CoW Swap on Specify, measured down to the swap
CoW Swap has run campaigns on Specify since early 2025. This is what they returned.
>1,500
Verified conversions
>$50M
Trading volume
6x
Measured incrementality
1.3%
Peak conversion rate
Figures from the campaign dashboard and the holdout test.
Conversions by month
Wallets acquired per month across 2025
Budget moved each month toward the audiences and placements that were converting.
The problem
Click attribution misses web3 users
Ad platforms optimize for clicks. Experienced DeFi users rarely click an ad: they see it, then open the app from a bookmark or their wallet. Click attribution misses those conversions, so the campaign gets optimized toward the wrong people.
What CoW Swap asked for
- Attribution from impression to onchain swap
- A holdout test, so lift is measured rather than assumed
- Cost per acquisition and return on spend
- Access to the underlying conversion data
How the partnership ran
From a prototype in January 2025 to continuous campaigns
January 2025: a prototype
When we first spoke to CoW Swap, Specify had no track record and a rough product. Ravi and the marketing team agreed to test the idea anyway: target wallets by onchain history and count swaps rather than clicks.
March to May 2025: first results
CoW Swap kept running campaigns while we rebuilt the platform, and Specify V1 launched in May. Their feedback set most of the roadmap.
May 2025 onward: back-to-back campaigns
CoW Swap has had a campaign live ever since. We meet monthly to go through the numbers and plan the next experiment.
Hundreds of hours
The numbers on this page took hundreds of hours of shared analysis. Most of the measurement methods Specify uses today were worked out with CoW Swap’s team.
What we found: clicks did not predict swaps
We expected the wallets that clicked to be the wallets that swapped. They were not.
What the data showed
Most wallets that swapped after an ad never clicked it. They saw the ad, then opened CoW Swap the way they usually do.
Optimizing for clicks
Budget goes to the audiences that click, and most of them never swap.
Optimizing for swaps
Budget goes to the audiences that swap, whether or not they clicked.
Onchain attribution is what made this visible. We match impressions to wallets and wallets to swaps, so a conversion counts with no click involved.
Why it worked
Onchain targeting
We picked wallets by what they had done onchain:
- • Swaps on competing DEXes and aggregators
- • Trade sizes where MEV protection saves real money
- • Activity on the chains CoW Swap was launching on
- • Recent DeFi activity
Wallet history is the strongest predictor of a swap we have found.
Chain launches: the peak
CoW Swap launched on nine chains in 2025. For each launch we targeted wallets already active on that chain.
>1%
impression-to-conversion
during peak launch windows
Those wallets already had the chain set up and gas to spend, so a new venue cost them nothing to try.
Conversion rate, compared
~0.05%
Display ads, typical
1%+
CoW Swap on Specify, peak
13-26x
Ratio
The holdout test
CoW Swap wanted to know how many of the swaps would have happened without the ads.
We held back a group of wallets matched to the targeted audience on activity and history, showed them nothing, and compared swap rates.
0.1%
Holdout group
(no ads)
0.6%
Exposed group
(saw ads)
6x
lift over the holdout group
What that means
Of the first 1,500 conversions:
- • About 250 would have happened without ads
- • About 1,250 would not have
In short
Five of every six conversions were caused by the ads.
Run the same holdout test on your protocol.
What CoW Swap changed in the product
Several features every advertiser now uses started as CoW Swap requests.
A/B testing
Two creatives or audiences, compared on conversions
Lifetime reporting
First swap, repeat swaps, and total volume per acquired wallet
Monthly reviews
A written report each month with what we plan to change
Cohort analysis
Retention by month of acquisition (coming soon)
What we took from it
Measurement
- Measure the onchain action itself
- A holdout test is the only way to know lift
User behavior
- Experienced DeFi users see ads and rarely click them
- Wallet history predicts conversion
Execution
- Per-conversion pricing keeps both sides honest
- A new experiment every month beat any single campaign
CoW Swap plans its Specify budget from two numbers, cost per swap and volume per swap, both measured onchain.
This was a marathon of experiments, not a one-off campaign. The results speak for themselves. The Specify team is deeply data-driven, open to hard questions, and willing to experiment continuously.
That's what helped us move away from guesswork and focus on real on-chain outcomes.
Ravi
Growth, CoW Swap
Run the same test on your protocol
CoW Swap started with one test campaign. Book a kickoff call and we set one up for you. The report is yours whatever it shows.