13 paid-ad results, each credited and linked to the source.
We have 173 Google Ads results on file, and every one reports an improvement, so read this page as a library of success stories rather than a fair sample.
The changes that show up most are audience and keyword targeting (63 results), campaign structure (58) and measurement fixes (47).
Only 18 are controlled split or geo tests, and 98 (57%) are vendor-reported, mostly Think with Google case studies.
Each result is graded by how it was measured; the evidence grades page explains the scale.
Counting the evidence field across all 173 members: 117 are before/after comparisons, 24 are campaign results with no baseline, 18 are split or geo tests, 7 are account snapshots, 5 are studies and 2 are bare claims.
Of the 98 vendor-reported results, 93 come from Think with Google, where Google publishes stories about its own products.
Fourteen of the 18 split tests are also Google-published.
That leaves 4 independent controlled tests, which are the strongest evidence on this page.
The clearest independent test is a brand search holdout from @runtraffic: incremental ROAS 8.17, but new-customer ROAS only 1.17. Brand search drove real extra revenue, but the post doesn't break down how much of it came from new versus returning customers.
Among the vendor-reported split tests, Kia UK's value-based bidding test against target CPA reported 20% more value, a 10% cheaper CPA and a 13% higher conversion rate.
Taxfix reported 9% more tax returns at a 44% lower cost per conversion with Performance Max than with Search alone.
Bol's 50/50 split of PMax against standard Shopping found a much smaller 9% revenue lift, which is a useful reality check.
The biggest before/after numbers are hard to compare because each one measures something different.
One bedding brand's 15%-off Labor Day sale (@SamPiliero) took revenue up 102% and blended Meta plus Google ROAS from 10.11x to 13.17x.
That gain came from the offer, not from a Google setting.
Counting lever tags (one result can carry several): audiences and keywords 63, structure 58, measurement 47, bidding 40, budgets 35, copy 17, creative volume 14, visuals 13, angles 12, offers 8. Creative barely features on Google, with only one result tagged for hooks.
Structure wins usually mean consolidating campaigns or splitting them by margin, category or customer type; see google ads campaign structure results.
Measurement wins usually mean feeding Google better signals, such as offline conversions, CRM lead quality or enhanced conversions, and the strongest of these are collected under google ads incrementality results.
Budget results are mostly gradual increases: one regulated ecommerce account (@thedanielbudai) went from $4.5k to $505k a month in revenue by raising budgets 10 to 15% at a time, with ROAS of 2 to 2.5x reported for the peak month (month one ran below break-even).
Results like this are self-reported and not typical.
More examples are in scaling google ads budget results.
Most Google results have no format tag, because the change was to targeting or bidding rather than the ad itself.
Only 39 format tags exist across the hub, led by video 13, UGC 6, AI-generated 5, catalog 5 and static 4. Format is therefore a weak basis for choosing a Google test.
We matched campaign names in each result's URL and headline, so these counts overlap and are approximate.
Search, broad match and keyword changes appear in 47 results, and Performance Max in 44. Demand Gen or Discovery appears in 8, YouTube in 8, App campaigns in 8, AI Max in 7 and Shopping in 7.
Performance Max is where the Google-published stories cluster.
In Castlery's five-week experiment (vendor-reported), online revenue rose 200% and ROAS 80%.
The full set is in the performance max case study hub, with separate pages for google search ads examples, google shopping ads examples and demand gen campaigns.
By vertical: ecommerce 35, home 22, apparel 21, beauty 14, finance 11, travel 10, SaaS 9, apps, auto, food and accessories 7 each, and smaller counts for the rest.
Finance, auto and SaaS results lean on lead quality and offline conversion values more than retail results do.
Half of the beauty results (7 of 14) are Estée Lauder and L'Oréal brands.
We took the headlines that state a Google ROAS or CPA percentage change and left out the Meta-only figures.
The ROAS changes range from +5.2% to +858.8%.
The CPA or cost-per-conversion reductions range from 3% to 96%, with a median of about 33%.
These are winners' numbers, so treat them as a ceiling, not a forecast.
Absolute levels vary far more. @runtraffic's ecommerce snapshots show account ROAS 6.96 (Search 9.37, PMax 7.65), but Demand Gen at 1.18 (mostly statics) and at 1.63 on $183k of spend.
For cross-platform context, see median CPA and ROAS by platform.
No member is tagged as lost or flat, which says more about who publishes results than about Google.
The warning signs are in the snapshots. In one declining Google ROAS audit, ROAS fell from 3.71x to 1.9x with no change named.
Scaling one lead-gen account's spend 5x in 3 days cost 0.25 points of CTR.
Another snapshot found Discover taking 26% of PMax spend.
Some vendor wins are also modest once you read them closely.
Uniqa's AI Max pilot added 13% conversions while CPA rose 3%, and the brand search holdout's new-customer ROAS (1.17) sits far below its incremental ROAS (8.17).
Expect smaller gains if your account is already consolidated, your conversion data is thin, or most of your revenue comes through brand terms.
Choose one lever from the top of the list (targeting, structure, bidding or measurement) and test it with Google's experiments tool or a geo split rather than a before/after comparison.
Hold budget steady, run at least a few weeks, and track new-customer revenue as well as platform ROAS.
The creative testing guide covers sample size and test design.
This hub collects 173 Google Ads results with the change made, the numbers and the source.
18 are controlled split or geo tests, 117 are before/after comparisons, and 98 are vendor-reported, mostly by Think with Google.
Most change targeting (63 results), campaign structure (58), measurement (47) or bidding (40), not the ad creative.
Across the headlines that state a Google CPA change, the median reduction was about 33%.
Yes. An ecommerce brand search holdout found incremental ROAS of 8.17 but new-customer ROAS of only 1.17, and Bol's 50/50 split of Performance Max against standard Shopping reported a 9% revenue increase.
Ecommerce is the largest vertical in the hub with 35 results.
Use them with care. 57% of the results here are vendor-reported, none are tagged as losses, and only 4 independent controlled tests exist.
Treat the reported gains as an upper bound.