In fact, our studies have shown that the average individual’s usage changes enough within 12 months, that any bill older than 3 months is unlikely to represent their future usage and may well skew our forecasted bill cost. This is why we have made a decision to focus on the most recent 3 bills for any given user, to get the most accurate forecast that matches their latest usage.

However, for a minority of individuals, this will work less well. Examples where this may be the case are:
a) Users who have used their phone abroad in the last 3 months but do not travel regularly – these users may find predicted bill costs include a high likelihood of roaming recurring, when this may not be the case.
b) Users whose mobile phone use may be “seasonal” (eg. roaming the same time each year), where an annual trend might be useful

We’re currently considering ways to allow people to include or exclude billing periods in their recommendation but we haven’t hit on anything that we’re happy with yet. Rest assured, we are fully aware of the issues some users face in this area and are working on how to fix it.