25 paid-ad results, each credited and linked to the source.
Across 247 published results that report ROAS, the most common winning changes were audience changes (87 results) and campaign structure changes (67), and the typical stated ROAS lift was +50% (median of the 120 results that give a clear percentage change).
Only 30 of the 247 are controlled split tests, and their median lift is smaller at +40%.
140 of the 247 (57%) are vendor-reported, so read the numbers as best cases, not forecasts.
How we counted: we took every member whose headline or numbers state a ROAS change as a percentage or a before and after pair (120 of 247), and dropped ones where the figure was a level, not a change, or came from counting extra channels.
Each result counts once under every lever it is tagged with, and 157 members are tagged with two or more levers, so these medians describe packages of changes, not single moves.
| Lever | Results with a % change | Median ROAS change |
|---|---|---|
| Angles and messaging | 13 | +72% |
| Audiences and targeting | 39 | +65% |
| Campaign structure | 38 | +63% |
| Bidding | 19 | +63% |
| Creative volume | 22 | +62% |
| Budgets | 26 | +55% |
| Visuals | 27 | +46% |
| Offers | 14 | +44% |
| Measurement and tracking | 22 | +41% |
The spread between levers is narrower than the spread inside each lever, so no single lever is a safe bet.
Copy changes showed the highest median (+120%), but only on 8 results, all bundled with other changes.
Browse the underlying entries in all results, or see the change types ranked by cost per purchase instead.
Four simple, single-change examples show the range. A store cut its price by €10 with the same ads and budget and saw CPA €170.34 → €93.51 (−45%) (a community post shared by @SamPiliero, before/after).
A bedding brand's 15% off Labor Day sale produced Revenue +102%, blended ROAS 10.11x → 13.17x (@SamPiliero).
Obvi split one CBO of 50 ads into 5 ad sets for NC-ROAS +55.65% (1.79) (@AshvinMelwani).
And a day-one product test turned $100 spend → $410 sales on day one (@EcomRedneck), one day of total store sales from a single campaign with no control; results like this are not typical.
Visuals (37 members), creative volume (38), angles (24) and hooks (9) together make up much of the hub.
The repeating pattern is new angles aimed at a specific buyer, tested in volume.
Chatpdf's weekly hook concepts on Meta came with ROAS +280% year over year and CAC −45% (Admiral Media, before/after), and a problem-led hook test at an 8-figure DTC brand gave +20.5% ROAS and −10.5% CAC over 75 days.
The cleanest creative evidence is in catalog ads, where vendors ran split tests on templates.
Twillory's review and bundle templates moved ROAS from 1.96 to 4.49 (+127%) against the control (Marpipe, vendor-reported).
See more hook changes, angle & message changes and creative volume & testing cadence changes.
20 members change the offer and 10 change the landing page.
The offers that worked were simple and visible: free delivery shown on catalog ads gave Raymour & Flanigan +54% ROAS vs control, and Latico's 20% off strikethrough design reached 3.52 ROAS against 2.66 for its other ads in May (+32%), both vendor-reported Marpipe tests.
Offers raise ROAS partly by lifting conversion rate, so check margin after the discount, not just ROAS. More in offer & price changes.
Structure (67), budgets (46) and bidding (35) are the largest account-side groups.
Consolidation shows up again and again: K-Republiq merged three campaigns into Advantage+ Shopping with cost caps and weekly ROAS went from 1.62 to 3.1 (Flighted), and Ando's side-by-side comparison found 40% higher ROAS on Advantage+ Shopping than broad CBO (Flighted).
Bidding has some of the strongest controlled results.
A dating app on Moloco moved Android D30 ROAS from 41.2% on CPA bidding to 117.8% on ROAS bidding at equal spend, though cost per registration was roughly 2 to 2.5x higher (Admiral Media).
Earnestree's TikTok Cost Cap test reported +38% ROAS vs maximum delivery (vendor-reported).
Compare more campaign structure changes and bid strategy changes.
Of the 120 results with a clear percentage, the middle half sit between +27% and +97%, with a median of +50%.
21 report less than +20%, and 28 claim +100% or more.
The bigger numbers mostly come from weaker evidence: before/after results have a median of +60% (70 results), while split tests sit at +40% (23).
Vendor-reported results have a lower median (+35%, 66 results) than non-vendor ones (+91%, 54).
Most non-vendor members are agency case studies with before/after framing and several changes at once, which inflates the number.
A realistic target for one well-run change is a lift in the 20-40% range, not a doubling.
No member in this hub is recorded as a loss, and only 2 of 247 are mixed, so this is a set of published wins, not a fair sample.
The mixed ones are instructive: Obvi's GLP-1 angle ran 123 ads at ROAS 1.06 and CPA $58.96, and a UK merch store hit ROAS 2.03 while spending £700+ in countries that never bought.
One audit showed Google ROAS falling from 3.71x to 1.9x with no change named.
Changes probably won't move ROAS when the product, price or margin is the real limit, when the account lacks the conversion volume for value bidding, or when spend rises fast; several winners held ROAS roughly flat while spend grew.
If you want help running these tests, see agencies who publish results like these.
ROAS is attributed revenue divided by ad spend, so it moves when the revenue you count moves.
46 members are tagged with measurement changes, and some lifts are partly counting: Topicals reported ~7.0 Total Commerce ROAS in Fospha vs 2.2 in TikTok Ads Manager, and a brand search test showed incremental ROAS 8.17 but new-customer ROAS 1.17 (@runtraffic).
Before you test, fix the attribution window, compare the same window before and after, and track new-customer or incremental ROAS alongside platform ROAS.
Run a split or holdout when you can.
The full ROAS formula and definition covers break-even ROAS and how it relates to MER.
In 247 published results, the most common winning changes were audience changes (87) and campaign structure changes (67), followed by measurement, budgets and creative volume.
The median stated lift was +50% across 120 results with a clear percentage, and +40% in controlled split tests.
Start with one change, test it against a control, and judge it on the same attribution window.
123 of the results here ran on Meta.
The repeating wins were consolidating campaigns into Advantage+ Shopping, adding cost caps or bid caps, testing new angles in volume, and fixing tracking.
For example, Ando's side-by-side comparison found 40% higher ROAS on Advantage+ Shopping than broad CBO.
65 results ran on Google. The most frequent changes were Performance Max restructures, value-based bidding, offline or enhanced conversions, and splitting brand from non-brand search.
Results with a clear percentage on Google had a median ROAS lift of +66% across 37 results, most of them before/after rather than controlled.
Low ROAS often comes from wasted spend (one store spent £700+ in countries that never bought), weak angles, or tracking that undercounts revenue.
Check geography and placements, confirm the pixel and conversions API send purchase values, then test new angles and offers before raising budget.