[tracker1]

I would think if it rotated on the first of each month, that would probably be sufficient... then you could get your counts for any given month (excluding first/last day) assuming most system check every week or two at least, and it would be pretty consistent.

[chii]

Rotating the identifier means you lose the information about attrition rate.

If you have some number of users leaving, but a similar number incoming, then it would look like you have a consistent usage. Losing the info about lost users means you don't improve in retention.

[nikajon_es]

Could you regain this info by adding a static prefix to the rolled id? So you know it was rolled, but not from which previous id. Where as new id's would have no prefix, so you can count new users as new.

[tracker1]

Could just be the date of install in UTC as a prefix, the other part randomized on the first of the month... they could still calculate relative drop off, and still get better stats more anonymously.