Since the graph is updated daily, the bar for the current month will be very low at the start of the month and rise gradually until the end of the month.
We simply count the number of unique IPv4 addresses that connect to the Qubes update servers each month (except for Tor connections; see [below](#how-has-the-methodology-for-counting-tor-users-changed)).
### How has the methodology for counting Tor users changed?
Before, we simply counted the number of unique Tor exit node IPv4 addresses that connected to the Qubes update servers each month.
However, this underestimated the actual number of Tor users, since many Tor users can use the same exit node.
The new methodology is to estimate the number of Tor users as a proportion of the total number of *requests* from Tor exit nodes on the assumption that the proportion of users to requests is roughly the same for both clearnet and Tor users.
We cross-reference the list of connecting IP addresses with [TorDNSEL's exit lists](https://metrics.torproject.org/collector.html#type-tordnsel) in order to distinguish Tor and clearnet IPs and requests.
For this purpose, we count an IP address as belonging to a Tor exit node if there was a Tor exit node active for that address within the 24-hour periods before or after it connected to the Qubes update servers.
