Most startups fail. A small number become very large. Nobody can predict each outcome correctly, but a fund can predict better than chance. A fund can also measure how good its predictions are.
To predict well, write each prediction as a probability, start from a base rate, and check the result later.
Write a probability, not a feeling
"I like this deal" is not a prediction. "This company has a 10% chance of an exit above $500M" is a prediction. You can check it.
A number makes you think about the base rate and the size of the outcome. It also lets you compare deals.
Start from the base rate
Find the outcomes of similar companies: the same stage, the same sector and a similar team. This is the base rate.
Then change the estimate for the facts of this deal. Make small changes for each fact. A good story often causes too large a change.
Signals that can help
Some signals are often useful: the past work of the founders, specific knowledge of the market, the speed of progress, and evidence that customers pay.
Each signal is weak alone. Use many signals together, and give each signal a weight that you can check.
Check your predictions
Record each prediction with the date. When the outcome is known, compare it with the prediction.
Check calibration: of all the companies that you gave a 10% chance, did about 10% succeed? If many more or many fewer succeeded, change your method.
Also check the deals that you did not do. A winner that you rejected teaches as much as a failure that you funded.
IdeaPulley keeps a ledger of its predictions and their outcomes, so that a fund can see how accurate the predictions are over time.
Free template: Investment decision log template