🎯 Decision Science for Product Leaders

What you’ll learn

1. Judge the Decision, Not the OutcomeSeparate decision quality from outcome ('resulting') and set the course through-line: calibrate confidence to evidence, match process to reversibility.In an uncertain world, good decisions can yield bad outcomes and vice versa, so judging decisions by results ('resulting') trains people to chase luck. The only controllable thing is decision quality. The course's aim is to decide well more often, not to predict better — calibrating confidence to evidence and matching care to reversibility.
2. The Biases Waiting to Ambush YouIntroduce cognitive biases (confirmation, sunk cost, anchoring, availability) and the principle of countering them with process, not willpower.Mental shortcuts cause systematic, predictable errors (Kahneman & Tversky). Four do the most product damage: confirmation bias, sunk cost fallacy, anchoring, and availability, and they interlock. You can't will them away; you build counter-processes (independent estimates, disconfirmation, forward-only framing). The sunk-cost cure: ask whether you'd start fresh today.
3. Base Rates: The Question Almost Everyone SkipsTeach base rates, base-rate neglect, and Bayesian updating as the core of calibrated reasoning.Intuition fixates on vivid case details (inside view) and neglects the background frequency (base rate / outside view), which is often more predictive. A worked hypothetical shows how a low base rate makes an 'accurate' detector mostly false-positive. Bayesian thinking starts from a prior and updates by evidence strength, reframing a decision as a confidence-weighted estimate you revise honestly.
4. Expected Value and the Shape of a BetTeach expected value, asymmetric bets, and the one-way/two-way door rule for matching process to reversibility.Ask 'what's the shape of the bet?' not 'will it work?' Expected value separates probability from payoff magnitude. The best bets are asymmetric (capped downside, large upside), like experiments; the worst hide catastrophic tails behind small upside. Match deliberation to reversibility: decide two-way (reversible) doors fast, deliberate carefully on one-way (irreversible) doors.
5. Evidence: Correlation, Causation, and the ExperimentDistinguish correlation from causation, establish the randomised experiment as the clean test, and use Kohavi's data to argue for humility.Observational data invites confounding and reverse causation (e.g. feature use and retention both driven by an 'engaged' user type). Only randomised A/B tests reliably establish causation, because randomisation balances hidden confounders — and an experiment is itself a cheap, reversible, asymmetric bet. Kohavi reports only ~1/3 of tested Microsoft ideas improved the target metric; ~1/3 flat, ~1/3 negative.
6. Prioritisation: Useful Frameworks and Their TrapsPresent prioritisation frameworks (RICE) and their traps — false precision and bias against asymmetric moonshots — and how to hold them loosely.Prioritisation frameworks like RICE (Reach × Impact × Confidence ÷ Effort) make triage explicit and embed calibration and expected value. Their traps: false precision (guessed inputs reverse-engineered to justify a pre-chosen idea) and a structural bias against high-effort, low-confidence long shots. Use them as aids to judgment, value the conversation they force, and reserve roadmap room for asymmetric bets.
7. Deciding Well, TogetherAddress group decision-making — groupthink, premortems, independent judgment, disagree-and-commit, and the decision log — and synthesise the course.Groups can decide worse than their smartest member via groupthink. Countermeasures: premortems (assume failure and ask why), independent judgments before discussion (avoid anchoring), and disagree-and-commit (argue then align). The decision log records reasoning and confidence, defeating hindsight bias and enabling calibration. The whole course reduces to: calibrate confidence to evidence and match process to reversibility.

Questions this course answers

Grounded in trusted sources

• Annie Duke — 'Thinking in Bets' (resulting; decisions vs outcomes)
• Daniel Kahneman — 'Thinking, Fast and Slow' (heuristics, biases, base-rate neglect)
• Ron Kohavi et al. — 'Trustworthy Online Controlled Experiments' and Microsoft experiment keynotes (~1/3 of ideas succeed)
• Amazon shareholder letters — one-way vs two-way door decisions; disagree and commit
• Intercom — the RICE prioritisation model (Reach, Impact, Confidence, Effort)
• Gary Klein — the premortem technique

Every Wunder lesson is built from real, reputable sources — never invented.

Related Science courses