In one line
Planning under uncertainty means judging a plan by the range of results it could produce, not by one forecast number. Run the same plan in many possible worlds and you see its middle, its spread and the odds of hitting a target.
What it does
- A forecast is a single number, usually an average or the most likely case. The year that actually happens is one draw from many that could have, each with its own demand, rival moves and news.
- A Monte Carlo simulation runs the same plan many times and lets the uncertain parts vary each time. The results form a distribution: its middle (the median), its spread (here the band that holds eight worlds in ten) and its tails. David Hertz set the method out for business decisions in 1964.
- Averages mislead when results are lopsided or there is a hard line to clear, such as a profit target. Sam Savage calls this the flaw of averages: a plan built on average assumptions does not deliver the average result, and the odds of beating a forecast can be far from even.
- Plans differ in spread as well as in their middle. A plan with a slightly lower middle and a narrower spread can be the better bet when missing a target is costly.
Business and brand
Impact on the businessP&L, cash, share price
- Boards set targets as single numbers, so the useful question is the chance of meeting them, not whether the forecast does.
- Cash and borrowing limits are tested in the bad years: plan for the one-in-ten world as well as the middle one.
- Two forecasts a little apart may be the same plan within its noise; money spent moving from one to the other buys little.
Impact on the brandMemory, loyalty, pricing power
- Brand building pays back over years, so judging it on one quarter's number understates it and reads its noise as failure.
- Volume bought with deep deals and loaded stock is the least certain kind: in the lab it gives Brightwater the widest spread of share.
- Share moves slowly where most buying is habitual, so a quarter's swing is often luck rather than the plan working or failing.
Try it
Run one plan in thirty worlds
Pick a plan for the year and a measure, then set a target. Each dot is one world: the same plan, with its own demand, rival moves and news. The dashed line is the forecast the desk would show.
Hold price, standard budget, a 60:40 media mix, light deals.
| Plan | Forecast | Middle | Eight in ten | Hit target |
|---|---|---|---|---|
| The team's plan | 16.3% | 16.8% | 16.0% to 17.5% | 10 of 30 |
| Chase volume | 17.3% | 17.6% | 16.5% to 20.3% | 19 of 30 |
| Build the brand | 17.5% | 17.9% | 17.2% to 18.9% | 28 of 30 |
Notice: The forecast says 16.3%, the average of the desk's three worlds. Across 30 worlds the middle is 16.8%, eight in ten land between 16.0% and 17.5%, and 23 of 30 beat the forecast. 10 of 30 (33%) reach your target of 17.1%. On this target, "Build the brand" gives better odds: 28 of 30.
Runs Countermove's full game: the reference soft-drinks market, three brands, rivals running their own policies, news and demand noise, in the thirty numbered worlds the desk's risk view uses.
Sources
- Hertz (1964), Harvard Business Review: risk analysis in capital investment
- Savage (2009), The Flaw of Averages, Wiley
- Kahneman & Lovallo (1993), Management Science: timid choices and bold forecasts
- Tetlock & Gardner (2015), Superforecasting, Crown