The data was collected from September 24 to October 29 by irregularly making a screenshot of the 'Usage' tab in the iOS preferences. I have a second hand iPhone 3GS of about 2.5 years old. The data I stored was Usage time, Standby time and battery percentage. On iOS, 'Usage' time indicates the time when the phone was used or actively doing something (such as syncing mail etc.), 'standby' time is the total time that the phone has been turned on since the last full charge.

Initially, I disabled 3G on my phone as I didn't have a dataplan yet, later I got a dataplan and turned 3G on and even later I upgraded from iOS 4 to iOS 5. Therefore, the data is separated into three subsets:

  • Set 0: All data (97 datapoints)
  • Set 1: Data with iOS 4 and 2G (63)
  • Set 2: Data with iOS 4 and 3G (15)
  • Set 3: Data with iOS 5 and 3G (18)

The only data missing therefore is iOS 5 with 2G, since I never turned 3G off again after I got a data plan.


For each dataset above, I plot and fit the data with the model

battery = 100 - A*time

because at time=0 the battery should be 100% by definition.

Since I use my phone more or less intensively on different days, I expect to get the average usage and standby times, including some measure for the uncertainty in this data given by the standard devitation in A.


The results are as follows:

  • Set 0 (All): Usage: 6.67 ± 0.591 hrs Standy: 26.2 ± 2.38 hrs
  • Set 1 (iOS4/2G): Usage: 8.02 ± 0.372 hrs Standy: 28.1 ± 3.08 hrs
  • Set 2 (iOS4/3G): Usage: 6.52 ± 0.424 hrs Standy: 22.1 ± 3.65 hrs
  • Set 3 (iOS5/3G): Usage: 4.49 ± 0.464 hrs Standy: 25.4 ± 5.89 hrs

And some plots of this data here:


It seems that battery usage life has gone down from iOS 4 2G to iOS4 3G to iOS5 3G, although this does not imply that the software is less energy efficient. Over time I also started using my phone more, so this is a mix of both my usage pattern and the actual battery life. Therefore I cannot draw any conclusions based on this data, except for that I definitely need to charge my phone every day.


The files used for this analysis (both data and Python script) are available on this gist, the most important files being: