💡 On forecasts & prediction error
Mobile data stats dive
One of my favourite annuals is the Ericsson Mobility Report. The telecoms infrastructure firm has been producing this report since 2011. The earliest version I could find online was the 2014 report1, so I snuck back and compared their predictions for 20202 to reality.
In short, Ericsson did pretty well…. except where they didn’t.
The table shows their 2020 predictions from 2014 compared to 2020’s actual data. If they were within 30% of their forecast, I’ve scored them as “accurate”. Worse than that as inaccurate. The green prediction errors are where they overestimated the change, the red is where the underestimated.
If you look at the forecasts, there seems to be a single dominant source of error: the impact of the smartphone. Start with subscriptions. Ericsson wildly overestimated the growth of mobile subscriptions, reckoning that by 2020, the average human would have 1.25 mobile phone subscriptions, and more than one mobile broadband subscription.
The firm was cruise-missile accurate w…