Penn Entertainment
Search & Discovery
Built ESPN BET's search experience from zero — cutting zero-result queries by 35%, nearly doubling search's share of total handle, and growing bet conversion through two years of iteration and a ground-up redesign.
Built for Bettors Who Know What They Want
ESPN BET launched without a search experience. I built one from scratch — iterated it for two years — then redesigned it entirely. Search went from a missing feature to a meaningful driver of handle and betting volume.
Context
By 2023, search had become table stakes in sportsbook products. Competitors had it. Users expected it. We didn't have it.
When a user arrives with intent — a team name, a player, a game — search is the fastest path from that intent to a placed bet. Without it, every one of those sessions defaulted to browse. Some users found what they wanted. Many didn't. The absence of search wasn't a UX gap, it was a conversion gap that touched every type of bettor, casual or experienced.
The timing added another layer of complexity. Penn Entertainment was simultaneously migrating Barstool Sportsbook off third-party vendors onto a proprietary tech stack — engineering was fully absorbed in platform work. Building search wasn't just a product priority, it was the primary lever I had to keep users engaged while the foundation was being rebuilt underneath them.
The Build: MVP → ESPN BET Launch → Redesign
2023: MVP & Launch
I owned search end-to-end from the start: scoping the MVP, defining relevance requirements, partnering with engineering on indexing and ranking logic, and establishing the measurement framework we'd use to iterate post-launch.
The MVP shipped first, ahead of the rebrand. After bug fixes and relevance tuning, a full launch followed later in 2023 as part of the ESPN BET rebrand window — one of the more operationally complex moments in Penn's product history — without slipping the launch timeline.
2023–2024: Iteration
Post-launch I established four core metrics to track search health and ran iterative improvements against them each release cycle:
| Metric | Before | After | Change |
|---|---|---|---|
| Zero-Result Rate | 20% | 13% | -35% |
| Browse-to-Add Conversion | 18% | 23% | +28% |
| Single Search Frequency | 28% | 23% | -18% |
| Browse-to-Click | 52% | 53% | +2% |
The zero-result rate drop came from a focused relevance tuning pass. One in seven search sessions that previously returned nothing now returned a result. Browse-to-Add growing from 18% to 23% meant users weren't just finding results — they were acting on them.
I was then pulled off search to lead other high-priority initiatives. Search held its gains.
2025–2026: Redesign
In 2025 I returned to search with a redesign mandate. The core new capability: users could now bet directly from search results without navigating away. The redesign launched in January 2026 on theScore BET.
What the Redesign Moved
The results validated the investment.
Volume & Handle
| Metric | Change |
|---|---|
| Bets placed within 1 min of search | +41% |
| Handle from search-driven bets | +59% |
Handle growing faster than bets signals users are finding and placing higher-value wagers through search, not just more wagers.
Search Share of Total Product
| Metric | Change |
|---|---|
| Share of all bets placed | Nearly doubled |
| Share of all handle | Nearly doubled |
Both metrics trended consistently upward through February — this looks like a durable behavioral shift, not a post-launch spike.
Primary KPIs
| KPI | Goal | Change |
|---|---|---|
| Search-to-Add Conversion | 30–35% | +27.5pp |
| Unique Searchers | Growth | +16% |
| Bet Placement Cannibalization | Additive | Remained additive |
Search-to-Add conversion was the headline KPI. The bet-from-search feature collapsed the funnel and the conversion rate reflects it. Cannibalization remained additive throughout, confirming search is growing the pie rather than redistributing it.
What I Learned
Search in a sportsbook is a different problem than search in most consumer products. The catalog is enormous, highly time-sensitive, and user intent is almost always transactional. A user searching "Chiefs" at 1pm on a Sunday isn't browsing — they want to place a bet in the next five minutes.
The redesign proved that the biggest remaining lever wasn't relevance — it was friction. Once users found the right result, the question became: how many taps stand between that moment and a placed bet? Removing that gap drove the conversion lift more than any ranking improvement had.
The broader lesson: search isn't a discovery feature. In a sportsbook it's a revenue surface. Treating it that way — measuring handle, not just clicks — changes every prioritization decision downstream.