Two Simple Tricks for Making Better Estimates

by Richard Harbridge on February 21, 2011


Douglas Hubbard is someone I respect immensely. He has an amazing book called “How to Measure Anything: Finding the Value of ‘Intangibles’ in Business” that I very highly recommend reading. In fact here is a link to that very book on amazon (buy it). What follows is some advice and simple tricks that can help anyone make better estimates regardless of their role or job.

Regardless of your role in your organization you are often requested to make estimates on how long something will take you. Often this is in a highly pressured situation where you don’t have time to do lots of research, run ‘complex’ math (which I love), and account adequately for risk. (Should you do all those things and more? Yes, but let’s be realistic it doesn’t always work under the constraints of every situation.)

The first tip I will give is to never ever give a single number. As an example when asked how long it will take to make ‘report A’ let the requester know a range that you feel comfortable with. Why a range? Realistically in a quick situation like that if you don’t give a range it won’t indicate how uncertain you are about the estimate.

If you were thinking about the request and felt like the task would take somewhere between 2-8 hours then communicating only a single value guess such as ‘6 hours’ can be misleading as it doesn’t include any information about how confident you are.

So the first thing we need to agree on is that quickly estimating a ‘single number’ (without math/careful consideration) typically leads to poor results in both setting the right expectation and being accurate.

Now onto the difficulties of even coming up with a range of possibilities. One of the hardest parts of estimating is coming up with a range you feel really confident in.

Basically when you give an estimate you should give a range that you feel 90% confident that the real value will fall within that range (90% is the suggested optimal confidence level due to the effort involved in getting more than 90%).

You don’t have to be perfect but you have to feel willing to bet on it.

In other words if you consider giving a range of 2-8 hours on a task as an estimate you should confirm that you are actually 90% confident that the total time will fall between that range. There is a trick (one of many) that Douglas mentions in his book that is easy to remember and use for testing whether you are actually close to 90% confident about an estimate.

Imagine that you win $2000 in one of two ways:

A) You will win $2000 if the true time it takes turns out to be between the upper and lower bounds you provided. If not then you win nothing.

B) You draw a M&M at random from a bag of 9 red M&M’s and 1 blue M&M. If the M&M is red you win $2000. If it is blue you win nothing.

Which option would you take? If you choose B) (which statistically most people do) then it means you might not be 90% confident but actually less confident (say 80%, 60% etc). If you choose A) it’s also not really what we want because it means you are probably over confident (especially if you felt strongly geared toward A). So you adjust the bounds (upper and lower) until you find a place where you feel indifferent between option A) and B) – that is ‘probably’ your 90% confidence level for this estimate.

It’s not as complicated (or in some ways as effective) as many other methods you can use to calibrate how confident you are about your estimates but it works and is easy to remember and start implementing immediately.

As individuals we should always be responsible for our estimates. By using simple tricks we can improve our estimate accuracy and effectiveness.

The two tricks we discussed are:

  1. Always give a range estimate (never a single value).
  2. Think carefully about your range estimate – are you 90% confident in that estimate?

I began exploring these methods a few years ago and slowly explored more and more advanced topics and scenarios around estimates (there are many). It felt like a good idea to share one of the first things that really stuck with me and that I still use (with slight variations) today.

Hope this helps,
Richard

VN:F [1.9.22_1171]
Rating: +3 (from 3 votes)
Print Friendly, PDF & Email

Previous post:

Next post: