My phone has a built-in ‘smart’ battery use monitoring system. Supposedly this helps to optimise the device somehow. The most visible way this works is that it will slowdown charging rates during what it determines ‘night time’ to be, in order to extend the overall battery lifetime. It doesn’t use the system clock for this, it tries to be smart and guesses based on device use when ‘night time’ is.

I don’t use my phone very much during the day. I prefer to do work using a PC. So my phone mostly stays in my pocket expect for the odd occasion where I need to use a cartography app. However, I frequently use my phone at night, to listen to things before going to bed, like podcasts or music.

You can probably see where this is going. With the abundance of data on how I use my device, it has come to the entirely incorrect assumption that night time is during the day, and daytime charging should be ‘optimised’ and slow.

This is an excellent and simple lesson for software development and data analysis. Just because you have a wealth of data, for example how a device or application is being used, does not mean that you have the full picture.

In my phone’s case, it is easy enough to disable the ‘optimised’ charge scheduler. That won’t always be the case on every device or in every application. Some takeaways:

1. Developers should always allow the end user to override anything a system might do to try and ‘help’
2. Data analysts should always consider how their data, or their interpretation, might be biased