Google Ads' 37-Month Limit: A SaaS Attribution Risk

Buried under the louder GA4 consent headlines this month is a change that will quietly hurt SaaS teams with long sales cycles: as of June 1, 2026, granular Google Ads reporting only goes back 37 months. As Invoca notes in its analysis of the broader June update, the retention cut landed at the same moment consent controls tightened, "shrinking your historical view at the same moment consent controls are tightening" (Invoca). For a consumer business with a same-day purchase, three years of history is plenty. For B2B SaaS measuring cohorts across multi-year lifecycles, a 37-month window is dangerously thin.

What a 37-month cap actually removes

The change does not touch the data you collected yesterday; it caps how far back the detailed view reaches. Granular reporting — the campaign-level, keyword-level, conversion-level detail you use to reconstruct what actually drove pipeline — now rolls off at roughly three years. Data older than the window leaves the reporting surface, and Google does not hand you a copy on the way out. If you have not exported it, it is gone from your reach at the resolution that matters.

Thirty-seven months sounds generous until you count it in annual cycles. It is three full years plus a single extra month — just enough to place one quarter against the same quarter a year earlier, and the year before that, with almost no buffer. For seasonal businesses that depend on comparing this Q2 against the last several Q2s to tell real growth from an annual rhythm, the cap removes the deepest comparison point right when you would reach for it. The history that gives a year-over-year number its meaning is exactly the history aging out first.

Why long SaaS sales cycles make this worse

B2B SaaS routinely runs sales cycles of six to eighteen months from first touch to closed-won, and the metrics that matter most — LTV, CAC payback, net revenue retention — only become legible long after acquisition. To judge whether a channel you invested in two years ago actually produced profitable customers, you need to follow that cohort from the click through the long tail of expansion and churn. That analysis is inherently backward-looking, and it depends on the granular history the 37-month cap is now eroding from underneath it.

Concretely, a customer acquired through a paid campaign in early 2023 may only have reached payback in 2024 and started expanding in 2025. To attribute that expansion revenue back to the original channel and compute a true payback period, you need the campaign-level acquisition detail from 2023 still queryable in 2026 — which is precisely what now falls outside the window. The same backward dependency runs through the cohort work in the SaaS product analytics metrics that predict growth: a metric that needs years of lineage cannot survive on a three-year memory you do not control.

The compounding effect with the consent split

This retention cut would be manageable on its own. What makes 2026 different is that it compounds with the consent changes that arrived in the same window. The June 15 ad_storage change reshaped what gets collected going forward, and we covered its mechanics in the GA4 and Google Ads consent split. The 37-month cap then attacks the other end of the timeline, limiting how far back you can look at what was already collected. One change narrows the present; the other erases the past.

The combined message is hard to miss: relying on Google's platforms as the long-term memory of your acquisition history is no longer safe. The collection layer is getting noisier as consent gates more of it, and the reporting layer is getting shorter as retention windows close. A SaaS team that wants stable, multi-year attribution baselines cannot assume the data will simply be there when it goes looking. Both ends of the visibility window are tightening at once, and the only durable response is to stop treating a rented reporting view as a system of record.

The fix: mirror your data into first-party storage

The defensible response is to mirror Google Ads conversion and campaign data into a warehouse you control, on a recurring schedule, and to backfill as much current history as Google still exposes before more of it ages past the window. Once the data lives in your own infrastructure, the 37-month cap stops being your problem — Google can truncate its reporting view all it likes, because your year-over-year baselines now sit in storage that answers to you. The export job is modest engineering; the cost of not having it is an attribution model with amnesia.

This is the same first-party posture that underpins resilient measurement across the board, the argument we make in full-funnel attribution for SaaS in a cookieless world: own the signal, and treat external platforms as sources you copy rather than systems you trust to remember. Keep your cohort and attribution models running on your warehoused copy so the logic and the history are both under your control. The teams that move now will preserve the long-cycle baselines that the slower teams will discover, too late, have quietly disappeared.

A practical timeline for SaaS teams

Treat this as urgent rather than routine, because every month of delay is a month of history that ages closer to the cliff. The first step is to export your full available Google Ads history — campaign, keyword, and conversion detail — into your warehouse immediately, capturing the deepest data Google still exposes today. The further back you can grab before the window keeps sliding, the more annual comparisons you preserve. This is a one-time backfill that gets strictly harder the longer it waits.

Then automate it: a scheduled job that pulls fresh conversion and campaign data into the warehouse so your first-party copy stays current without anyone remembering to run it. Rebuild your year-over-year and cohort reporting on that warehoused data rather than on Google's native view, and validate that your payback and retention models still resolve against cohorts from three-plus years ago. Do this, and a retention window that breaks other teams' baselines becomes, for you, a non-event — which is exactly what good measurement infrastructure is supposed to feel like.