Rainfall and Olympic pools

We can't imagine scale. Our ability to estimate anything large, complicated, or opaque is poor.

Area is often referred to 'as the number of football fields,' and volume is 'the number of Olympic pools.' Do you know the area or volume of either? No. Me neither.

As a side note, in looking up the number of Olympic pools in Sydney harbour (238,000), I came across that 'Sydharb' is a unit of volume of water in Australia - 562 gigalitres.

Last year, I helped someone build a shed (7m by 12m). During the week, it drizzled on and off most days. Only when we had the roof up but not installed the gutter could you see the water running off the shed. In this brief construction period, it was apparent how much water was falling over a relatively small area.

If you'd asked me to estimate the volume of water coming off the roof based on the rain I’d seen all week, I would have been off by more than I could imagine. I've spent considerable time in the rain and am generally good at estimation.

Did it matter that I was wrong with my estimate? No. The water would be stored in a tank, and we didn’t need to do anything with this information.

Whatever you're estimating should matter more, and you're probably no better than I am.

Next time you're in a meeting, and someone says, 'I think it's about', remember that a poor guess is worse than no guess.

If you're going to bother guessing, spend the time and effort to come up with a reasonable approximation.

Why is this hard to implement?

In the same way, it's hard to avoid sugar in your diet; it's hard to avoid poorly conceived guesses in business. They've become so commonplace that we've just accepted them as part of work.

The difficulty you'll face is identifying what's a poorly conceived guess and what's meaningful information.

There are times when you might need to guess. We know using 'calibrated experts' will produce a better outcome than just guessing—see Doug Hubbard for more on this approach. I’ve argued that when you need to use this method, you should consider alternative ways to find data.

If you read about the standard intake questions for a management consulting role, a common question is to estimate some obscure value - how many coffees are sold daily? Those conducting the interviews say, 'It's about the process, not the answer'. They want to see that you can grasp onto something and use this information to piece your way to an answer. In the case of coffee – the expected answer is that you'd identify the different types of places that sell coffee, estimate the number of each type of place, and then estimate the number of cups sold daily. You might go as far as to identify that you’d sell more coffee early in the morning than in the afternoon.

Multiple these all together, and there's your 'number'.

This approach sounds good, but it's more flawed than my estimate of water falling off the roof. The entire basis of statistical sampling is using approaches that avoid these errors. In the case of the ‘management consulting approach’ the very first decision leads to everything else you do being incorrect.

How could you do this better? Instead of guessing who sells coffee, you could sample people drinking coffee as a starting point. You’re now measuring what you’re interested in, cups of coffee, not the number of locations that sell coffee. At least you know you've avoided the first mistake. Another approach would be to call up the people that sell the takeaway cups, not those that sell coffee. Management consultants should be able to determine why measuring coffee sold would be wrong. Try going beyond the obvious.

If you want to know the capability of your analytics team - ask them how to estimate the number of cups of coffee sold and see which approach they take. When you want them trained to think in a way that doesn't involve poor guesses, let's do some work together.