We evaluate journal cost effectiveness by a metric we call Net Cost Per Paid Use (NCPPU).
NCPPU is similar to the more familiar Cost Per Use (CPU), but different in what we consider some very important ways. We'll describe the differences here by braking apart the phrase, starting with the end first:
Use is traditionally considered just downloads. This is great, but should also include other kinds of value, like citations to the journal and authorships in the journal. NCPPU includes this kind of Usage, as described more in the Impact documentation.
Don't pay for free! When articles are available for free, usage of those articles shouldn't be considered a cost. Therefore NCPPU excludes all Usage to free content, such as articles available as Open Access, your Backfile, and (configurably) articles available via Academic Social Networks. More details in the fulfillment documentation and open access documentation.
The cost in NCPPU isn't just the subscription cost, because the alternative to subscribing doesn't cost $0. When libraries don't subscribe to a journal they still offer access to the journal via ILL, and this has a cost for the library. So we consider the cost to be the difference between the projected subscription cost and the projected ILL cost, were the library to not subscribe. Details are available in the subscription costs documentation.
So that's what NCPPU is! The net cost of subscription minus ILL, divided by holistic usage to articles which aren't available for free.
It's fun to note that NCPPU can actually be negative! If the subscription cost is relatively low, and the Usage is relatively high, the estimated cost of ILL can be higher than the cost of subscribing to a journal!