One Hundred Days to the UA Sunset: The Triage Edition

One hundred days from the July 1 shutoff, the industry’s migration-lag surveys and the anecdotes agree on an uncomfortable fact. A startling share of Universal Analytics properties, the estimates run from a third to well over half, remain effectively untouched. No GA4 property, or a property auto-created and never configured, which for continuity purposes is the same thing.

Last March’s post was a boarding call with fifteen months of runway and a full work list. This one is triage, and the tone changes accordingly. The sequence below is ordered by irreversibility. With a hundred days left, the question is no longer ‘what is the ideal move.’ It is ‘which losses last forever, and in what order do we stop them?’

The sites that dual-tagged in 2020 are configured and historied. They can skip ahead to the side-doors section. That advice is now on its third annual repetition, with results to show. Everyone else: the list, in the order the calendar enforces, starting with the item that should be done before you finish reading this.

The triage sequence, by irreversibility

First, today, before anything else: GA4 collection live on every property. The reasoning has not changed in three years and is now merely urgent. The models do not reconcile. Compare-to history exists only if collected. History you never recorded is the only loss you can never undo in this entire move. Every other item below can be done late, badly, and then redone. This one either happened or it did not. The minimal version is an afternoon. The GA4 property created, the tag deployed through the governed container, default events flowing. Its completion this week is the difference between entering the cutover with a quarter of history and entering it blind.

Second: conversion and audience reconstruction for active campaigns. The Ads linkages wired to UA (conversion imports, audiences, bidding signals) break silently at the sunset. Campaigns keep spending. Bidding gets worse against dying signals. Paid teams discover the failure in performance data, not in alerts. The rebuild belongs to April, with the check pass before the June spend cycles. GA4 conversions defined on purpose, not auto-migrated. Links re-made. Bidding checked against the new signals.

Third: the history-export call, with honest expectations. UA properties go read-only in July and are slated for deletion after. So the export question has a real deadline. The honest sort: the data-heavy sites with warehouse habits take the BigQuery path. Structured, queryable, genuinely useful later. Most teams should instead pull the reports they actually reuse into tidy archives, scoped by what future reporting will really ask. The annual summaries, the channel baselines, the campaign benchmarks. The drawer-of-CSVs export that tries to save everything preserves nothing usefully.

Fourth: the keep-the-reports-making-sense memo. The one-page brief to every dashboard reader. The model changed, the numbers will not reconcile, here is what we compare instead, here is the break annotation. Sent around before July, it spares the analytics team a summer of executive confusion, and it is the cheapest trust insurance in the whole move.

The two legitimate side-doors

• The GA-alternatives look, while the wiring is open anyway: the privacy-first analytics tier, the Matomo and Plausible class, has grown into a real option. Simpler models. First-party designs. EU-friendly designs. It fits sites whose needs are honest page-and-event counts rather than the ad-wired depth GA4 assumes. A forced migration is the one moment the switching cost is already sunk. The look belongs in the project for any team whose privacy stance, consent burden, or complexity budget makes ‘simpler and owned’ appealing.

• Consent architecture, cheapest to fix mid-migration: the tag layer is open and the events are being redefined. The consent-scoping work, collection conditioned on consent state, costs least when done inside this project rather than as its own. Sites that bolt consent on after the migration will pay for the same surfaces twice.

The framing that salvages strategy from the scramble

The paragraph for the teams arriving at this late and annoyed, because they deserve more than a scolding. Forced replatforming is also the moment of maximum permission. The organizations that use it well convert a compliance scramble into the measurement rethink they were never otherwise going to schedule.

The questions the migration forces are the ones a mature measurement practice wanted asked anyway. What actions actually matter. What names the events deserve. Who reads which reports and why. What gets collected under what consent. The UA-era configuration being abandoned was, at most sites, a decade of buildup nobody would rebuild on purpose. The GA4 property being configured this month can be the deliberate version, at the modest extra cost of running the taxonomy conversation while the tags are open.

That is the salvage available even now, at a hundred days. Not the ideal migration, that ship boarded in 2020, but a deliberate one, executed under deadline with the strategy questions asked rather than deferred. It beats the alternative on offer. The auto-made, never-set-up property that exists on paper and answers nothing. Found out as such during the first post-sunset reporting cycle, by exactly the executives the memo was never written for.

Final boarding call

The market note, delivered without pleasure. The consultancy pricing shifts next quarter, per the scarcity dynamics we warned about last March. The sites that convert this month’s list into completed work will spend that quarter watching the market from the completed side.

The item-by-item version, in brief for the meeting this post should trigger. Collection today, the afternoon that prevents the permanent loss. Conversions and linkages by April, the silent breakage with the spend attached. Exports decided by May, scoped by realistic reuse. The continuity memo by June, the trust insurance. And the side-doors weighed inside the same project window (the alternatives, the consent plumbing) wherever the needs or the privacy stance argue for them.

One hundred days is enough for all of it, done deliberately, starting now. It is not enough for any of it deferred another quarter. We have covered this migration since 2019, so nobody can locate the surprise. July 1 does not move. The afternoon is available today. This is the final call.