Four times in two days… People told me that their win rates have changed dramatically.

Q. How are you computing that?

A. Won/(Won + Lost)

What about your open deals?

Turns out that 80% of people responding to our survey use this formula above which omits open deals. About 17% include the open deals in their measurement. And 2% claim to do something else. We wondered what those folks were doing so we interviewed them. They were all using both formulas.

What you need to understand about that formula above is that it is a bad estimate of something you probably don't want. Here are the issues:

Selection bias. Many deals (in some cases, most deals) are never marked Closed Lost. Even though they are. The metric will be optimistic. Think about it. If you never mark deals closed lost, then your win rate would be 100%.

If you consistently and rigorously scrub your pipeline of Closed Lost deals just before you measure win rate, and you measure over a long period (comparable to your sales cycle); then that formula is an estimate of the long-term win rate. The chance that a deal will ever be won. Why “ever?” Because there are no constraints on how old a deal is to be included in the measurement.

“Consistently and rigorously” means, applying the same judgement providing accurate predictions of each deal’s ultimate outcomes. That’s hard to do. You must mark every deal that will not be won as closed lost. Or consistently make the same errors. You can’t know and do that.

If you don’t measure over a long-enough period, and consistently and rigorously scrub lost deals, you will have a volatile metric. You can’t make meaningful comparisons across periods or groups.

If you are using it to make predictions or to allocate resources, you likely want a shorter term win rate. For instance, the chance that a deal will be won in say the next 90 days.

So what?

You can’t use this formula to predict short-term outcomes. It will be optimistic due to selection bias and because it's an estimate of the long-term win rate.

You probably can’t use it for long-term predictions because it will be wrong and volatile.

You can’t use it to make comparisons (period-to-period or group-to-group).

It's a bad estimate of something you probably don't want.

Including open deals in the measurement helps—a bit. But that formula has other problems. Skip that and use the time-to-event (TTE) method.

Want to learn more?

Here's why something's wrong if your forecasts are right.

Here's what we can learn about win rates from epidemiologists (the folks who invented the TTE method).

## Comments