In August 2018, Flatstudio started exploring a question most sportsbooks were still sidestepping: what happens when a sports product is designed around ongoing relevance, not just the moment of placing a bet?
The result was Rewind. It never shipped. But it made a fairly clear argument about where sports products were heading — and why most of them weren't structurally ready to get there.
The problem wasn't the interface
The standard critique of sportsbooks at the time was that they were visually overwhelming. Too many markets, confusing layout, hard for newcomers to parse. That was all true. It was also the wrong diagnosis.
The deeper problem was behavioural. Betting-only products had weak retention because they were built around a transaction, not a relationship. Users arrived with intent, placed a bet, and left. Nothing held them between events. No engagement loop that worked on a quiet Tuesday.
There was also a paradox: the same product was simultaneously too complicated for casual users and too shallow to hold engaged fans. Rewind was designed to answer both sides at once, which is the harder version of the problem.
The feed was the product
Rewind was built as a sports ecosystem — personalised feed, match centres, team and player profiles, community and media layers, betting — rather than a sportsbook with content bolted on.
The distinction matters. When personalisation is cosmetic, it sits on top of a static navigation structure. In Rewind, the feed was the structural core. What users followed determined what the product showed them. Removing it wouldn't have degraded the experience — it would have eliminated it.
Most sportsbooks assumed users already knew where things lived. Rewind assumed the product should do more of that work.
In dense sports products, clarity is not just a usability issue. It is also a trust issue. If users cannot quickly understand what they are looking at, they hesitate — and hesitation in real-time environments usually means drop-off.
Underneath that was a simple constraint: any piece of information a user might need should be reachable in roughly two interactions. Not because two-click rules are sacred, but because retrieval costs compound quickly in dense products. Once users have to stop and decode structure before they can act, engagement starts leaking.
The bot wasn't a support widget
The conversational layer is the decision most likely to be misread in retrospect, so it's worth being specific.
It wasn't customer service. It was designed as a retrieval layer embedded inside the product — including inside conversations between users. The idea was that it could surface relevant facts mid-discussion, contextualise statistics, or answer quick questions without anyone leaving the app to search elsewhere.
It also addressed a genuine usability problem: betting slip complexity. Accumulator logic, potential returns, market-specific rules — real barriers for less experienced users. The bot was designed to lower that barrier contextually, without static help documentation.
The logic was simple: access through intent rather than menu depth, with context preserved across the interaction. Today that gets described as agent-like. In 2018, it was just a better way to reduce retrieval cost.
What happened afterward
Rewind did not sit in a portfolio archive. Its logic resurfaced in later engagements, parts of it proved reusable, and the design files themselves were eventually sold in 2023.
The clearest adaptation came in Alta SyncReplay. Different domain, same class of problem: multiple live data streams, high-stakes decisions, fast retrieval, and key moments that need to be reviewed and contextualised in real time. The surface changed. The underlying challenge didn't. That's usually a reasonable indicator that the thinking was structural rather than category-specific.
As for the market: DraftKings, FanDuel, and eventually ESPN Bet all moved toward broader sports-fan positioning, deeper personalisation, and engagement logic that extended beyond the bet. These are large organisations with their own research and their own forcing functions. Different scale, different incentives, same directional shift: the product had to become broader than the bet itself.
Why it didn't get built
Several later attempts to turn Rewind into a real product ran into the same wall: live data infrastructure, content licensing, real-time engineering at scale, and sports data economics that are genuinely not forgiving.
The honest version of the story is that the concept was directionally right before it was economically convenient. The idea preceded the conditions that would have made building it tractable. That's not a failure, but it's a fact worth naming.
Rewind was an attempt to solve a structural problem — how to build a sports product that was accessible to casual users and genuinely retentive for engaged ones. The answer it arrived at was a product architecture answer, not a visual one. The visual work followed from structural decisions, not the other way around.
That's the kind of problem worth designing for: products where the real challenge is what to surface, when, and why — and where getting that wrong has behavioural consequences that no amount of visual polish will fix.
It's a bridge between a moodboard and a mockup. It shows how your brand feels (textures, colors, typography) in one wide image. It ensures we align on the visual direction before spending time on the logo.
It's a bridge between a moodboard and a mockup. It shows how your brand feels (textures, colors, typography) in one wide image. It ensures we align on the visual direction before spending time on the logo.
It's a bridge between a moodboard and a mockup. It shows how your brand feels (textures, colors, typography) in one wide image. It ensures we align on the visual direction before spending time on the logo.
.png){kind=small}
