Agentic Experimentation

The Challenge

Advertisers frequently receive experiment recommendations, but only a small portion convert them into live tests. Setting up an A/B test requires duplicating variants, configuring parameters, and navigating complex workflows, creating significant friction between intent and action.

The Solution

I designed an agentic workflow that turns recommendations into pre-configured A/B tests. Instead of manually building experiments from scratch, advertisers review key decisions while the system handles setup, configuration, and execution.

40% of Recommendations

AI-powered experiment recommendations account for nearly 40% of all MAIBA recommendations.

28% Higher Adoption

Identified a major drop-off in experiment creation and redesigned the workflow to reduce setup friction.

+0.035% iRev

Experiments launched through MAIBA generated measurable incremental revenue impact.

+2.437% WAU

MAIBA-driven experiments achieved significantly higher advertiser engagement compared with other recommendation types.