Migrating off Google Analytics 4 without losing the plot

Most GA4 migration guides are written by the people you are migrating to, which makes them optimistic about how much of your setup transfers. This one is too — but the parts below that will cost you a weekend are marked as such, because discovering them mid-migration is worse than knowing up front.

Your numbers will not match. That is normal.

Every migration produces a moment of alarm when the new tool reports fewer sessions than GA4 did, or more. Before you debug anything, know that an exact match is not achievable and not the goal — the two tools define their terms differently.

• Sessions. GA4 restarts a session after 30 minutes of inactivity by default, and also on a new campaign source mid-visit. Tools that do not do the campaign-restart part will report fewer sessions from identical traffic.
• Users. GA4 counts a cookie-backed identity that persists for months. A cookieless tool counts a daily-rotating hash, which means a person visiting on Monday and Thursday is two visitors, not one. Expect a higher user count and a lower "returning" rate.
• Bounce rate. GA4 does not have one in the classic sense — it reports engagement rate and derives the inverse. If your new tool uses the traditional definition, the two are not comparable at all.
• Sampling. GA4 samples large or complex reports. Alternatives generally do not. On a high-traffic site the new numbers may be more accurate while looking wrong.

The useful check is not whether the totals match. It is whether the two tools agree on the shape — same top pages in the same order, same traffic sources in the same proportion, same daily rhythm. If the shape agrees and only the absolute numbers differ, your setup is fine.

Historical data does not come with you

Nothing exports your GA4 history into another tool in a form that tool can report on. This is the genuinely painful part and no vendor can fix it, because the underlying models differ too much for the transfer to mean anything.

Three realistic options, in order of how much they preserve:

• BigQuery export, if you already have it. Raw event-level data you own outright, queryable forever. If you enabled this at any point, you have real history.
• Scheduled CSV exports of the reports you actually use. Not the whole property — the eight or ten reports anyone opens. This takes an afternoon and covers most of what people miss later.
• Screenshots of your annual summaries. Unserious sounding, entirely sufficient for the once-a-year question of what traffic looked like in 2024.

bigquery.sql

-- If you have the BigQuery export, your raw events are already
-- outside GA4 and this is the only history that survives properly.
SELECT
  event_date,
  event_name,
  COUNT(*) AS events,
  COUNT(DISTINCT user_pseudo_id) AS visitors
FROM `project.analytics_123456789.events_*`
WHERE _TABLE_SUFFIX BETWEEN '20240101' AND '20241231'
GROUP BY event_date, event_name
ORDER BY event_date;

Do this before you delete the property. Google retains GA4 data for a maximum of 14 months by default, and a deleted property is not recoverable.

Re-map the events by hand

GA4's event model — an event name plus up to 25 arbitrary parameters, with nested arrays for commerce — is more elaborate than most alternatives. Automatic conversion tends to produce a flattened mess of unused fields.

tracking.ts

// GA4: every interaction is an "event" with up to 25 parameters.
gtag("event", "purchase", {
  transaction_id: "T-1234",
  value: 49.0,
  currency: "USD",
  items: [{ item_id: "pro", item_name: "Pro plan" }],
});

// Most alternatives: a named goal plus a small, flat property bag.
// The nesting is what does not survive — flatten it deliberately
// rather than letting a migration tool guess.
track("purchase", {
  transactionId: "T-1234",
  valueCents: 4900,
  currency: "USD",
  plan: "pro",
});

Take the opportunity. In practice most GA4 properties have accumulated dozens of events, of which perhaps six are ever looked at. List the ones that have informed a decision in the past year, port those, and leave the rest behind.

Conversions become goals

GA4 marks an existing event as a conversion. Most alternatives define a goal as a URL pattern or a named event. URL-based goals are usually simpler and less fragile — a thank-you page path does not change when someone refactors the frontend, whereas an event name in a click handler does.

Running both at once

Two analytics scripts on one page is not a problem. They do not share state, and the second one costs a few kilobytes. Keep the overlap long enough to cover a full business cycle including a weekend — a week minimum, a month if your traffic is seasonal.

layout.html

<!-- Run both for one full reporting cycle before you switch.
     Two trackers on one page is fine; they do not interfere. -->
<script async src="https://www.googletagmanager.com/gtag/js?id=G-XXXX"></script>
<script defer src="https://analytics.example.com/tracker.js"
        data-site="abc123"></script>

During the overlap, check the shape agreement described above weekly. If a specific page or source disagrees badly between tools, that is worth investigating before you commit; a uniform offset across everything is not.

What you can turn off afterwards

If you move to a cookieless tool, the consent banner GA4 required may no longer be necessary — and removing it typically recovers more measured traffic than the migration itself costs, because everyone who would have clicked "reject" is now counted.

Confirm it applies to your full stack before removing anything: the banner covers every cookie on the site, not just the analytics one. If you still run ad pixels or a session recorder, the banner stays. See do you need a cookie banner for analytics for the specifics.

The order to do it in

• Export the history you care about, while the property still exists.
• List the events that have actually informed a decision.
• Install the new tracker alongside GA4. Change nothing else.
• Wait one full reporting cycle.
• Compare shape, not totals. Investigate disagreements per-page.
• Re-create the six goals that matter. Not all thirty.
• Point your dashboards and reports at the new source.
• Remove the GA4 tag. Keep the property read-only for a quarter.
• Only then, revisit the consent banner.