Which is better: An estimate of “roughly 2 weeks” or an estimate of “4.75 days”? Personally I favour accuracy over precision. “4.75” is attractive because it very precise and a smaller number, but “roughly 2 weeks” might be more accurate. And estimating must be accurate to be any use for planning.

Defining accuracy and precision

In common parlance “accuracy” and “precision” are used interchangeably but in the scientific world they are different. Accurate means the measure is correct. Precise means the measure is consistent with other measurements. Of course the ideal is where a measurement is both accurate and precise. The nightmare scenario is when your measurement is both inaccurate and imprecise.

Accuracy vs Precision (Source: Hotdesign)

Accurate Estimation

Accuracy: The closeness of a given measurement to its true value, i.e. whether the measure is correct.

If I have to choose between accuracy and precision then I’ll go with accuracy. If our example item really does take “roughly 2 weeks” then the estimate is accurate. If most of the estimates are accurate then I can work with them and put together a schedule, which in my world means a Release Plan.

Precise Estimation

Precision: The stability of that measurement when repeated many times, i.e. whether the measurement is consistent with other measurements.

“4.75 days” is a very precise estimate. The number of significant figures increases with the precision of a measurement (NDT: Significant Figures (Digits)). So a high number of significant figures (Significant digits) implies a high precision. In this case the estimate suggests we’ve built up an estimation process that allows us to estimate items precisely to the quarter day. Usually only possible by collecting and comparing a lot of estimates.

The two significant places in our “4.75 days” estimate says nothing about the accuracy of the measurement. Who cares about that last 0.75 days of the estimate if the whole estimate is out by 10 days!

In contrast “Roughly 2 weeks” is not at all precise. Anything between 10 and 20 days might be “roughly 2 weeks”. That might bother other people but not me. I don’t really care whether a particular item takes 10 days or 20. I care more that the average for all “roughly 2 weeks” items is 14 days.

I’m also conscience that a precise estimate often comes from a longer estimation process. The time invested in estimation is time people are not building stuff. I’d prefer they were developing so try to minimise estimation effort.

A rough but accurate estimate is good enough for me.

References

NDT Resource Center. (n.d.). NDT: Significant Figures (Digits). Author.

Wikipedia. (n.d.). Accuracy and Precision. Author.