What we log when automation touches member data

When a person moderates a post, you can ask them why. When a script does it, you need a log. After a couple of "who changed this?" mysteries on our network, I sat down and wrote out what we record every time an automation touches member data.

One line per action

Each entry is short and dull by design:

• when it happened, in UTC
• which bot profile did it
• what triggered it: a schedule, an incoming webhook or a member's action
• what changed, with the old and new value
• whether it worked, and the error if it didn't

We don't log the content of private messages or anything a member wouldn't expect us to keep. The log answers "what happened and why", not "what did people say".

Questions it has already answered

In the first month: why a member's event RSVP disappeared (a cleanup job with the wrong date filter), why the weekly digest went out twice on a Monday (a server restart in the middle of a scheduled run), and why one group's cover photo kept resetting (an old integration nobody remembered switching on). Each took minutes to find instead of an evening.

Keep it short-lived

We keep 90 days of entries and then delete them. That's long enough to look into a complaint and short enough that the log doesn't turn into one more pile of personal data to protect.

If you run a community of any size with automations, I'd start here before adding anything smarter. Being able to explain what happened is worth more than another clever feature.