What is what-if analysis?

What-if analysis means asking a model questions of the form, what happens if this input changes? Each answer comes from re-running the model with the new value and comparing the output with a base case. It matters because it lets decision makers test choices safely, before spending money or time on them.

How it works

Start with a base case that reflects the current plan. Then define one or more alternatives, such as a higher price, a bigger team, or a delayed launch. Run the model for each and compare the key outputs side by side, for example profit, cash, and customer count after 12 months. Changing one input at a time shows its effect clearly, while changing several together shows combinations and interactions.

What-if analysis is the plain version of scenario analysis, which bundles inputs into named stories, and of sensitivity analysis, which ranks inputs by effect. When the model is random, compare ranges rather than single numbers, as in stochastic simulation. Models used this way should be validated first, because a flawed model gives confident but misleading answers.

Example

Say a company has 5,000 customers paying 20 per month, with 3 percent monthly churn. The base case after 12 months is about 3,470 customers and monthly revenue of about 69,400. What if the price goes to 24 and churn rises to 4 percent? Customers fall to about 3,060 and revenue becomes about 73,500, a gain of about 6 percent despite the loss of customers. The test supports the price rise unless churn goes above about 4.5 percent. This is an illustration with round numbers.

How to model it in Machinations

Put each input into a Register that a visitor moves as a Lever, such as price and churn. Build customers as a Pool with a Source and a Drain, and revenue as a Register. Play for the base case, then move the Levers and Play again, comparing the chart lines. Use a Forecast when the model has random elements.