A 3-year-old advocacy org with no paid media history and $10,000 to build a public signatory base. One goal: convert strangers into public signatories. Here's exactly how we did it.
Most advocacy campaigns measure success in impressions and reach — proxies that feel good but don't actually prove anything. This client wasn't interested in proxies. They needed pledges: real people putting their name behind a commitment to responsible AI development. That's a conversion, with all the friction that implies. You have to stop someone mid-scroll, give them a reason to care about an issue they may not have thought about today, and ask them to take an action that requires a name and an email.
The budget was $10,000. The flight was 26 days. The target: as many pledges as possible at the lowest sustainable cost per result. No brand awareness objectives. No "let's build the funnel for next quarter." This was a direct response campaign with a hard deadline, a hard budget, and one number to hit.
This organization had been operating for three years building credibility in AI policy circles — white papers, coalition relationships, earned media. They had never run a paid campaign. Their signatory base had grown through organic outreach: email lists, partner shares, conference sign-ups. That approach had a ceiling, and they'd hit it.
The pledge campaign had a hard deadline tied to a policy moment they couldn't move. There was no runway for a test-and-learn phase, no budget for a second attempt if the first one underperformed. The entire paid media history of this organization would be determined by one 26-day flight. That's not a constraint you manage around. That's the entire brief.
A full-funnel Meta structure isn't complicated. But most campaigns don't actually follow it. This one did.
Systematic creative testing isn't about having an opinion about what will work. It's about being willing to be wrong — and moving budget the moment the data tells you to.
| Creative | Format | Distribution | Pledges | Cost Per Pledge | Verdict |
|---|---|---|---|---|---|
| "It's Not Inevitable" | Image | Broad | 425 | $1.92 | Winner — Most efficient |
| "Steering AI" | Image | Broad | 561 | $2.58 | Strong — Volume leader |
| "Steering AI" | Image | Retargeting | 182 | $2.29 | Solid — Right audience |
| "Trillionaires" | Video | Broad | 56 | $2.92 | Underperformed — Deprioritized |
By week two, the algorithm had reallocated. Image creatives were generating pledges at $1.49 to $1.92. The video — "Trillionaires" — was generating pledges at $2.92. We didn't override the data. We let it run the campaign. By the final week, 94%+ of spend was going to image creatives.
This is the part most campaigns get wrong. They build a video they're proud of and defend the budget allocation long past when the data says to move. The willingness to be ruthless about creative performance is what kept this campaign at $1.52 per pledge instead of $2.50+.
There's a version of this campaign that spends $10,000 and gets 3,000 pledges at $3.33 each. That version launches broad and retargeting at equal budget split, runs all four creatives at equal budget, and doesn't move allocation when the data comes in. That version is what happens when you set up a campaign and walk away.
What got this campaign to $1.52 per pledge was structure and discipline. The right funnel architecture from day one. A creative testing framework that was designed to produce a clear winner, not to validate a pre-existing opinion. And the willingness to let the algorithm — and the data — make the decisions that humans are usually too attached to their assumptions to make.
The broad targeting ad set alone generated 5,300 pledges at $1.49 each — cheaper than the retargeting segment. That's counterintuitive. The instinct in advocacy is to narrow: find the most likely converts and target only them. The data consistently shows the opposite. Give Meta a large, unconstrained pool and it finds people your interest-based segments would have never reached. That's not an accident. It's what the algorithm is built to do when you give it room.
Fixed-flight advocacy campaigns, policy moments, cause launches — any situation where there's no second attempt. If the structure is right from day one, the data does the work. This is what that looks like.
Let's talk →