Improving performance ratings of taxi drivers

By MUNGAI KIHANYA

The Sunday Nation

Nairobi,

23 May 2021

I took a ride in a “digital taxi” recently and had an interesting conversation with the driver. It was about newcomers in the business and how unknowledgeable they are about finding their way around the city even though the Smartphone Apps have maps that show the quickest way.

My driver advised me that it is easy to know a newcomer: they will usually have a very high rating! “If you have been in the business for long, there is no way you will have an average score near 5.”

This counter-intuitive conclusion is a good illustration of a situation where the mean value can be misleading. A newly enrolled taxi driver could have a high mean score simply because there isn’t large enough data to work with.

As time goes by and more customers rate the drive, a proper pattern is established and the mean value remains relatively stable. To understand this, suppose the driver has carried 10 customers and 5 gave him a score of 4 while the others gave 5. The mean is (20+25)/10 = 45/10 = 4.5.

If the next customer – the 11th – gives a score of 3, the new mean comes to (20+25+3)/11 = 48/3 = 4.36; or, if we round it up, approximately 4.5 One low score has made a noticeably significant change in the mean.

Now compare with another driver who has carried 100 customers and suppose half od them gave him a score of 4 and the other half 5. The mean, again, comes to (200+250)/100 = 450/100 = 4.5.

If the next customer (is it the 101th or 101st?) gives a score of 3, the new mean for the more experienced driver will be (200+250+1)/101 = 451/101 = 4.47. Rounding this up gives 4.5. The single low score does not make noticeable difference in the average.

Clearly, then, for an experienced driver who has carried many customers, the mean score hardly ever changes! Thus, we may ask whether this scoring system is really helpful to customers. I don’t think it is.

The calculation can be improved by restricting the number trips included; for example, to the last 50 trips. This way, the mean value will change continuously depending on how the driver conducts himself.

If he improves his overall conduct, his mean score will go up and, when he slackens, it will go down. This puts just the right amount of continuous pressure for the driver to perform at his best – all the time.