Ads Experimentation Platform
TikTok advertisers needed a better way to validate campaigns before scaling. I led the design of a unified experimentation platform that enables advertisers — especially SMBs — to run A/B tests, analyze results, and confidently scale high-performing campaigns, all without leaving TikTok.
View Creator HubMost advertisers on TikTok lacked a structured way to validate campaign strategies. Small to medium-sized businesses (SMBs) especially struggled with ad spend inefficiency — they were either investing blindly or abandoning the platform entirely due to poor ROI visibility. TikTok needed a system to democratize experimentation.
Business Goal
Increase ads spend revenue and advertiser lifetime value by enabling strategic experimentation that helps advertisers confidently identify winning campaign strategies — moving from guesswork to data-driven decisions.
User Goal
Build a unified experimentation platform where advertisers — especially SMBs with limited resources — can easily design, run, and analyze A/B tests to validate messaging, targeting, and creative approaches without leaving TikTok.
Approach
The experimentation initiative was split across three different teams — Brand Lift Studies (BLS), Conversion Lift Studies (CLS), and Split Testing — each with separate roadmaps and resources. My role was to unify the fragmented experience, understand each team's unique problem space, and align on a cohesive MVP that served all three while maintaining product coherence.
I held a series of workshops with engineers, PMs, and researchers from all three teams to identify overlapping pain points, map user journeys, and establish shared design principles. This collaborative approach ensured every team felt heard while building a unified platform that scaled across all experiment types.
Discovery
Workshops with each team surfaced distinct pain points: BLS users needed brand-impact measurement, CLS users wanted conversion tracking, Split Test users wanted creative comparisons. But everyone needed the same thing — statistical confidence and a clear next step.
Interviews with advertisers across budget tiers exposed the real gap: 73% relied on gut instinct over data, SMBs couldn't afford external testing tools, and most campaigns lacked statistical rigor. The breakthrough — one unified platform for all three experiment types, built around statistical confidence and actionable recommendations.
From there, I ran workshops with each team — individually and together — to shape a vision for a unified Experiment Manager: mid-fidelity flows mapping all three experiment types, plus a roadmap from MVP to advanced statistical models, automated recommendations, and cross-experiment insights.
Unifying Three Experiment Types
The platform needed to support three distinct experiment workflows, each with different measurement needs:
Brand Lift Studies
Measure brand awareness, recall, and brand perception. Requires holdout group methodology and demographic targeting refinement.
Conversion Lift Studies
Track direct business outcomes: clicks, conversions, purchases. Focused on immediate ROI measurement and budget scaling.
Split Testing
Compare creative performance, messaging, and targeting variations. Rapid iteration and A/B testing at scale.
The design challenge was creating one unified hub where advertisers could easily understand which experiment type suited their goal, run it without switching platforms, and interpret results with statistical confidence — regardless of measurement complexity.
Discovery distilled into one vision: a unified Experiment Manager where BLS, CLS, and Split Tests could be created, monitored, and reviewed in a single workflow — turning fragmented tools into one source of truth for campaign success.
The methodology was tiered by sophistication — simple single-variable tests for SMBs, multi-armed bandits and statistical models for enterprise advertisers — balancing rigor with accessibility for every skill level.
Vision to MVP
Post vision formalization, I worked with individual teams to formulate an MVP version, going through iterations for each project — ensuring users, especially SMBs, had clarity, insight, and ease in starting their first experiment.
Second Iteration
A second round of iterations refined the experience further — sharpening flows, tightening visual design, and resolving edge cases ahead of the MVP launch.
The MVP version of EM unified two main experiments — Split Test and BLS — with a unified list view, filter system, and reporting.
The Big Picture
Split Test also existed on its own horizontal track, and I made sure that experience reflected the same cohesive vision as the rest of the project. Working with my cross-functional team, we ensured SMB users had a consistent experience and a strong, seamless transition between the two surfaces.
The experimentation platform unlocked substantial advertiser ROI improvements and platform growth. Advertisers who used the platform scaled spend by average 40%, SMBs increased campaign frequency by 3.5x, and overall platform ad revenue grew significantly as advertisers moved from conservative spending to confident scaling.