Combine the inputs by

Fixed cost subtracted

TargetThe outcome you need to clear. The tool reports the chance the result lands below it: your risk of falling short. Set it to 0 to read the chance of an outright loss. to clear

Uncertain inputs (2–5)

centre = your best guess · 2.5% / 97.5% = the 95% range · which shape?How each input is spread around its centre, in plain terms:

Normal: a symmetric quantity, equal room above and below your best guess (an uplift that could go either way, a forecast error, a measurement).
Log-normal: a skewed positive quantity, usually modest, occasionally much larger, never below zero (revenue, deal size, duration, cost).
Gamma: counts of events in a period, or the wait between them (leads per week, support tickets per day, time between sign-ups).
Beta: a probability, rate or share, held between 0% and 100% (conversion rate, click rate, market share).
Triangular, the simplest: three honest guesses, lowest, most likely and highest, for when you have no data to fit (a new product’s first-year sales).

Not sure? Normal or Triangular are safe starts.

Plain guide: distributions explained → · how to get these ranges →

Median outcome

90% of outcomes land between

Chance of missing target

Which input's uncertainty moves the outcome · sensitivity tornado

The outcome distribution · 20,000 simulated runsEach run draws one value from every input's range at once and pushes the set through your model. The pile of 20,000 answers is the distribution of outcomes that is impractical to work out on paper. Red is the part below your target. A 24% chance of missing simply counts runs: about 24 in 100 simulated futures fall short.

Method. Each input is turned into a distribution from your three numbers: a Normal with the spread implied by the 95% range, or a log-normal (median at your centre) for positive, right-skewed quantities. The tornado moves one input across its range at a time, holding the rest at their centres. The distribution draws all inputs together, 20,000 times (a Monte Carlo simulation). The median is the middle of those runs: half the runs land above it, half below. At least one treats every input as a probability and reports the chance that at least one fires: one minus the product of the misses. It assumes the risks are independent; correlation between them moves the true combined chance either way, and this tool does not model it. No data leaves your browser.

Method

From point
guess to range

A back-of-envelope model is a chain of uncertain numbers. Give each one a centre and a 95% range, from data or from structured judgement (elicitationWhen you have no data, structured judgement still gives you a range. The Elicit tool turns three plain-language numbers into a calibrated distribution you can paste straight in here.), and this runs the model thousands of times to show the whole spread of outcomes, which input drives the risk, and the chance you miss your target.

Every input is a range

A single best guess hides the uncertainty that actually decides the call. A centre plus a 95% range (a range the true value should escape only about 1 time in 20) says what you believe and how sure you are, whether the numbers come from data or from judgement.

Sensitivity, one at a time

The tornado swings each input across its range with the others held still. The longest bar is the input whose uncertainty matters most: the one worth measuring better before you commit. With the defaults above (leads × conversion × deal size against the target), deal size throws the longest bar.

iii

Simulate the whole system

Real outcomes move every input at once. Sampling all of them together, thousands of times, gives the full spread, and the number a hard-to-reverse decision cares about most: how often it falls short.

Monte Carlo Simulation

From pointguess to range

Every input is a range

Sensitivity, one at a time

Simulate the whole system

From point
guess to range