# Read the activity log and export it

## Purpose
Find out who did what in the account, check whether the system's own messages went out,
and take a copy away as a spreadsheet.

## Who this is for
Account administrators, and anyone holding account-log, campaign-log or communication
permissions.

## Where this lives

Sidebar **Activity Logs**, titled **Activity log**: who did what in this account,
newest first.

Platform administrators also see a **Message delivery** section under the log: the
notification pipeline, dead-letter recovery and the in-app notification inbox. Account
members, whatever their role, do not. The section says so to the administrators who
see it.

## The audit trail

The main table lists, newest first:

| Column | What it holds |
| --- | --- |
| When | The day it happened, with the time underneath |
| Who | Who did it, by name where the server could resolve one; otherwise by what they are: "The system", "A respondent", "A team member", "A platform administrator", "An API integration", "The platform team" (a platform script or set-up kit), or "Someone outside this account". "Unknown user" is left only for an id nothing could classify. The raw id is in the tooltip |
| What happened | A sentence: the action and what it was done to ("Created a role assignment or invitation", "Locked a team member's access"). Under it, **Technical details** opens the stored action, the record's type and id, the actor's id and the names of the changed fields |
| Changes | How many fields changed. A count, not a diff: there is no before/after viewer to open |

Three controls narrow it: a search box (**Search by person, event or record id…**, which
matches the name, the sentence and the ids, so an id pasted from a support thread still
finds its row), an **All actions** dropdown listing the actions actually present, and an
**All kinds of record** dropdown listing the kinds of record. The count on the right reads
"*n* events logged" and follows your filters.

The page loads the 100 most recent entries. The export is not limited to what is on
screen (see below).

## Export to CSV

Choose **Export CSV** in the header.

- The download **honours the search term** currently in the audit-log search box, so
  filter first if you want a subset.
- It streams the full filtered set from the server, capped at 1000 rows, not the 100
  the page is displaying.
- The file is named from the server's own content-disposition header, falling back to
  `activity_log_<accountId>.csv`.
- On success the page confirms "Exported activity log to CSV." Failures report the
  server's reason in the same place.

The **All Actions** dropdown filters the on-screen table only; it does not narrow the
export. Use the search box for that.

## Message delivery (platform administrators)

This section is shown to platform administrators only.

Five counters (**Waiting to send**, **Failed**, **Retried events**, **Total retries**,
**Dead-lettered**) sit above **Notification events**, the lifecycle notifications queued
by campaign, assessment and registration activity. Each reads as a sentence ("Who can
take an assessment changed"), with the campaign it belongs to by name and how delivery
went ("Waiting to send", "Sent", "Failed"); the event code, the record, the channel and
the actor's id are under **Technical details**. **Show events for** narrows the events to
one campaign, chosen by name; it does not touch the audit log above.

Two buttons in the section header act on the account's queue:

- **Send waiting messages now**: pushes the queued backlog through now instead of
  waiting for the scheduled run.
- **Retry failed messages**: re-attempts the ones that failed.

Events that exhausted their retries appear under **Dead-letter recovery** with their last
error. Tick the ones worth another attempt and re-queue them. Treat a growing
dead-letter list as a signal to look at the underlying cause rather than as a queue to
keep draining.

**In-app notifications** lists notifications delivered inside the product, with
read/unread tracking. Mark one read, or mark all read.

In-app notifications go to the people who run the account: members holding a role that
gives them some permission here. A member who is only a respondent (an invited
participant, whose role grants nothing in the console) and a reviewer restricted to one
assessment receive none, and their inbox is refused, because the account's events name
other respondents, their uploads and the published results.

## Limits and gotchas
- Retention is not configurable from the interface. There is no log-policy screen.
- The actor is shown by name where the server could resolve one, with the raw id in the
  tooltip; the CSV carries both, as `userId` and `actorName`. Entity targets stay as ids
  in both.
- A campaign's own log (reached from **Logs** in the campaign header) is titled
  **Campaign Audit Trail** and carries that campaign's audit trail (and, for platform
  administrators, its message delivery). What lives only here is the account-wide view:
  every campaign, and the account-level actions that belong to no campaign at all.

## Related pages
- [Review campaign analytics and logs](/operate/tenant-admin/review-campaign-analytics-and-logs)
- [Invite members and assign roles](/operate/tenant-admin/invite-members-and-assign-roles)
