Time For Training

This post is inspired by Simon Guilfoyle and his excellent blog on train targets and performance measures. If you haven’t read it, go and do so before coming back to read this! You can find Simon’s post here: Bad Performance Measurement on Tour

Two trains in stationRead it? Good. Now we can carry on.

Simon’s blog showed that the management of one train operator believes that anything up to 5 minutes late is actually ‘on time’ and the management of the other is happy to count anything up to 10 minutes late as ‘on time’.

How late does a train have to be before you consider it late? I don’t know about you, but if my connecting train leaves 9 minutes after my scheduled arrival time, then 10 minutes late doesn’t really feel like ‘on time’.  Still, at least I know that management will consider the service a success!

Anyway, back to my main point. Consider yourself a passenger who has to choose between catching these two trains…

The one on the left runs an average of 5 minutes late (or ‘on time’ in management speak), whilst the bad boy on the right staggers to its’ destination a poor 7 minutes late on average. Any sane person would choose the left train and expect to rock up a couple of minutes earlier, wouldn’t they?

Consider the below chart for the train that averages 7 minutes late:

More stable; low variation chart

The chart shows the lateness of each individual journey over time and just look at that awful performance! It’s ALWAYS 7 minutes late and NEVER hits management target of less than 5 minutes late. 100% failure = franchise lost (unless the tender is faulty, but that’s a whole different blog!) The train is predictably 7 minutes late (the average is 7.1) each and every day.

I think we’ve picked right here, but let’s just take a look at the detailed figures for the other train that averages 5 minutes late;

Less stable; highly variable chart

Well that’s great for management; the train is often less than 5 minutes late and often only 1 minute late; targets met and franchise safe!

But hang on a minute, how predictable is the performance. How many times is this train more than 10 minutes late? At its’ worst, it’s 29 minutes late. The average (mean) is 5.2 minutes late, but it will be up to 17 minutes late predictably. Ouch! As Simon so rightly points out, averages do not tell you the true picture and only looking at data in time series will give you useful measures of performance.

So let me ask again; which train would you catch?

I know which one I would go for and when my connecting train leaves 9 minutes after my predictably 7 minute late first train, I’ll look back with just a hint of smugness at all you people who believed the published performance measures and were regularly left on the platform after picking the ‘faster’ train.

Photo credit: Flickr user Gene Hunt

One comment on “Time For Training

  1. Pingback: (Not So) Bad Performance Measurement On Tour (#3) | InspGuilfoyle

Leave a Reply

Your email address will not be published. Required fields are marked *


HTML tags are not allowed.