GA4 Data Retention, Sampling & Annotations

GA4 Data Retention: The Setting Most Teams Never Touch

Two weeks ago I opened a GA4 exploration for a client who wanted to compare Q1 and Q2 funnel performance. The exploration loaded, but only the last eight weeks of data appeared. Everything before that was gone. Not hidden -- deleted. The ga4 data retention default of two months had been in place since they migrated from Universal Analytics in mid-2023, and nobody had changed it. Three years of event-level data, wiped on a rolling basis.

This is not a rare case. I see it in roughly half the properties I audit. The retention period defaults to two months for both event-level and user-level data. Unless someone deliberately switches it, your explorations, funnel analyses, and segment comparisons are built on a shrinking window that never reaches far enough back to answer the questions that matter.

Data retention, sampling limits, and annotations share a trait: they are easy to configure and do nothing until the day you need them. By then the damage is done.

How GA4 Data Retention Actually Works

GA4 stores two categories of data, and the retention rules differ for each.

Aggregated data powers the standard reports you see under the Reports tab -- Traffic Acquisition, Engagement, Monetization. This data is not subject to the retention setting. It stays indefinitely regardless of what you configure (GA4 data retention documentation).

Event-level and user-level data powers everything in the Explore tab -- funnel explorations, path analyses, segment overlaps, cohort reports. This is the data that the retention settings control. When it expires, those explorations return empty or partial results. No warning, no error banner. The report simply renders with whatever data remains.

Your two options (free properties)

Setting	Retention period	What it covers
Default	2 months	Event + user-level data
Maximum (standard)	14 months	Event-level + user-level data

If you are on Analytics 360, event-level data can be retained for 26, 38, or 50 months. User-level data caps at 14 months even on 360 (GA4 data retention documentation).

How to change it

Admin > Data collection and modification > Data retention. Switch "Event data retention" from 2 months to 14 months. Do the same for "User data retention." You need at least Editor permissions.

One critical detail: the change is not retroactive. If you ran at two months for a year and then switch to 14 months, you do not get the previous ten months back. Data that has already expired is gone. This is why I flag it as the single highest-priority setting in every GA4 tracking audit.

The "Reset user data on new activity" toggle

Below the retention dropdown, there is a toggle labeled "Reset user data on new activity." When enabled, the clock restarts for a user every time they generate a new event. A returning user who visits monthly will never have their user-level data expire, because each visit resets the countdown (GA4 user data reset documentation). Leave this on. There is no reason to turn it off unless you have a specific compliance requirement that demands fixed-duration deletion.

Beyond the GA4 Data Retention Window

Even at 14 months, the ga4 data retention period is short for most B2B companies with sales cycles that stretch six to twelve months. If you need historical event-level data beyond the ga4 data retention window, the only real solution is the GA4 BigQuery export. Once event data lands in BigQuery, it stays until you delete it, and you query it with SQL instead of the GA4 interface. The export does not backfill -- it captures data from the moment you enable it. So if you have not set it up yet, do it today. Future-you will be grateful.

GA4 Sampling: When Your Explorations Start Lying

Data retention controls how far back you can look. Sampling controls how accurate the data is when you do look. Both erode trust in your numbers, and most teams do not notice either one until they try to make a decision.

What triggers sampling

GA4 standard (free) properties apply sampling when an exploration query processes more than approximately 10 million events (GA4 sampling thresholds). The ga4 sampling limit is not based on total property size -- it is per query. A simple date-range exploration over a low-traffic site may never hit it. A segmented funnel with multiple breakdowns over 12 months on a high-traffic site will hit it quickly.

When sampling kicks in, GA4 analyzes a subset of your event data and extrapolates the results. You will see a green shield icon (unsampled), a yellow shield icon (sampled), or a red shield icon (heavily sampled) in the top-right corner of the exploration. The percentage displayed tells you what fraction of available data was actually scanned.

Analytics 360 properties raise the threshold to approximately one billion events per query (GA4 sampling thresholds), which effectively eliminates sampling for most use cases.

I saw the real cost of this last quarter. A SaaS client was running a 6-month funnel exploration segmented by acquisition channel and plan tier. The shield icon showed 14% sampling -- meaning GA4 scanned roughly one in seven events and extrapolated the rest. Their exploration reported a 3.2% trial-to-paid conversion rate. When I rebuilt the same query in BigQuery against the full unsampled export, the actual rate was 4.1%. That 0.9-percentage-point gap was enough to change how the team allocated paid-search budget for the next quarter. If a decision depends on precision, always validate sampled explorations against BigQuery.

How to reduce sampling in free GA4

You cannot turn sampling off, but you can minimize how often you hit the threshold:

1. Shorten the date range. The single most effective lever. A 30-day window processes far fewer events than a 12-month window.
2. Remove unnecessary dimensions and segments. Every additional breakdown multiplies the query complexity.
3. Use standard reports instead of explorations when possible. Standard reports are pre-aggregated and never sampled.
4. Export to BigQuery and query there. BigQuery processes your full, unsampled event stream. If accuracy matters for the analysis, this is the path. I cover the setup and first queries in my BigQuery export guide.

Sampling vs. thresholding vs. cardinality

Teams often conflate these three. They are different problems with different symptoms.

Issue	What happens	How to spot it
Sampling	GA4 scans a subset of events and extrapolates	Shield icon turns yellow or red in explorations
Thresholding	GA4 withholds rows to protect user identity when Google Signals is active	Rows disappear, totals do not add up, "(other)" grows
Cardinality	A dimension has too many unique values, so GA4 groups low-volume ones into "(other)"	"(other)" row absorbs a large share of the data

If you are seeing "(other)" swallow meaningful data, the first thing to check is whether Google Signals is enabled. Turning it off (and relying on Google Ads linking for demographics instead) often resolves thresholding without losing much functionality.

GA4 Annotations: Context Your Future Self Needs

Google removed annotations when it launched GA4 in 2020 and did not bring them back until early 2025. For nearly five years, there was no built-in way to mark a date in GA4 and say "this is why the numbers changed." Teams relied on spreadsheets, Slack messages, or memory -- all of which decay faster than you want.

Now ga4 annotations are back, and they are worth using systematically.

What to annotate

Annotations add a note to a specific date (or date range) visible to anyone with access to the property. The obvious use cases:

• Tracking changes. New GTM container published, consent mode updated, new events deployed. If you follow a tag governance process, annotate every container version push.
• Campaign launches and pauses. Budget shifts, new creative, channel experiments. When traffic spikes or dips three months from now, the annotation tells you why without digging through email threads.
• Site changes. Redesigns, URL migrations, new landing pages, checkout flow updates.
• External events. Algorithm updates, seasonality markers, PR coverage.

How to create an annotation

In any standard report or overview, click the annotation icon in the upper-right toolbar (next to share and insights). Select a date, add your note, and optionally color-code it. Each property supports up to 1,000 annotations (Google Analytics Help). Annotations are visible to all users with report access, but only Editors and Administrators can create or edit them (Google Analytics access roles).

Why most teams still skip them

Because nobody is responsible for doing it. The fix is simple: make it part of your deployment checklist. Every time you publish a GTM container version, you annotate the date. Every time a campaign goes live, you annotate. It takes ten seconds and saves hours of forensic work when someone asks "why did conversions drop on June 12th?"

I annotate aggressively in every property I manage. When a client calls about a sudden traffic change, the first thing I check is whether an annotation explains it. If it does, we move straight to impact analysis instead of spending two hours on root-cause investigation.

GA4 Data Retention, Sampling & Annotation Checklist

If you have read this far and are not sure whether your property is configured correctly, here is the quick check:

1. Data retention: Admin > Data collection and modification > Data retention. Confirm your ga4 data retention setting is 14 months and "Reset user data on new activity" is toggled on.
2. BigQuery export: Admin > Product Links > BigQuery Links. Is a link active? If not, set it up now. Data only flows forward. Once it is running, you can build a Looker Studio dashboard on top of the unsampled data.
3. Sampling awareness: Open any exploration with a 90-day date range. Check the shield icon. If it is yellow or red, narrow the range or simplify the query.
4. Annotations: Pick the last three significant changes to your site or campaigns. Add an annotation for each. Then make it a habit.

These are not advanced configurations. They are defaults that Google set conservatively, and most teams never revisit. The cost of ignoring ga4 data retention alone is invisible until the day you need six months of funnel data for a board meeting and discover it was deleted four months ago.

If that sounds familiar -- or if you suspect there are other settings quietly degrading your data -- a tracking audit will surface every one of them. I go through the full checklist, flag what is broken, and hand you a prioritized fix list.

FAQ

Not sure your GA4 settings are right? Book a tracking audit -- I will go through every configuration, flag what is costing you data, and give you a prioritized fix list.