croissant
Home / Blog / Reporting

GA4 data retention: plan historical explorations before data expires

Croissant editorial team · · 5 min read

Plan agency reporting around GA4 retention limits, distinguish Explorations from standard reports, and handle expired historical detail responsibly.

A client asks for a year-over-year funnel analysis, but the GA4 Exploration cannot show the older period. Check the property's retention setting and the requested reporting surface before promising the comparison. Google says retention limits the user- and event-level data available to Explorations, while standard aggregated reports are not affected by that setting. [1] A standard report may still answer a narrower question, but it does not automatically replace the missing funnel detail.

Know which question needs historical detail

Start with the decision, metric, population and time window. A standard aggregate trend and a custom Exploration can answer different questions. For example, an aggregate count may describe total activity, while a funnel asks how people progressed through specified steps. Record the exact fields, filters and reporting surface before a client deadline.

Google documents that a report opened in Explorations excludes dates outside the property's user and event data retention period. Reports and Explorations also support different fields and filters, so a copied view may not preserve the same question. [2] Check the resulting settings rather than assuming two screens are interchangeable.

Check retention before agreeing the analysis

Ask an authorized property administrator to inspect the current setting and its implications for the planned date range. Google lists two- and fourteen-month choices for user-level data and key events; other event data has additional options for eligible 360 properties. Some property-size and data-category limits apply. [1] The available setting is not a promise that every requested field will remain available for the same period.

Google says increasing the period applies to data collected and not already deleted; it does not restore data already deleted. Reducing the period can trigger deletion in the next monthly process, and changes have a 24-hour delay. [1] Treat any change as a client-owned privacy and governance decision, not an automatic fix by a reporting workflow. Consider the purpose, applicable obligations, access and the client's retention policy before an authorized person changes a setting.

Worked example: the missing historical funnel

Imagine a fictional agency in November preparing a client's year-over-year funnel comparison. The current property has two-month retention, and the older user- and event-level data needed for the requested Exploration has already been deleted. The team can see an aggregate historical standard report, but that report does not contain the same funnel steps and filters. These are invented conditions for illustration, not evidence about a real property.

The analyst records the requested comparison as unavailable in that Exploration and keeps the old-period funnel conversion rate blank. They do not calculate a percentage from partial data or substitute the aggregate count as a funnel rate. An account lead may propose a separately labelled aggregate trend if it answers a useful narrower question, then ask the client to agree to the changed scope. The report should state what could not be reconstructed and why, if the cause has been confirmed.

A suitable note is: ‘The requested historical Exploration is unavailable because its required data has expired under the property's retention setting. The standard report below shows an aggregate trend and does not measure the same funnel.’ If the team has not confirmed the cause, say the detail is unavailable and identify the next check instead. The missing-data guide covers how to keep unavailable values distinct from zero.

Build a repeatable planning check

For each recurring report, keep a small register in an approved workspace: property, owner, reporting surface, required fields and filters, longest lookback, current retention setting, reviewer and next review date. Check the register when a client asks for a new quarterly or annual comparison and when a workflow changes. The owner should verify the setting in the actual property rather than rely on a template.

If a future analysis needs a longer lookback, discuss the setting before data expires. If an approved alternative archive or export is considered, document what it contains, its access controls, retention and costs, and whether it can answer the same question. An export is a separate system with its own limits and responsibilities; it is not a magic recovery button for already deleted data.

Test the reporting workflow with synthetic cases: a complete requested period, a period that crosses the retention boundary, an unknown setting, and a changed scope. The draft should carry the limitation into charts and narrative. A named reviewer checks the source view and approves the client-facing report before distribution.

Keep other causes and decisions separate

Retention is one possible reason for incomplete detail. Google also identifies differences in supported fields, filters, low-user thresholds, behavioral modeling and processing time between reports and Explorations. [2] Inspect the actual view and its quality information before assigning a cause. For sampling, use the separate sampled-report review; a sampled result and expired data require different responses.

A small agency can manage this with a reviewed register and calendar reminder. For a recurring multi-client workflow, the reporting demo provides context for collecting exceptions and routing drafts for review. Discuss a fictional example through the contact form; keep client exports and credentials out of the initial enquiry. A person retains responsibility for client communication, property settings and consequential decisions.

Implementation checklist

  • Name the decision, period, fields and reporting surface before promising a comparison.
  • Have an authorized owner check the property's retention setting and applicable limits.
  • Identify whether the requested historical data is still available.
  • Record missing detail as unavailable rather than zero or a partial rate.
  • Label any alternative report and get approval for the changed question.
  • Review privacy, access and cost before changing retention or adding an archive.
  • Test boundary and unknown-setting cases with synthetic data.
  • Have a person verify the source and approve the final client report.

The sources behind the sauce.

Primary research and official documentation behind the cited points. The workflow examples and checklists are Croissant’s own guidance.

  1. Google Analytics Help — Data retention
  2. Google Analytics Help — Data differences between reports and explorations