The Situation
Mighty needed to drive account creations and shop launches through Meta Ads and Google Ads. The objective was efficient user acquisition, but early campaigns were constrained by sparse conversion data.
Without sufficient event density, automated bidding could not stabilize, lookalike audiences lacked depth, and increasing spend would have amplified inefficiency rather than improved it. Signal density was the real bottleneck.
The Goal
Establish a disciplined acquisition framework capable of producing reliable account creation conversions while preserving room for controlled experimentation. Stability first, then scalable growth.
Our Approach
We prioritized structure over scale.
On Meta, we shifted campaigns to optimize directly for account creation events and used short-window website visitor audiences as an interim signal source while account-based audiences matured. On Google, we separated testing from stable acquisition flows to protect performance while refining keyword structure and bidding strategy.
The operating principle was isolation: isolate audiences, isolate creative variables, isolate keyword themes. Give each platform cleaner signal to learn from, and protect the campaigns already producing revenue from the noise of experimentation.
Execution Highlights
Meta Ads
Campaigns were restructured around account creation as the optimization event, with Campaign Budget Optimization consolidating learning across ad sets.
Because account creation data was still limited, recent website visitor audiences carried the load as an interim signal source while stronger account-based audiences matured. Testing showed that 1% lookalikes built from recent website visitors were the strongest interim targeting layer, and a high-performing demographic cluster emerged: women aged 35 to 44, primarily reached through Instagram mobile feed.
Creative testing clarified what the market responded to. Parent-and-child imagery consistently outperformed child-only variations, the homepage beat alternative landing pages, and “Learn More” confirmed as the strongest CTA within the test conditions.
Google Ads
Campaigns were structured around Target CPA bidding with disciplined budget-to-target ratios, and a dedicated testing campaign was built to protect stable acquisition from experimentation.
Converting search terms were consolidated into tightly themed exact-match clusters. Queries around “business ideas for kids and teens” and “how to start a business for kids” emerged as the highest-intent drivers, giving the account a stronger foundation for keyword expansion.
Results
At a steady $2,000 per week in spend, CPA improved by approximately 60% from the launch baseline while conversion volume increased at the same budget level.
The account also moved out of exploratory testing with a clearer acquisition system: validated audiences, validated creative direction, stronger keyword structure, and a clean separation between scaling campaigns and testing environments. The result was not only a lower CPA but a more stable account architecture that could keep learning without letting experiments contaminate the campaigns already producing results.
Constraints We Navigated
The system was built under real constraints. Limited early conversion data restricted access to robust lookalikes and aggressive bid automation. Testing had to run without disrupting live acquisition performance, which required clean structural separation from day one. Signal accumulation could not be forced, so interim audience sources carried the load while conversion density grew.
Progress was made within those constraints rather than delayed by them.
Why This Worked
The account did not need more variables. It needed cleaner ones.
By separating stable acquisition from testing, each platform had a clearer job. Meta could learn from stronger interim audience signals while conversion data matured. Google could expand around proven search intent without letting exploratory keywords drag down the core acquisition flow.
That structure gave the account room to improve without forcing scale before the signal was strong enough to support it.
