Future & Contribution
Choices That Hold Up in Many Futures
Also called: Scenario planning and robust decision-making
- Personal interest
- Formal theory
- Working interpretation
When I cannot know how the future will turn out, I prefer choices that work reasonably well across several believable futures. Models that use probabilities help show where a plan is fragile, without pretending to know what tomorrow will bring.
Which choice still works if my main prediction turns out to be wrong?
Why it attracts me
A plan built for the average day is fragile in a world where few days are average. I like the shift in question this idea brings. Instead of asking what will happen, I ask which choice I can live with across the futures that might happen.
The idea
Scenario planning sketches several believable futures, each different in a way that matters. A robust choice is not the best one under any single forecast. It is the one that does acceptably well across most of them. Simulation helps because it plays a plan out under many conditions instead of one (Simulating a System Event by Event). As evidence arrives, I can also change how likely I think each scenario is (Adjusting My Confidence as Evidence Arrives), without throwing the others away.
An example
The Queueing & Staffing Lab on this site makes this concrete. With its starting settings, 12 inspection requests an hour, 20 minutes per job and a 15-minute target, six inspectors answer about 94 percent of requests on time. Now suppose demand turns out to be 14 an hour instead. The same six inspectors fall to about 83 percent. Seven inspectors clear the 90 percent goal in both cases. If I am unsure which demand is coming, seven is the choice that survives being wrong.
What I think (and don't know)
Robustness has a price. The seventh inspector sits partly idle on quiet days. I am still learning how to weigh it honestly against the cost of a plan that breaks.
What this does not establish
Testing a plan against several futures shows where it is fragile. It does not predict which future will happen or guarantee that the plan will work.
Questions I'm still exploring
- How many futures do I need to test before a choice counts as robust?
- When is it worth giving up the best result in one future for safety in many?
- How do I avoid imagining only the futures that suit what I already want?
Sources and further reading
- Pierre Wack, "Scenarios: Uncharted Waters Ahead", Harvard Business Review (September 1985) — An early account of scenario planning at Royal Dutch Shell.
- Robert J. Lempert, Steven W. Popper and Steven C. Bankes, Shaping the Next One Hundred Years: New Methods for Quantitative, Long-Term Policy Analysis (RAND, 2003) — Sets out robust decision-making: choosing plans that do well across many possible futures.
Working interpretation: drafted from my notes and interests for review. It is not a direct quotation, and I may still change it.