The stats page shows published and in-progress translations. The in-progress translations includes:

• Translations that users just started and will become published soon.
• Translations that get published but remain as in-progress because of small changes.
• Translations that got abandoned, and will be never published.

The abandoned translations are the problematic ones since they prevent other users from translating and may have good work that is lost. So we may want which percentage of the in-progress translations are actually abandoned.

We could measure "abandoned translations" with a query such as "in-progress translations without activity in the last 30 days for which the target article does not exist". This needs to be based on some estimations (is a month enough to declare a translation not active?) but with reasonable estimates it may provide a good enough perspective on the nature of in-progress translations.