GA4 Retention May Have Eaten Your Black Friday 2025 Data

GA4 keeps event-level and user-level data for two months by default. The setting can be raised to 14 months for a standard property, but raising it does not restore rows already deleted. Black Friday 2025 ended ten months ago. If nobody changed the setting before late January 2026, the session-level detail from last year's peak has expired from Explorations, and no vendor can bring it back.

Octalysis calls this an Evanescent Opportunity: a thing that was available and quietly stopped being available, with the loss felt only afterwards. Most brands feel it in October, when someone tries to build last year's Cyber Week funnel in an Exploration and gets a thresholded, partial, or empty result.

What retention controls, and what it does not

The retention setting governs the data behind Explorations: funnels, path analysis, segment overlap, user explorer, cohort tables. Those need event-level rows, and event-level rows older than the window are deleted. The standard reports are built from aggregated tables and keep showing last year's sessions, revenue and channel groups regardless of the setting.

So the aggregated answer survives: how much revenue GA4 attributed to each default channel group on 28 November 2025. The new question does not: which sequence of channels the buyers who came back on Cyber Monday had touched in the two weeks before. If you did not ask it while the rows existed, the rows are gone.

The setting to check today, and the date to remember

Admin, Data settings, Data retention. Two options for a standard property: 2 months or 14 months. Set it to 14 months now and the events from Cyber Week 2026 stay available until late January 2028, which covers the whole of next year's planning window and the Black Friday 2027 comparison. It cannot recover anything already deleted.

The Price of Being Found notes a related finding from its publisher's own warehouse: the analytics history available for the coverage analysis was eleven months, which the authors treat as a finding about how long anybody's measurement memory actually is. Two months is shorter than that.

The file that does not expire

There is one artefact of last year's peak that no retention setting touches: an export you made. A GA4 export to CSV, or the raw event tables in BigQuery if the link was on, lives in your own storage under your own rules. You don't own your attribution data makes the wider case; this is the narrow, dated version of it. The export is also the input to a causal attribution read, which estimates per-channel incremental contribution with intervals from the data you hold rather than from the platform's memory of it.

If you exported Cyber Week 2025 at the time, you still have the option the retention setting removed. If you did not, export Cyber Week 2026 in the first week of December, before any of this repeats.

Why this is a Black Friday problem specifically

Peak weeks are when the interesting questions live. The rest of the year, the difference between a 2-month and a 14-month window is academic; nobody rebuilds a quiet March funnel in October. Cyber Week is the one period a brand wants to interrogate a year later, with a new hypothesis, and it is precisely the period that a default setting makes unavailable exactly when the next one is being planned.

What to do this week

• If you own the budget: open the retention setting and read it. If it says 2 months, change it, and write the date down somewhere that survives staff changes.
• If you have to defend the number: check whether a BigQuery link or a dated CSV export of November 2025 exists anywhere in the company. If it does, it is your only route to a like-for-like Cyber Week comparison, and the input to the read you would otherwise be estimating from memory.

The Black Friday 2026 calendar lists the other three deadlines. Two of them are still open.

GA4 retention options as documented for standard properties on 9 September 2026. The eleven-month warehouse observation is from The Price of Being Found (Edition 2.10), Chapter 8.