Google Ads Data Retention: 37 Months from June 2026 (Export Plan)

On May 1, 2026, Nadine Wang from the Advertising and Measurement APIs Team announced a new data retention policy on the Google Ads Developer Blog: starting June 1, 2026, Google is cutting the retention window for granular reporting data to 37 months. What sounds like a technical footnote is a concrete problem for advertisers with long account histories. Anyone running year-over-year comparisons at week or day granularity will only have a rolling 37-month window from June 2026 onward.

The backstory is absurd: Google had only extended retention from earlier shorter windows to 11 years in November 2024. The 11-year policy was marketed back then as a lever for long-term analysis. Now, less than 18 months later, comes the reversal. The signal to advertisers is clear: don’t rely on Google’s retention promises. Hold your own data.

What changes on June 1, 2026

The new policy is tiered by data type and aggregation level. Three different retention windows apply in parallel:

Data type	Retention from June 1, 2026	Examples
Hourly, Daily, Weekly (granular reports)	37 months	Daily campaign performance, weekly Search Terms reports
Monthly, Quarterly, Annual (aggregates)	11 years	Monthly spend, annual campaign roll-ups
Reach & Frequency	3 years	Unique Users, Average Impression Frequency, 7-Day/30-Day Frequency, Distribution 1+, 2+, 3+, 4+, 5+, 10+

The tightest cutoff hits Reach and Frequency data: only three years. That matters for Display, YouTube and Demand Gen advertising, less so for search-only accounts.

Anyone clicking a report in the Google Ads UI that is older than the relevant window will see no data from June 2026 onward. The same applies to programmatic queries via the Google Ads API and Scripts.

Which APIs and tools are affected

The policy applies not just to the Google Ads web interface, but to every surface that serves Google Ads data:

• Google Ads UI: reports older than 37 months disappear from the surface
• Google Ads API: GAQL queries using segments.date or segments.week for data older than 37 months return DateRangeError.INVALID_DATE. A more precise error, DateRangeError.REQUESTED_DATE_GRANULARITY_NOT_SUPPORTED, is planned for a future version update.
• Google Ads Scripts: same cutoff as the API
• Google Analytics Data API: metrics like Advertiser Ad Cost, Clicks and Impressions are silently truncated to the last 36 months, without any error message. That is more dangerous than an explicit error, because the script keeps running as if everything were fine.
• BigQuery Data Transfer Service: stops backfilling for data older than 37 months

Not affected: the DV360 and CM360 APIs. Both keep their own 24-month retention policies, which remain unchanged.

The BigQuery trap that’s actually dangerous

A detail that gets lost in most discussions: anyone using the BigQuery Data Transfer Service to mirror Google Ads data into BigQuery has a blind spot. Manually triggered GA4 backfills for data older than the 37-month cutoff can overwrite existing historical BigQuery tables with empty values. If BigQuery is your current backup for Google Ads data, audit your transfer configuration before June 2026.

Recommended in practice: rather than relying on Transfer Service, write your own export to a native BigQuery table (a real snapshot, not a Transfer Service sync). That is a one-time effort, but it protects against silent data loss.

Why this is a 180-degree turn

In October 2024, Google extended retention to 11 years (effective November 13, 2024). At the time it was marketed as a benefit for advertisers: long historical analysis, granular YoY comparisons, market-trend observation across years.

Less than 18 months later, the reversal to 37 months. Google has not communicated an official reason. Plausible explanations from the PPC community: storage cost, data-protection adjustments, or preparation for stricter GDPR readings that follow data-minimization principles. Whatever the reason, the message to advertisers is unambiguous: don’t depend on Google’s retention guarantees. Keep your own copy.

This lesson is not new in 2026. It was already a key argument in the Server-Side Tracking post for building first-party databases. Data retention is the flip side: not just what gets collected, but how long it remains accessible.

Who actually has a problem

Not every account is hit equally. Three scenarios from the field:

Account under 3 years old: Low priority. The account does not yet have 37 months of history, so the cutoff only becomes relevant in the future. Setting up a recurring monthly export job for granular data is worthwhile, but there is no urgent action item.

Account 3 to 5 years old: Medium priority. Granular data older than 37 months today disappears in June 2026. Anyone running daily YoY comparisons or multi-year Search Terms histories should export proactively. According to the Google Ads Help Center, the rolling window will only cover the last 37 months of granular data.

Account over 5 years old, complex history: High priority. This is where the biggest data asset sits: years of daily numbers, Search Terms reports, Reach-Frequency data from YouTube campaigns. Failing to back this up means losing a hard-to-reconstruct foundation for bid models, seasonality analysis and audits.

Particularly exposed by industry: seasonal businesses (travel, fashion, gifts) with clear YoY patterns that rely on granular data. Accounts with short conversion paths (lead-gen for local services) are less affected because long historical comparisons matter less there.

What to export before June 1, 2026

Three datasets stand out as priorities:

1. Reach and Frequency reports (urgent, because the 3-year cutoff is the tightest):

• Per campaign and asset group: Unique Users, Average Frequency, 7-Day and 30-Day Frequency, Frequency-Distribution buckets (1+, 2+, 3+, 4+, 5+, 10+)
• Export at maximum available range
• Primarily relevant for Display and YouTube, both of which have existed long enough for multi-year reach data to accumulate. Performance Max only went GA in late 2021, Demand Gen in late 2023; for these campaign types there is simply no historically deep data that would be lost to the 3-year cutoff.

2. Daily Performance reports going back to account start (37-month cutoff):

• Per campaign and ad group: Impressions, Clicks, Cost, Conversions, Conversion Value, CTR, CPC, CPA
• Search Terms at daily granularity, where available
• By device type, geo segment, time of day (where useful for audits)

3. Weekly Search Terms reports (37-month cutoff):

• Per campaign: all matched queries with Impressions, Clicks, Conversions
• Essential for negative-keyword workflows and keyword expansion, see Negative Keywords in Google Ads

Monthly and annual aggregates do not need to be exported. They keep their 11-year window and cover trend comparisons at month level automatically.

How to solve this technically

Three realistic options, sorted by effort:

Option A: Manual CSV export from the Google Ads UI

• Click “Download” on each report in the UI, account by account
• Pro: no technical setup, works immediately
• Con: at multi-account scale and many report types, time investment becomes massive. Fine for a single account, painful for agencies or multi-account setups.

Option B: Google Ads Scripts with Sheet export

• A script at account or MCC level pulls data into Google Sheets
• Pro: runs automated, configurable per account
• Con: Sheet cell limits hurt at large accounts, scripts have to be installed per account, scope limit for Reach-Frequency

Option C: Python script with Google Ads API direct

• Script using customer IDs of the accounts to back up, plus GAQL queries for the three priority datasets. Writes to CSV or directly to a native BigQuery table as a snapshot (not via Transfer Service).
• Pro: scales across all accounts at once, one-time setup, runs unattended, bypasses the BigQuery trap
• Con: requires a Google Ads API refresh token and a developer token with Basic Access

For multi-account setups, Option C scales significantly better because the one-time setup amortizes across all accounts. For a single account without programming experience, Option A is the fastest path.

Recommendation for the next weeks

A reasonable sequence regardless of whether you are securing one account or a multi-account portfolio:

1. May 2026: Reach- and Frequency reports first, because the 3-year cutoff is the most acute. Maximum available range.
2. May through mid-June 2026: Daily Performance reports going back to account start, per campaign and ad group.
3. Mid-June 2026: Weekly Search Terms reports, back to account start.
4. From June 2026 onward: a monthly refresh job that appends new data to the snapshot table so the backup keeps pace with the rolling 37-month window.

Exports should land in a clearly documented format (column schema, time range, export date), so that even five years from now it is still traceable where the data came from.

FAQ

As a small business with one account, do I really need to export?

If your account is under 3 years old, nothing is at risk yet. With 1 to 2 years of history, you are out of reach of the cutoff. It gets critical around 3 to 5 years of active advertising. Anyone running daily YoY comparisons or relying on Search Terms history for negative-keyword maintenance should export.

Is the BigQuery Data Transfer Service enough as a backup?

No, at least not without additional safeguards. The Transfer Service itself stops serving data older than 37 months from June 1, 2026. Plus the GA4 backfill trap, which can overwrite existing historical tables. If you depend on Transfer Service sync, you should either pull a one-time snapshot before June 2026 or harden the transfer configuration.

What happens to audits or reports that depend on older data?

After the cutoff, older data is reachable neither via the UI nor the API. Audits that must hit granular data older than 37 months are no longer possible after June 2026 unless the data was exported beforehand. Clean audit preparation is therefore part of the pre-cutoff checklist, see also the post on common tracking issues, which often relies on historical data for diagnosis.

Will Google Analytics 4 data in BigQuery also be deleted?

A GA4 property is governed by its own retention settings inside the GA4 property itself, typically 2 to 14 months for event data, per Google’s GA4 retention docs. The 37-month policy discussed here applies to the API surfaces that serve Google Ads data. If you have a GA4 property exporting to BigQuery, you are governed by GA4 retention rules, not the Google Ads policy. Related but independent topics to audit.

What does this mean for my Smart Bidding data?

Smart Bidding trains on recent data; the model itself does not reach back 37 months of history. From an optimization standpoint, nothing changes. What changes: your own ability to look back at how Smart Bidding performed over multiple years. Anyone who documents bid-strategy changes over years needs the export.

Are Microsoft Ads and Meta also affected?

No, at least not directly. Microsoft Advertising has its own retention policy (typically 36 months for detailed reports), Meta similar values. The change discussed here is Google Ads specific and explicitly does not extend to its sister products DV360 and CM360. For a multi-platform stack, each platform should be reviewed separately.

Sources

• Google Ads Developer Blog: New Data Retention Policy for Google Ads starting June 1, 2026 (Nadine Wang, Advertising and Measurement APIs Team, May 1, 2026)
• Google Ads Data Retention Policy, Google Ads Help Center
• PPC.land: Google Ads cuts granular data access to 37 months starting June 2026
• Surfside PPC: Google Ads reduces data retention to 37 months starting June 2026
• Search Engine Roundtable: Google Ads Updates Data Retention Policy

Your next steps

Anyone managing multiple accounts or one with a long history should treat May 2026 as a hard deadline. Three concrete steps for the coming weeks:

1. Account inventory: which accounts are older than three years? Which run YoY analysis at granular level?
2. Identify priority datasets: Reach-Frequency first (tightest cutoff), Daily Performance next, Weekly Search Terms last
3. Pick an export architecture: manual CSV for single accounts, script-based for multi-account with a BigQuery snapshot as backup

If you need help with the technical implementation or want a clean audit before the cutoff, you can book an intro call. Data preparation is part of the standard audit routine.

→ Book a free intro call