We cut an AI personal assistant's cost per signup 35% while scaling spend 2.7×.

A launch with no data to lean on.

The client is a venture-backed AI personal assistant that makes calls, answers email and books things for its users. It brought Valid in at launch: $0 revenue, no ad history, and a goal of reaching $1–2M a month in spend. The brief was clear:

• Get Google and Meta live from zero
• Ramp spend quickly without losing efficiency
• Win users who use the assistant, not just sign up

Spend went up. Cost per signup came down.

We raised budgets twice in week 2 and every efficiency metric improved with them. Daily spend grew 174%, cost per signup fell 35%, and 18% of clicks became signups, up from 7%. Performance Max carried the ramp, scaling 6× as its cost per signup dropped from $27.96 to $16.84.

Fix the data, then push budget.

In launch week, Google could only match 17 of 55 signups to an ad click, so its bidding had almost nothing to learn from. We made the client's product analytics the source of truth, ran Search on Maximize Clicks to collect click IDs, and fixed the dropped emails and click IDs with the client's engineers. Google now sees 99% of signups.

Cheap signups weren't the goal. Active users were.

We launched Meta in week 2 and judged every channel on what users did after signing up. Valid's Meta campaigns brought in people who connect their Gmail account right after signing up, the step that lets the assistant start working for them, at 2.2× the rate of Google signups. They also reach a 5th task at 6.8× the rate. We moved both platforms to optimize for Gmail connections, and Meta's cost per Gmail connection fell from $98 to $50 in two days.

Paid became the biggest channel.

Daily signups went from about 28 before launch to over 300. Valid's Google and Meta campaigns supplied up to 70% of a day's signups while organic kept growing alongside, for 714 paid signups at $20.09 each in the first 14 days.