Annie Duke on Better Decisions, Kill Criteria, and When to Quit

Annie Duke on Better Decisions, Kill Criteria, and When to Quit

transcriptlennys-podcastdecision-makingquittingleadershipstrategypokerbehavioural-economics

Annie Duke on Better Decisions, Kill Criteria, and When to Quit

Source: Lenny’s Podcast Speaker: Annie Duke Date: May 2024 Link: https://www.lennysnewsletter.com/p/making-better-decisions-annie-duke

Key ideas

  • Only discussion belongs in a meeting (nominal group technique). People assume meetings serve three purposes: discover, discuss, decide. Only discussion belongs in the room. Discovery — gathering individual opinions, forecasts, and estimates — must happen independently and asynchronously before the meeting, so that crosstalk, the loudest voice, and status effects cannot contaminate inputs. Decision should also happen outside the room (a single DRI or an independent vote). Focusing discussion on divergent opinions rather than double-clicking on agreement multiplies the value of the time together.
  • Make the implicit explicit. The core lever in decision quality is externalising intuition into rubrics, structured criteria, and explicit forecasts. Implicit models cannot be examined, compared, or improved. Applied at First Round Capital over five years: structured investment rubrics with one-to-seven ratings on market, team, and product, plus explicit probability forecasts (e.g., likelihood of Series A funding). The data now feeds back to each partner to reveal where their judgements are predictive and where they are not — including where confident table-pounding is correlated with nothing.
  • There is no such thing as a long feedback loop. Feedback loops are as long as you choose. For any long-horizon outcome, identify leading indicators that are necessary (though not sufficient) for it — things you know must happen on the way to the outcome. In VC, fund at Series A is a measurable signal in ~16 months, not 10 years. Long feedback loops often serve as psychological cover: if the loop is long, you cannot be proven wrong, and past lucky outcomes (e.g., an early Uber investment) can masquerade as skill indefinitely.
  • Pre-mortem + kill criteria = pre-commitment. A pre-mortem alone rarely changes plans. The real power comes from converting predicted failure signals into kill criteria with attached pre-committed actions. Example from a sales team: if the prospect only wants to discuss price and won’t demo → kill the lead; if no decision-maker appears after initial meetings → offer executive alignment, and kill if declined. Pre-committing actions before sunk-cost and status-quo bias set in is what makes a pre-mortem operationally useful rather than a cathartic exercise.
  • By the time you’re thinking about quitting, you’re probably already late. Three compounding biases delay quitting: sunk cost (not wanting to “waste” prior investment), endowment effect (we over-value what we own relative to identical things we don’t), and identity threat (internal and external self-image tied to continuing). The Glitch→Slack story: Stewart Butterfield quit when back-of-the-envelope CAC maths told him the business was not venture-scale, while the company still had $6M in the bank and great press — not when forced. The invisible cost of persisting is the opportunity cost of alternatives you cannot yet see.

Overview

Annie Duke — author of Thinking in Bets and Quit, special partner at First Round Capital, former professional poker player — delivers a tightly structured episode on how to close the gap between knowing about biases and actually changing decision-making behaviour. Covers the nominal group technique, the First Round Capital structured-decision experiment, feedback loop design, pre-mortems with kill criteria, and the Glitch→Slack story as a masterclass in timely quitting. Also includes a tribute to Daniel Kahneman (adversarial collaboration, humility about non-replicating findings) and a brief parenting segment (mental time travel; nevertheless).