Google Ads Conversion Lift Goes Self-Serve (October 2026): What Small Teams Should Know Before Trusting Their Brand Search CPA

Google Ads now lets eligible advertisers set up user-based Conversion Lift studies for Search and Performance Max campaigns without a Google representative, a change Search Engine Land reported on October 7, 2026 and Search Engine Roundtable confirmed on October 8 after Google rewrote its help page. The entry ticket is at least 1,000 observed conversions and a $5,000 minimum budget for the campaigns in the study, which means the people who most need the answer are mostly the ones who can’t run the test.

The tool exposes a belief most small advertisers hold without checking: that a conversion Google reports is a conversion Google caused. Here’s who can now test that inside the account, and a smaller test for everyone else.

What changed with Google Ads Conversion Lift in October 2026?

Search and Performance Max joined Display, Video, Demand Gen and App as campaign types that can run self-serve, user-based Conversion Lift. Before this, Search and PMax studies needed a Google rep, as PPC News Feed’s Hana Kobzová first flagged.

The requirements listed on Google’s help page:

• At least 1,000 observed conversions. Modeled “supplementary” conversions, such as store visits, don’t count.
• A $5,000 minimum budget for the campaigns in the study.
• At least one conversion action that works with Conversion Lift.
• Each campaign can be in only one study at a time (Brand Lift, Search Lift or Conversion Lift). iOS app campaigns and Travel Ads aren’t supported.

Setup is under Campaigns › Experiments › Lift studies. You choose a holdback from 1% to 50%, and the tool estimates “study power” between 50% and 95%. Google notes access still varies by account.

Does a conversion in Google Ads mean the ad caused it?

No. Attribution records that someone clicked an ad before converting. It can’t tell you whether they would have converted anyway. That’s the belief this release corrects, and it costs small advertisers more than any bidding setting does.

Here’s how a lift study works. Google randomly holds back a share of eligible users, your chosen 1–50%, who never see the campaigns being measured. Everyone else sees ads as normal. When the study ends, Google compares the conversion rate of the exposed group with the control group’s. Only the difference counts as incremental. Divide spend by those incremental conversions and you get incremental CPA (iCPA). If you measure by value, you get incremental ROAS (iROAS).

This logic lets you predict results before you test. The more of a campaign’s conversions come from people who already chose you, the bigger the gap between reported CPA and iCPA. Brand search, remarketing-heavy Performance Max and any campaign that “wins” the last click before a booking will show the widest gaps. Non-brand searches from people who have never heard of you will show the smallest.

Why does The Trusted Local Choice make this test matter more, not less?

A business that has already become the trusted local choice pulls in demand by name, so a large share of its “ad conversions” are customers who were coming anyway. That’s why I’m reading this news through The Trusted Local Choice rather than pure direct response. The question isn’t whether ads produce conversions. It’s whether they produce conversions that your reputation didn’t already secure.

Here’s an illustrative case. A two-location veterinary clinic spends $4,000 a month on Google Ads, and 60% of that ($2,400) goes to brand search. Google reports 200 conversions on the brand campaign, a $12 CPA, so it looks like the best line in the account. But clients searching “[clinic name] hours” also see the clinic’s Maps listing and organic result. If only 15% of those 200 bookings needed the ad, the clinic is paying $80 per incremental booking, not $12. That’s $2,400 a month, nearly $29,000 a year, held in a campaign whose real value is a fraction of what the dashboard shows.

The lens cuts both ways. Brand search isn’t worthless for a trusted local business. If competitors bid on your name, or a delivery aggregator outranks your own site, the ad may be what keeps the booking yours. Reported CPA can’t tell you which situation you’re in.

Is Google’s self-serve lift test neutral?

It’s useful, but read it knowing Google designed the test, chose the defaults and grades its own ads. Three details in the help documentation are worth weighing.

First, Google says conversion-based bidding (Target CPA, Maximize Conversions, Target ROAS) and data-driven attribution are “associated with higher lift.” That’s an association across advertisers, not proof that switching bid strategy causes lift. Accounts that already use those settings tend to be larger and better tracked. Second, the Summary and Overall tabs include modeled “supplementary” conversions such as store visits; the Conversion goal tab separates them, so record the observed figure. Third, Google labels results with study power between 50% and 90% as “directional.” An account that barely clears 1,000 conversions will often land there. A directional result that says your brand campaign is incremental shouldn’t settle a budget argument on its own.

Search Engine Land calls the change making incrementality “more accessible.” For a small business near the threshold, it’s accessible on paper and often inconclusive in practice.

How should a small team run a lift test this month?

If you clear the thresholds, run one clean Google study before the holiday auction. If you don’t, run a manual brand-search holdout. Both fit into a week of part-time work.

1. Check eligibility in five minutes. Set the date range to the last 12 months and look at observed conversions on primary actions, excluding store visits and other modeled actions. Under 1,000, skip to step 6.
2. Pick one question. “Is brand search incremental?” or “Is PMax adding customers we wouldn’t get otherwise?” Not both. Google advises including every campaign of the same type and targeting, because measuring only some of them makes lift harder to detect.
3. Choose a conversion that happens often. A booking request or add-to-cart may give you enough volume where a completed purchase won’t. Google says upper- and mid-funnel actions tend to show more measurable lift.
4. Set the holdback and read the power estimate before you launch. If power shows below 90%, either extend the dates or accept that the result will be directional and say so in your notes.
5. Freeze creative and audiences for the whole run. Google says bid and budget changes are acceptable, but creative and audience changes muddy the result. Run at least 14 days, and 21 is safer for anything with a consideration cycle.
6. Under the threshold? Run a manual holdout. Pause brand search on alternating weeks for a month: on, off, on, off. Track total bookings, calls and orders from your own system, not from Google Ads. If totals barely move during the off weeks, the campaign is mostly paying for customers you already had.
7. Write down iCPA next to reported CPA. Use the ratio, not the dashboard CPA, the next time you decide where a dollar goes.

If your account is already deciding budget splits for you, see our breakdown of Google’s Holistic and Growth budget modes, because a lift result is exactly the evidence those modes don’t use.

When should you run a lift study around Q4 2026?

Between now and mid-November, or wait until mid-January. Holiday demand inflates both groups in ways that make a short study hard to read.

• October 12, 2026: Google begins switching on automated promotions by default for some Search and PMax campaigns, as Search Engine Roundtable reported. A discount that appears mid-study changes the creative. Check the setting before you launch.
• October 13 – November 10: the cleanest window for a 21–28 day study before Black Friday pricing.
• November 26–30, 2026: Thanksgiving through Cyber Monday. Avoid having a study running across these dates.
• January 11 – February 8, 2027: the post-holiday fallback. Demand is lower, so plan for longer runs to reach the same power.

What should you do with the result?

Move money only when the gap is large and the result is not merely directional. If iCPA comes in within about 1.5× of reported CPA, the campaign is doing real work, so leave it. If it’s 4× or more, cut that campaign’s budget in steps of 25% and watch total bookings for two weeks after each step. If you also run conversion uploads, check how those values feed into reporting first. Our guide to multi-source conversions covers the overwrite risk. Tools like campaignpress.ai can help a small team keep that test log in one place, but the decision rule above works in a spreadsheet.

Frequently asked questions about Google Ads Conversion Lift

Does Conversion Lift cost extra in Google Ads?

There’s no separate fee mentioned in Google’s documentation. The real cost is opportunity: the holdback group, which can be 1% to 50% of eligible users, doesn’t see your ads during the study. Larger holdbacks give clearer answers sooner but forgo more conversions; smaller ones need longer runs.

Can I test brand search and Performance Max at the same time?

Not with the same campaign in two studies. Google allows a campaign in only one Brand Lift, Search Lift or Conversion Lift study at a time. A small account rarely has the volume to power two separate studies anyway, so pick the campaign with the biggest budget and the least certainty.

What if my account has fewer than 1,000 conversions?

You won’t qualify for the Google tool, but you can still learn the main thing. Pause one campaign on alternating weeks for a month and compare total business outcomes from your booking system, phone log or store backend. Treat small differences as noise and act only on large swings.

How long should a Conversion Lift study run?

Google allows seven days but recommends more than 14. It says that cutting studies short on products where people take a while to buy has reduced measured Absolute Lift by up to 17%. For local services with same-week bookings, 14–21 days is reasonable. For considered purchases such as furniture or B2B services, plan for 28 days and schedule around holiday peaks.