Decide better, on purpose.

A working guide to the mental models and habits that decision research supports, with a protocol, a journal and a 30-day plan for putting them to use.

Read this first. No model guarantees a good outcome. What the evidence supports is a small set of practices that reduce predictable errors. Each entry below is marked for how strong its support is, so you can weigh it accordingly, and the advice at the end is to start with three, not twenty-two.

  • StrongReplicated studies, meta-analyses or large field results.
  • GoodSeveral studies or a solid mechanism, with thinner evidence on real-world outcomes.
  • HeuristicWidely used by practitioners, little direct testing. Treat as an experiment.

Size your decision

Most wasted effort comes from treating every decision the same. Answer two questions.

If it goes wrong, how hard is it to undo?

How much is at stake (money, time, relationships, reputation)?

The models

Twenty-two tools in six groups. Each says what it is, what the research behind it shows, and one concrete thing to do with it.

Bias counters

The models exist mainly to counter a handful of reliable errors. When you notice one of these pulling at a decision, reach for its counter.

BiasHow it distortsWhat to do about it
Confirmation biasYou look for and favor evidence that agrees with what you already think.Ask what would change your mind. Write the case against your choice.
OverconfidenceYour certainty runs ahead of your accuracy.Give probabilities and ranges, then score yourself in the journal.
Planning fallacyYou underestimate time, cost and risk.Use the outside view: how long did comparable projects really take?
AnchoringThe first number you see drags your estimate toward it.Form your own estimate before seeing anyone else's. Collect several reference points.
Sunk costWhat you have already spent pushes you to keep going.Ask, "Knowing what I know now, would I start this today?" Set kill criteria in advance.
Status quo and loss aversionLosses feel bigger than equal gains, so the default feels safe.Reverse test: if I were already in the alternative, would I switch to where I am?
Hindsight biasAfter the outcome, you remember having predicted it.Write predictions down beforehand.
AvailabilityVivid or recent stories feel more likely than they are.Look up base rates instead of reasoning from anecdotes.
Narrow framingYou see one or two options and treat them as all there are.Generate at least three, including a small test version.

The protocol for decisions that matter

This is the models assembled into one sequence. Use it for anything the sizer marks as major, and a shortened version (steps 1, 3, 4, 6 and 8) for meaningful decisions.

  1. Frame the decision

    5 minutes

    Write the decision as a question. Add what success looks like in concrete terms and the date by which you must decide. Check that you are solving the real problem and not a symptom.

    Uses one-way and two-way doors

  2. Set a time budget and a state

    2 minutes

    Decide how long this deserves. Then check your own state: if you are angry, anxious, elated, hungry or exhausted, schedule the decision for later.

    Uses cool-state rule

  3. Widen the options

    10 to 20 minutes

    List at least three real options, including a smaller test of the big one. Name the opportunity cost of your front-runner. If you can test cheaply before committing, design that test now.

    Uses widen your options opportunity cost

  4. Take the outside view

    15 to 30 minutes

    Find comparable cases and how they turned out. Write your starting estimate from that base rate, then adjust only for specific, nameable differences. Put numbers on your confidence.

    Uses outside view probabilistic thinking

  5. Score against criteria you set first

    15 minutes

    Choose four to six criteria and weights before rating options. Rate each option on each criterion separately. If the total disagrees with your gut, investigate the gap instead of overruling either.

    Uses structured scoring expected value

  6. Stress-test the leader

    20 minutes

    Run a pre-mortem on your front-runner, write the strongest case against it, and trace the consequences two or three steps out. Ask one or two people for independent opinions before you tell them your leaning.

    Uses pre-mortem consider the opposite second-order thinking

  7. Get distance, then sleep on it

    One night

    Apply the friend test and the 10-10-10 check. Remember that you will adapt to the outcome faster than you expect. If the decision is hard to reverse, let a night pass before committing.

    Uses self-distancing 10-10-10 affective forecasting

  8. Commit and record

    10 minutes

    Log the decision, your prediction and your confidence. Write kill criteria and a review date. Turn the first action into an if-then plan so follow-through does not depend on willpower.

    Uses decision journal kill criteria if-then plans

Script: a 10-minute pre-mortem
  1. Describe the plan in two sentences and set a 10-minute timer.
  2. Say, "It is one year from now. This failed badly."
  3. Write every reason you can think of, without judging them. Aim for at least eight.
  4. Circle the two or three most likely or most damaging.
  5. For each, either change the plan or write an early warning sign you will watch for.
  6. If you are doing this with others, have everyone write privately first, then share.
Script: an outside-view check
  1. Name the category your decision belongs to (for example, "first-time founders opening a food business" or "job changes to a new industry").
  2. Find three to five comparable cases from research, industry data, or people who have done it.
  3. Write down how they turned out: success rate, typical time, typical cost.
  4. Set your starting estimate to that range.
  5. List specific reasons you differ. Move your estimate only as far as those reasons justify, and write them down.
Script: the friend test and 10-10-10
  1. Write your situation in the third person, using your own name or "a friend".
  2. Write the advice you would give that person.
  3. Write how they would feel about each option in 10 minutes, 10 months and 10 years.
  4. Compare the advice with what you were about to do. If they differ, write down why.

Decision journal

The single most useful habit here is recording decisions before you know how they turn out. Entries are saved in this browser on this device only. Nothing is sent anywhere, so keep in mind that clearing your browser data will remove them.

A 30-day plan to make it stick

Adding everything at once is a reliable way to keep nothing. This plan builds one layer per week. Tick items as you complete them; your progress is saved in this browser.

If you only do three things: keep the decision journal, run a pre-mortem before big commitments, and check the outside view before you trust your own estimate. Add the rest once those feel routine.

The rhythm after 30 days

Daily, 2 minutes
Log any decision that mattered, with a prediction.
Weekly, 20 minutes
Review due entries. Ask: which decisions did I make this week? Did I size them? Where did I notice a bias? What is one thing I will try next week?
Monthly, 30 minutes
Check your calibration summary. Are you consistently over- or under-confident? Adjust your confidence numbers accordingly.
Quarterly, 1 hour
Update your personal checklist. Drop any practice that has not earned its time and add one you are missing.

Where the evidence is thinner

Good guidance should say where it is uncertain.

Decision fatigue is shakier than its reputation. The well-known parole-judge study (Danziger, Levav and Avnaim-Pesso, 2011) has been criticized on methodological grounds, and a large multi-lab replication of the broader ego-depletion effect (Hagger et al., 2016) found little to no effect. Incidental emotions do leak into unrelated choices (Lerner, Small and Loewenstein, 2004), so the cool-state rule still stands, but for that reason rather than because willpower runs out.

Your gut is not always the enemy. Kahneman and Klein (2009) agreed that intuition can be trusted where the environment is regular and you have had plenty of fast feedback, as in chess or firefighting. It is unreliable in noisy, low-feedback domains such as long-range forecasting or new ventures. Lean on the models most where you lack that kind of practice.

Good process does not guarantee good outcomes. Luck is real. Judge a decision by what you knew and how you reasoned, which is why the journal asks you to separate the two when you review.

Knowing many models is not the same as using them. The evidence supports specific practices: the outside view, pre-mortems, if-then plans, structured scoring and feedback. I am not aware of good evidence that memorizing a long list of models improves decisions on its own.

Effect sizes come from particular settings. Many findings come from lab studies, hospitals or organizations. Applying them to personal decisions is a reasonable extrapolation, not a proven result. Treat the heuristics especially as experiments and let your own journal tell you what works for you.

Sources and further reading

Listed so you can look them up. Check the originals before quoting specific figures.

Books

  • Daniel Kahneman, Thinking, Fast and Slow (2011)
  • Kahneman, Sibony and Sunstein, Noise (2021)
  • Philip Tetlock and Dan Gardner, Superforecasting (2015)
  • Chip Heath and Dan Heath, Decisive (2013)
  • Annie Duke, Thinking in Bets (2018) and Quit (2022)
  • Atul Gawande, The Checklist Manifesto (2009)
  • Gabriele Oettingen, Rethinking Positive Thinking (2014)
  • Howard Marks, The Most Important Thing (2011)

Papers and articles

  • Arkes and Blumer (1985), The psychology of sunk cost
  • Buehler, Griffin and Ross (1994), Exploring the planning fallacy
  • Fischhoff (1975), Hindsight is not equal to foresight
  • Frederick et al. (2009), Opportunity cost neglect, Journal of Consumer Research
  • Gilovich and Medvec (1995), The experience of regret
  • Gollwitzer and Sheeran (2006), Implementation intentions and goal achievement: a meta-analysis
  • Grossmann and Kross (2014), Exploring Solomon's paradox
  • Grove et al. (2000), Clinical versus mechanical prediction: a meta-analysis
  • Hagger et al. (2016), A multilab preregistered replication of the ego-depletion effect
  • Iyengar, Wells and Schwartz (2006), Doing better but feeling worse
  • Kahneman and Klein (2009), Conditions for intuitive expertise: a failure to disagree
  • Kahneman and Lovallo (1993), Timid choices and bold forecasts
  • Klein (2007), Performing a project premortem, Harvard Business Review
  • Kross et al. (2014), Self-talk as a regulatory mechanism
  • Lerner, Small and Loewenstein (2004), Heart strings and purse strings
  • Lord, Lepper and Preston (1984), Considering the opposite
  • Mellers et al. (2014), Psychological strategies for winning a geopolitical forecasting tournament
  • Mitchell, Russo and Pennington (1989), Back to the future: temporal perspective in the explanation of events
  • Nickerson (1998), Confirmation bias: a ubiquitous phenomenon
  • Nutt (1999), Surprising but true: half the decisions in organizations fail
  • Pronovost et al. (2006), An intervention to decrease catheter-related bloodstream infections in the ICU, NEJM
  • Samuelson and Zeckhauser (1988), Status quo bias in decision making
  • Schmidt and Hunter (1998), The validity and utility of selection methods in personnel psychology
  • Schwartz et al. (2002), Maximizing versus satisficing: happiness is a matter of choice
  • Staw (1976), Knee-deep in the big muddy
  • Trope and Liberman (2010), Construal-level theory of psychological distance
  • Tversky and Kahneman (1974), Judgment under uncertainty: heuristics and biases
  • Wilson and Gilbert (2005), Affective forecasting: knowing what to want