Shortening Feedback Loops

There's no such thing as a long feedback loop—find intermediate signals correlated with success

Annie Duke
This will make you a better decision maker

Shortening Feedback Loops

"There is no such thing as a long feedback loop. And the way you choose to shorten the feedback loop is to say, what are the things that are correlated with the outcome that I eventually desire?" - Annie Duke

What It Is

Shortening Feedback Loops is the practice of identifying intermediate signals that correlate with your ultimate goal, so you can learn and improve faster. Rather than waiting months or years to know if a decision was good, you find leading indicators that give you signal in weeks or days.

The insight comes from Annie Duke's work with venture capital firms. When she first talked to VCs about decision quality, many said "Our feedback loops are a decade—we don't know if an investment was good until exit." Her response: "Do you invest and then go to sleep like Rip Van Winkle? Or are there all sorts of things that happen in between?"

The feedback loop is only as long as you choose it to be. If you're not tracking intermediate signals, you're voluntarily living in the dark.

How It Works

The framework has two components:

1. Identify Necessary-But-Not-Sufficient Conditions

What must happen for your ultimate goal to be achieved? Work backward from success:

  • For a successful exit → Company must fund at Series A → Must achieve product-market fit → Must show traction
  • For a successful product launch → Must hit adoption milestones → Must get positive user feedback → Must ship on time

These intermediate milestones are signals. A company that doesn't fund at Series A has never exited for $1B. That's a necessary condition. Track it.

2. Find Correlated Signals

Beyond necessary conditions, what observable factors correlate with eventual success? These don't have to be causal—correlation is enough for learning:

  • Quality of founding team (rated on specific dimensions)
  • Speed of initial traction
  • Ability to retain key talent
  • Customer engagement patterns

Then track your predictions against these signals.

How to Apply It

For Venture Investing (First Round Example)

Annie helped First Round create structured evaluation forms with specific components:

At Investment:

  • Rate market quality (1-7)
  • Rate founder quality (1-7)
  • Rate product quality (1-7)
  • Forecast: probability company funds at Series A

At Checkpoints (16 months later for Series A):

  • Did they fund at Series A?
  • Compare to forecast

Now you can ask: When partner X rates a market as 7/7, how does that correlate with company performance? Is their market judgment predictive? What about founder judgment?

Instead of waiting 10 years, you learn in 16 months whether your judgment is calibrated.

For Product Development

Ultimate Goal: Successful product launch with strong adoption

Intermediate Signals (2-4 weeks):

  • Sprint velocity—are we hitting estimates?
  • Stakeholder engagement—are decision-makers showing up?
  • Technical risk resolution—are unknowns being retired?

Intermediate Signals (6-8 weeks):

  • Internal dogfooding feedback
  • Alpha user engagement metrics
  • Technical debt trajectory

Intermediate Signals (12+ weeks):

  • Beta user retention
  • NPS scores
  • Support ticket volume

Track your predictions at each stage. "I think we'll hit 80% sprint velocity" → actual: 65%. Adjust.

For Hiring

Ultimate Goal: Great hire who performs well for 2+ years

Intermediate Signals (2-4 weeks):

  • Did they pass the interview loop with strong scores?
  • Did they accept the offer? (Reveals something about role attractiveness)

Intermediate Signals (3-6 months):

  • 90-day review performance
  • Manager feedback
  • Peer feedback
  • Self-ramp assessment

Structured Evaluation at Interview:

  • Rate specific competencies (not just "gut feel")
  • Forecast: probability this person is a strong performer at 1 year

Now compare your interview ratings to 6-month performance reviews. Which competencies actually predict success? Where is your intuition calibrated vs. miscalibrated?

Why People Resist This

Annie identifies psychological reasons people prefer long feedback loops:

1. Protection of Reputation "If I was early into Uber, I don't really want to know if I'm good or not. People already think I'm good." Shortening feedback loops risks revealing that a lucky outcome wasn't actually skill.

2. Aversion to Feeling Wrong Tight feedback loops surface mistakes faster. That's uncomfortable in the moment, even if it helps in the long run. Humans notoriously trade long-term improvement for short-term comfort.

3. Power Law Obscures Signal In domains with extreme outcomes (VC, startups), a few massive wins can mask poor decision quality on average. Intermediate signals reveal the true hit rate.

The reframe: If you're willing to do the work of shortening feedback loops and others aren't, you have an enormous edge. Most people don't.

When to Use It

  • Long-term investments: VC, real estate, major strategic bets
  • Product development: Multi-quarter initiatives
  • Hiring decisions: 2-year performance horizon compressed to 90-day signals
  • Career decisions: "Will this role work out?" tracked through intermediate milestones
  • Customer relationships: Multi-year contracts with early warning indicators

The Core Question

For any decision with delayed feedback, ask yourself:

"What are the things that are necessary but not sufficient for success? What are the things correlated with the outcome I eventually desire? Am I tracking them?"

If the answer is no, you're choosing to live with a long feedback loop. That's a choice, not a constraint.

Source

  • Guest: Annie Duke
  • Episode: "This will make you a better decision maker"
  • Key Discussions:
    • (00:45:43) Brett Berson's question about decision quality with long-term outcomes
    • (00:48:46) "Do you invest and then go to sleep like Rip Van Winkle?"
    • (00:49:55) Intermediate signals: Series A funding, product-market fit, traction, talent retention
    • (00:50:51) "There is no such thing as a long feedback loop"
    • (00:56:05) How First Round now tracks partner forecasts and ratings against outcomes
    • (00:58:58) Feeding back accuracy information to improve decision-making
  • YouTube: Watch on YouTube

Related Frameworks

Back to Decision Making