Probabilistic Thinking

Thinking ToolsAny stepAlso known as Thinking in Bets

From Bayesian reasoning and decision theory, which have no single author, carried to managers by Daniel Kahneman (Thinking, Fast and Slow, 2011), Philip E. Tetlock and Dan Gardner (Superforecasting, 2015), and Annie Duke (Thinking in Bets, 2018).

Holds beliefs as probabilities, updates them as evidence arrives, and judges a decision by the information behind it, separately from how it turned out.

Lease a second warehouse or rent overflow space: 100 points of belief across four futures

Spread 100 points across the four outcomes of the next two years. Each slider names what each choice earns in that outcome, in two-year profit. The figure weighs the payoffs by the points and shows each choice's expected value, its chance of a loss, and which wins. The opening spread is the team's first; try 25, 45, 20, and 10, the spread after the largest customer renewed. Points that miss 100 are read as shares of what is placed. Illustrative.

15 pts
40 pts
30 pts
15 pts
Points placed
Short of 100: place the restAll 100 placedOver 100: take some back
Expected value, lease
$145K
Expected value, overflow space
$160K
Chance of a loss, lease
45%
Chance of a loss, overflow space
15%
The better bet on these odds
Rent overflow spaceSign the lease

Points on each outcome

025507510015403015SurgeSteady growthFlatDecline025507510015403015SurgeSteadygrowthFlatDecline

An example

A building products distributor was running out of warehouse space, and its operations team wanted to sign a five-year lease on a second building. The case rested on one belief: demand would keep growing. With steady growth, the lease beat renting overflow space month to month by $150,000 over two years.

The finance director asked the team to turn the belief into odds. They named four outcomes for the next two years, a surge, steady growth, flat demand, and a decline, and spread 100 points across them: 15, 40, 30, and 15. Beside each outcome they wrote what each choice would earn. Read that way, the lease was worth $145,000 on average against $160,000 for overflow space, and it lost money in 45 percent of the futures the team believed in.

They rented overflow space and paid the landlord to hold the building for ninety days. Then their largest customer renewed with higher volumes, the points moved to 25, 45, 20, and 10, and the lease, now worth $305,000 against $205,000, was signed.

Why it works

A single belief, such as demand will grow, hides everything else that could happen and what each would cost. Spreading 100 points across the outcomes forces the trade the single belief skips: every point given to growth comes from somewhere, so the team must say how likely the flat and falling years are too.

Once the odds sit beside what each choice earns in each outcome, two numbers settle most cases: the expected value, each payoff weighted by its odds, and the chance of a loss. A choice can win in the most likely outcome and still lose on both. The points also make updating exact. New evidence moves points between outcomes by degrees, and the choice changes only when the weighted sums cross.

The thinking behind it

Annie Duke Annie Duke, a former professional poker player who writes on decision making under uncertainty, argued in Thinking in Bets: Making Smarter Decisions When You Don't Have All the Facts (2018), a popular book, for treating beliefs as bets held with degrees of confidence. A belief held that way can be stated as odds and weighed against what rides on it. She also named resulting, the habit of judging a decision by how it turned out.

Philip E. Tetlock Philip E. Tetlock, a political psychologist who studies forecasting and judgment, led the Good Judgment Project, which competed from 2011 in a forecasting tournament run by IARPA, the US intelligence research agency. Superforecasting: The Art and Science of Prediction (2015), written with the journalist Dan Gardner, describes the habits that set the project's most accurate volunteers apart from the rest, a popular account of research that is Tetlock's own.

Daniel Kahneman Daniel Kahneman, a psychologist, showed with Amos Tversky that intuitive predictions follow how closely a case resembles a category, and so neglect base rates and regression to the mean. Thinking, Fast and Slow (2011) is his popular synthesis of that work and carried it to managers. Its lesson for this tool is the starting point: a probability should begin from how often things of this kind happen before the vivid details of the case move it.

How to run it

  1. Name the outcomes. List three to five futures that do not overlap and together cover everything, the bad ones included.
  2. Spread 100 points across them. Give each outcome its share of belief, starting from how often such futures have come about before; the points total 100.
  3. Write each choice's payoff per outcome. For every option, estimate what it earns or loses if that outcome happens.
  4. Weigh the payoffs by the points. Multiply and add for each choice's expected value, and add up the points on outcomes where a choice loses money.
  5. Choose with the downside in view. Take the higher expected value unless its chance of a loss is more than the business can carry.
  6. Name the evidence that would move points. Say what news would shift belief between outcomes and when it arrives; when the call is close and waiting is cheap, wait for it.
  7. Update and record. Move points by degrees as evidence lands, keep the dated spreads, and score them against what happened.

Where it goes wrong

The outcomes overlap or leave a gap. Steady growth and strong growth both get points, and nobody puts any on a decline. Check that the list covers everything and that exactly one outcome can happen.

The points stop adding up. A good sales call raises the odds of the best case without lowering anything else, and the total drifts past 100. Hold the total fixed, so every increase is paid for.

The average hides the downside. A choice can win on expected value and still lose money in many of the futures, or risk a loss the business cannot absorb. Read the chance of a loss beside the average before choosing.

When to use something else

  • Thinking in Bets. When the result is in and the question is whether the call was sound apart from how it turned out: sort it by the quality of the bet on record and the result.
  • Bayesian Updating. When one belief must be updated through a series of signals, each with known rates.
  • Decision Tree Analysis. When the choice unfolds in stages, with later decisions that depend on what happens first.

Where it sits in SOLVE eX

  • Any step: Tools that serve every step: habits of thought and checks on the thinking itself.

Go to the source

  • Thinking, Fast and Slow, 2011
  • Superforecasting: The Art and Science of Prediction, 2015
  • Thinking in Bets, 2018

Back to the catalog