quality
A good result tells you nothing about whether you had an insight or your buddy just happened to start Uber. Without explicit forecasts you cannot separate the two — this is 'resulting'.
Is it for you?
Best for
Investors, founders and executives making high-stakes bets whose terminal outcomes are years away and who want to actually improve rather than accumulate unfalsifiable reputation
Not ideal for
Domains with genuinely no observable intermediate correlates, or teams unwilling to be measured — measurement imposed on unwilling people just produces gaming
From the transcript
“there is no such thing as a long feedback loop”
“shorten the feedback loop is to say what are the things that are necessary but not sufficient”
“do you invest and then you go to sleep like Rip Van Winkle”
“it's very very difficult for human beings to deal with feeling wrong in the moment even if it helps them in the long run”
From the episode
This will make you a better decision-maker
Annie Duke (author of “Thinking in Bets” and “Quit,” former