Research Portfolio
A couple of studies done at Meta recently: from uncovering platform frictions to informing 0-1 product launch
Meta · 2025
Meta · 2025
Meta · 2026
Meta · 2026
Meta · 2026
Meta · 2026This work is confidential. Please enter the password to continue.
A couple of studies done at Meta recently: from uncovering platform frictions to informing 0-1 product launch
Meta · 2025
Meta · 2025
Meta · 2026
Meta · 2026
Meta · 2026
Meta · 2026
MetaHigh-intent shoppers who clicked through ads were completing purchases on the system browser instead of inside the in-app browser (IAB). Prior research couldn't isolate the reason because it surveyed all abandoners, and the dominant answer was always "I wasn't ready to buy." The actual friction signal was buried in low-intent noise.
The question wasn't why do users leave? It was why do users who clearly intended to buy still leave?
MetaThe team's existing performance targets were borrowed from industry standards calibrated for users who actively choose to browse the web. But in-app browser users didn't come to browse — they tapped an ad. Their willingness to wait is likely lower than standard benchmarks assume.
The hypothesis: Industry-standard latency targets are too lenient for ad-context users. Behavioral data should reveal tighter thresholds where bounce risk escalates sharply.
MetaA shoppable video feature's CTA click-through rate had declined following a reactive copy change, falling short of the team's performance target right before a major market rollout. The team needed evidence-based guidance on which CTA text would maximize noticeability and engagement intent, set accurate post-click expectations, and remain acceptable to creators.
The stakes: The CTA is the sole entry point for the entire shoppable experience. Every downstream metric depends on users first noticing and tapping it.
MetaVideo Ad Chaining delivers consecutive video ads to high-engagement Reels users. Early revenue metrics were positive, but monetization metrics don't tell the full story. If the ad experience degrades user sentiment or drives disengagement, it could signal long-term retention risks even if short-term revenue improves.
The stakes: If sentiment degradation was real, it could justify pausing or capping the feature. If manageable, the team could scale confidently. Research needed to answer both "is it hurting?" and "how much can we scale before it starts hurting?"
MetaA sibling platform had recently merged their equivalent shopping entry points, projecting significant reach gains. The team faced a high-stakes strategic question: follow suit and merge, or keep the experiences separate?
The reach gains were real and attractive — but the two platforms differ in fundamental ways. Getting this wrong could leave growth on the table or erode a commerce-first experience the team had been deliberately building.
MetaA new shoppable video feature was approaching its first major market rollout, but no systematic quality measurement existed. The team was relying on anecdotal feedback and backend metrics that couldn't capture the human perception gap between "the algorithm thinks this matches" and "a user would agree this is the same product."
The core tension: ML models optimize for algorithmic similarity, but users judge quality by visual and conceptual exactness. Without a structured quality bar, there was no defensible basis for saying the product was ready to launch.