Google Ads 37-Month Data Retention: How to Export Your Historical Data Before It Ages Out
Google Ads now retains hourly, daily, and weekly data for only 37 months. Here's exactly what to export, how to do it, and where to store it.
Google Ads 37-Month Data Retention: How to Export Your Historical Data Before It Ages Out
As of June 1, 2026, Google Ads retains hourly, daily, and weekly reporting data for only 37 months. Monthly, quarterly, and yearly reporting still goes back 11 years, but the granular data that powers most year-over-year analysis is now on a rolling expiration. The google ads 37 month data retention change also stopped BigQuery Data Transfer backfills beyond 37 months, so the historical archive most agencies rely on is no longer available through Google's pipes. If you want hourly, daily, or weekly history older than 37 months, you have to export it yourself, store it yourself, and rebuild any reporting that depends on it.
What changed and why it matters
Before June 1, 2026, Google Ads stored granular reporting indefinitely for most reports. You could pull daily-level performance from 2018 if the account existed back then. That data lived in Google Ads, in the API, and in BigQuery Data Transfer.
The new policy keeps that level of detail for the most recent 37 months only. Anything older is rolled up into monthly, quarterly, or yearly aggregates. Aggregates are fine for high-level trend reporting, but they don't help with the things granular data is actually used for:
- Year-over-year day-of-week comparisons
- Seasonality modeling that needs daily resolution
- Historical incrementality testing
- Debugging anomalies more than three years old
- Training models that need granular history
If you don't take action, in 37 months your archive starts deleting itself one day at a time.
The retention rules in plain language
There are two retention buckets now:
37-month bucket (rolling):
- Hourly reports
- Daily reports
- Weekly reports
- Most segment-level breakdowns at granular time resolution
11-year bucket:
- Monthly reports
- Quarterly reports
- Yearly reports
- Account-level aggregates at coarse time resolution
Practical translation: you can still pull a monthly P&L style report from 2017. You cannot pull the daily detail underneath it.
BigQuery Data Transfer's backfill window, which used to allow advertisers to populate BigQuery with multi-year history on first connection, was also capped at 37 months. Setting up a new transfer today won't get you the old data. That's the part most teams discover after the fact.
What to export, in order of priority
If you can only do a few exports before time runs out, do them in this order.
Tier 1: granular performance history
Daily-level performance for every active campaign, ad group, and keyword for the full 37 months currently available. This is the highest-value export because it's what's actively rolling out of the system.
Fields to capture at minimum:
- Date
- Campaign ID and name
- Ad group ID and name
- Keyword or query (where relevant)
- Impressions, clicks, cost, conversions, conversion value
- Network and device segmentation
- Conversion action breakdowns
Tier 2: search terms reports
Search term data is the most expensive to recreate and the most useful for ongoing optimization. Pull daily search terms reports for every Search and Performance Max campaign across the full 37 months. PMax search terms data is more limited than Search, but capture what's available.
Tier 3: change history and account structure
Account structure changes over time (new campaigns, paused campaigns, bid changes, budget moves) are not in the standard reports. Pull change history at the account level. This is what lets future-you understand why performance shifted in 2024.
Tier 4: audience and demographic detail
Audience and demographic performance segmentation by day. Useful for any account doing audience-led bidding or PMax optimization.
Tier 5: asset-level reporting
For accounts running RSAs and PMax, pull asset-level performance data. This is the data that gets lost first when ads are paused or replaced.
Three ways to export
Pick the method that matches your team's setup.
Option 1: Google Ads Editor
Best for: small accounts, one-time exports, teams without engineering support.
Google Ads Editor can download account data and let you export to CSV. It's manual, but it's the lowest barrier to entry. You'll need to repeat the process per account, and there are practical limits on how much data you can pull in one session.
Steps:
1. Download the latest Google Ads Editor
2. Open the account and download all data
3. Use Account > Export to CSV
4. Repeat for each report type you need
This works for a handful of accounts. It doesn't scale to MCC structures with dozens of accounts.
Option 2: Google Ads API
Best for: agencies, in-house teams with engineering support, any setup with more than five or six accounts.
The Google Ads API supports all the reports you need and can be scripted. Most teams already use it for daily reporting; the difference here is running a one-time historical pull across 37 months.
Practical notes:
- Use Google Ads Query Language (GAQL) for granular queries
- Pull data in monthly chunks to avoid timeouts and stay inside row limits
- Store raw responses as Parquet or compressed CSV in cloud storage
- Don't deduplicate or transform on the way in; keep the raw export
If you're going to invest engineering time anywhere, this is where it pays off. Once the script exists, you can re-run it monthly to keep the archive fresh.
Option 3: BigQuery Data Transfer (going forward only)
Best for: ongoing capture from this point forward.
BigQuery Data Transfer will continue to pull current data into BigQuery on a daily schedule. What it won't do is backfill more than 37 months. So set up Data Transfer for everything going forward, but understand that the historical archive has to come through Editor or the API.
If Data Transfer is already running for an account, the existing tables stay intact. The cap is on new transfers and on backfill requests.
Where to store it
Three reasonable destinations, in order of how much engineering they require.
Google Drive or shared cloud storage. Fine for CSV exports from Editor. Set up a folder structure by account, then by year, then by report type. Document what's in each folder.
BigQuery. The natural fit if you're using the API. Land raw exports in a raw dataset, then build views on top. Partition by date and cluster by campaign ID for cheap queries later.
Snowflake, Redshift, or another warehouse. Same pattern as BigQuery, just in your existing warehouse. The principle is the same: raw landing zone, then transformed marts.
Whatever you pick, make sure two things are true: it has redundancy (backups or replication), and it has clear access control. Granular ad data is sensitive.
Backfill checklist for reports that already depend on this data
If you have dashboards or reports already running against Google Ads data, audit them now.
- Looker Studio dashboards connected directly to Google Ads. These will silently stop showing granular data older than 37 months as time progresses. Switch the data source to your warehouse so you control the history.
- Year-over-year reports. Anything comparing rolling daily or weekly data more than three years back will break. Move the source to your archive.
- Forecasting models. If a model trains on multi-year daily history, retrain on your archived data, not on the live Google Ads connection.
- Client-facing performance reports. If a client deck shows multi-year monthly summaries built from daily underneath, switch to the monthly-level data directly (which still goes back 11 years) or to your own archive.
Common mistakes to avoid
A few patterns worth flagging:
- Assuming BigQuery Data Transfer covers history. It doesn't anymore. New transfers only backfill 37 months.
- Exporting once and forgetting. The 37-month window rolls forward every day. Build a monthly refresh job.
- Skipping search terms. Search terms data is hard to reconstruct and high-value. Don't skip it.
- Trusting Google's UI to surface what you lost. It won't. The UI shows aggregates beyond 37 months without flagging the missing detail.
- Storing exports without documentation. Future-you needs to know what's in each file, when it was pulled, and which account it came from.
Frequently asked questions
What is the Google Ads 37-month data retention policy?
As of June 1, 2026, Google Ads retains hourly, daily, and weekly reporting data for 37 months on a rolling basis. Monthly, quarterly, and yearly reports remain available for 11 years. Anything older than 37 months at granular resolution is no longer accessible through the Google Ads UI, API, or BigQuery Data Transfer.
Does the 37-month limit apply to BigQuery Data Transfer backfills?
Yes. Google Ads BigQuery Data Transfer no longer backfills beyond 37 months. New transfers and re-initialized transfers will only pull data inside that window. Existing transfers that have been running keep their accumulated history.
How do I export historical Google Ads data?
The three options are Google Ads Editor (for small accounts and CSV exports), the Google Ads API with GAQL (for any account with engineering support or more than a handful of accounts), and BigQuery Data Transfer (for ongoing capture going forward, not for backfill).
What data should I export first?
Start with daily-level performance for campaigns, ad groups, and keywords across the full 37 months, then search terms reports, then change history, then audience and demographic segmentation, then asset-level reporting.
Can I still see monthly performance from five years ago?
Yes. Monthly, quarterly, and yearly reports remain available for 11 years. The change only affects hourly, daily, and weekly resolution.
How often should I refresh my export?
At minimum once a month. Daily is better if the export is automated through the API. The 37-month window rolls forward every day, so any data inside that window today will eventually leave unless you've captured it.
Bottom line
The 37-month rule is permanent and it rolls. If you want hourly, daily, or weekly Google Ads history beyond 37 months, you have to own the archive. Start with the API or Editor, store raw exports in a place you control, refresh monthly, and update any reports that currently depend on Google holding the data for you. The cost of building this is small. The cost of discovering it doesn't exist when you need it is much larger.
*Want help building a Google Ads historical data archive for your accounts? Book a call with us.*
Related articles
- How to Track Google Ads Calls From Your Google Business Profile (the Hidden Phone Number Setting)
- How to Remove "Found by Google" Products From Google Merchant Center
- Ask Advisor and AI Performance Insights: What Just Launched in Google Merchant Center on June 23, 2026
- The DSA to AI Max Migration: What's Changing in September 2026 and February 2027
- The August 17 Google Ads Bidding Target Change: What Limited-by-Budget Campaigns Are About to Do Differently
- Google Will Use UK and EU IP Addresses for Ad Personalization on August 3, 2026. Here's What Advertisers Need to Do.
- All Kanopy resources
- SEO & web services
- Paid ads services
- AI outreach services