Google is deleting three years of historical ad data, export before June

Google is deleting three years of your historical ad data in June. The official framing is "data policy update."

What is actually happening: any reporting data older than 37 months will be permanently purged from the platform. Unique users, impression frequency, 7-day and 30-day frequency distribution, reach metrics. Gone.

For the average account, this means losing the only baseline you have for understanding long-term audience saturation and frequency trends. The benchmarks that tell you whether this year's CPMs are genuinely higher or whether your reach is compressing. All of it disappears, and the comparison window resets to three years maximum going forward.

The framing vs the read

Google's framing is data hygiene. Privacy alignment. Simplification of the platform.

The inconvenient read: long-term historical data is also the primary tool advertisers use to identify gradual performance deterioration. Frequency creep, reach plateau, cost inflation. Three years of evidence is enough to spot a pattern. Shorter windows are easier to explain away as market conditions or seasonality.

The trade is not new. Every platform reduces the amount of historical data advertisers can audit independently, and every time the explanation is technical or privacy-related rather than commercial. The pattern is consistent enough that it stops looking like coincidence.

What you actually lose

This matters more in some accounts than others.

• High-frequency display and video accounts lose the ability to validate frequency curves with multi-year baselines. If you have ever defended a frequency cap to a client using historical reach data, that argument now has a 3-year ceiling.
• Branded search accounts lose the ability to compare today's branded volume against a pre-pandemic baseline. For brands that grew significantly between 2022 and 2024, the question "are our branded queries plateauing or growing" gets harder to answer with reliable data.
• Seasonal businesses lose the multi-year seasonal index. Black Friday 2023 vs Black Friday 2024 vs Black Friday 2025 was a clean comparison. After June, you are working with two data points instead of three.
• Auction-insight comparisons get truncated. The slow rise of a competitor's impression share over a 3-year window is visible. Over a 12-month window, it looks like noise.

What to actually do

Before June, run these exports and store them somewhere you control:

• Frequency reports across all video and display campaigns for the maximum date range available. CSV, with all dimensions.
• Reach and unique users data per campaign, monthly granularity, full history.
• Auction insights reports for your top 20 keywords, exported monthly across the full window. These are the most useful for spotting competitor behaviour over time.
• Search term reports for any account that has run more than 18 months, exported in full. Aggregated query trends inform negative keyword strategy and search-volume tracking.
• Demographic and device performance reports at the historical granularity Google currently provides. The newer reporting cuts the data more aggressively than the older API.

Store the exports in a system you control, ideally a data warehouse or at minimum a structured cloud-storage bucket with consistent naming. Do not rely on Google Ads Editor or third-party tools to retain this for you.

The deeper question

What else does Google have a structural incentive to make harder to see over time?

Cross-channel attribution data, query-level insights, audience overlap reporting, view-through behaviour. None of these are getting more transparent. The direction has been one-way for five years, and there is no reason to expect it to reverse.

The advertisers who notice these trade-offs early are the ones building their own measurement infrastructure: server-side conversion APIs, independent customer-data platforms, warehouse-based attribution models. Not because platform reporting is unreliable, but because reliance on platform reporting alone is a strategic vulnerability.

If you want help working out which exports matter most for your specific account before June, book a free audit and we will scope the historical data worth keeping.