From Umami

Get the export

Umami exports events. Take the event dump rather than a report: it is one row per event, which is the shape that imports cleanly and keeps the hour-of-day detail.

Import it

Upload on the site’s Imports screen, or point the importer at a URL. The importer resolves columns through a table of aliases, so Umami’s naming is read without reshaping.

Like every import, it is idempotent by replacement: every row carries an import_id, so re-running replaces rather than duplicates and undoing is a delete, and resumable through a staging table, so a job that died at row 400,000 resumes at 400,001.

Vocabulary

Umami	Micaforge
Event	An event with kind custom and an event_name.
Event data	props, filterable as props.<key>
Session	A session, with the same 30-minute inactivity rule.
Visit	A session.
Referrer	referrer, resolved into referrer_host, channel and source.

Umami’s per-event data becomes props verbatim, so a property you were filtering on there is filterable here under the same name.

What changes

Bots. Umami filters them. Micaforge names them and counts them separately, so nothing disappears and nothing is folded into the human numbers.

Answer engines. A top-level channel here, rather than a referrer among referrers.

Web Vitals and performance. Collected by the tracker from the first pageview after install. Nothing to import; it starts from the day you install.

After the import

Recreate your goals (they are not in an event dump) and check a week where both tools have data. Pageviews and sessions should agree closely, because both count the same things the same way. Where they differ, it is almost always bot filtering, and Micaforge can now show you exactly which agents made the difference.

Then turn on the server log feed, which is the half neither tool had.