Reference Classes: How Often Does This Kind of Thing Happen?

A new bubble tea shop opens downstairs. It’s beautifully decorated, and there’s a queue every day of its first week. Will it survive a year? Most people focus on the details in front of them — good location, decent tea, friendly owner — and land on an optimistic number. But there’s a better starting point: of shops like this, how many usually make it past a year?

Inside view vs. outside view

• Inside view: start from the specifics of this case and imagine how it will unfold.
• Outside view: treat it as one instance of a kind of thing, and look at how that kind of thing usually turns out.

The historical rate for that kind of thing is the base rate. It gives you a starting point that isn’t colored by your feelings about this particular case.

Which reference class?

The tricky part: the same case can belong to several classes, each with a different base rate. For the tea shop (figures for illustration only):

• All newly opened food and drink businesses: about 60% still open after a year
• Franchises of well-known chains: about 75%
• The 10 shops that opened on this street in the past five years: only 4 lasted a year, or 40%

Which is right? There’s no single answer, but some principles help:

• The more similar, the better. Match on the factors that drive the outcome — industry, size, location, experience.
• Don’t go too small. Ten shops is worth a look; two or three is mostly noise.
• Try several. If different classes give similar numbers, you can be fairly confident. If they’re far apart, the case is genuinely uncertain, so keep your probability away from the extremes.

Switch between reference classes below to see how the base rate shifts, and ask which one best matches the shop you’re forecasting.

Using it in forecasts

In the “Base rate / historical reference” field, write down the class you chose and its historical rate. Even a rough estimate beats pure gut feeling.