A world‑class poker player can make the right move and still lose a huge pot. A CEO can follow a rigorous plan and still miss targets. In both cases, the audience cheers or boos based only on the final score—without ever seeing how good the decision actually was.
A bad outcome doesn’t automatically mean you made a bad decision. That sounds obvious in theory, but in practice we collapse the two constantly. A product launch flops, a hire doesn’t work out, a trade loses money—and we retroactively “explain” why the choice was flawed, even if we’d happily repeat it with the same information.
In this episode, we’ll dig into the core discipline of thinking in bets: separating decision quality from outcome quality. We’ll look at how to evaluate choices *before* results come in, and how to review them *after* without rewriting history. This isn’t about dodging responsibility; it’s about creating a feedback loop that actually teaches you something. When you can say, “That was a good decision with a bad outcome,” you unlock a quieter superpower: the ability to keep taking smart risks without being whipsawed by luck.
In most organisations, outcome and process are welded together. Hit the quarterly target? The project was “brilliant.” Miss it? “What were we thinking?” That shortcut feels efficient, but it quietly trains people to avoid anything uncertain, even when the odds are in their favour. The result: safer choices, slower learning, and a culture where no one wants to own a bold call. To change this, we need a separate lane for judging *how* we chose, not just *what* happened. That means making assumptions, probabilities, and alternatives visible enough that they can be inspected, challenged and improved over time.
Nine times out of ten, Amazon’s experiments “fail.” Internal reports and outside analyses suggest the vast majority of A/B tests don’t beat the existing version. Yet those few that do win, stacked over thousands of disciplined trials, create billions in value. If they judged ideas only by short‑term outcomes, most of those breakthroughs would never have survived their first wobble.
This is the shift: from “Did it work?” to “Given what we knew, how well did we *set up* the decision?” That requires getting much more explicit about what “good” looks like *before* you act.
One practical move is to separate your work into two artefacts:
- A **decision memo**: a brief snapshot of the choice at the moment you commit. - An **outcome log**: what actually happened, recorded later.
The memo focuses on structure: clear objectives, the few key options you seriously considered, your best guess at the upside, downside and most likely case for each, and what evidence you’re relying on. The outcome log captures the result, timing, and any major surprises.
When you review both side by side, you get leverage. Four helpful questions:
1. *Process / good outcome*: Did we succeed **because of** the process, or in spite of it? 2. *Process / bad outcome*: Was this just the 1‑in‑20 downside, or did we miss something systematic? 3. *Pattern spotting*: Are there recurring weak spots—rushed data gathering, narrow option sets, optimistic timelines? 4. *Calibration*: Were our rough odds in the right ballpark, or way off?
Over time, this turns “bad result = blame” into “bad result = data.” You’re training yourself and your team to care more about repeatable edges than about any single flip of the coin.
Here’s where simple tools help. Many high‑reliability teams use lightweight **checklists** not as bureaucracy, but as guardrails: Have we defined what “success” means? Have we sought at least one disconfirming view? Have we written down the main ways this could fail? Confirming you’ve done that thinking doesn’t guarantee a win—but it tilts the odds.
In earlier episodes we framed decisions as bets under uncertainty. This is where that mindset pays off: you’re less obsessed with being right this time, and more focused on becoming the kind of decision‑maker who’s right often enough, over a long series, that the variance stops scaring you out of good moves.
A venture investor faces this tension every day. They can rigorously screen 20 startups—founder track records, market size, technology risk—and still watch their “best bet” fold while an average pick becomes a unicorn. If they judged themselves only by which logos hit the news, they’d either spray money everywhere or retreat into ultra‑safe, low‑return deals. Instead, good funds score themselves on whether each investment matched their written criteria at the time, not on whether it later 10x’d.
Same with product teams rolling out a new feature. Suppose they pre‑commit: “We’ll ship if we estimate at least a 60% chance this improves activation.” Six months later, adoption is flat. Process‑oriented teams don’t just kill the idea; they ask, “Where exactly did our reasoning depart from reality?” Maybe users loved the concept but onboarding confused them; maybe seasonality skewed early tests. That nuance is what lets the next experiment be sharper, not just safer.
When AI agents start making thousands of micro‑choices for you—routes, prices, even hiring—this separation of process and outcome becomes non‑negotiable. We’ll need “flight recorders” for decisions: lightweight traces showing what was considered, ignored, and why. Think of it as a portfolio statement for your judgment: not just the balance, but every trade. As tools surface this history on demand, upgrading how you decide may matter more than what you decide.
Treat each choice like tending a garden: some seeds thrive, some don’t, but the habits—soil, spacing, watering—compound. Your future self is the main beneficiary. As AI, markets and careers shift faster, this skill becomes a quiet hedge: even when single bets misfire, your overall “portfolio” of decisions can keep drifting upward.
To go deeper, here are 3 next steps: (1) Grab Annie Duke’s *Thinking in Bets* and, as you listen to tomorrow’s episode or work on a real decision, use her “resulting” concept to label three past outcomes as either “good decision/bad outcome” or “bad decision/good outcome.” (2) Install a simple decision journal tool like Notion or Logseq and, for your next two meaningful choices (e.g., job move, pricing change, product feature), log your options, probabilities, assumptions, and what “a good decision” looks like *before* you know the result. (3) Watch Michael Mauboussin’s talk “The Success Equation: Untangling Skill and Luck in Business, Sports, and Investing” on YouTube, then build a two-column table for one area of your life (like investing or hiring) and explicitly sort which factors are mostly luck vs. mostly skill, using his examples as a guide.
