A Second Start for Streaming

Cafe Bazaar Video brought movies and TV series into Cafe Bazaar, an Android app store in Iran. Starting in 2024, I designed its relaunch and later redesigned the player, Continue Watching, and Video Detail.

Cafe Bazaar had operated a video service before, leaving designs and code we could reuse. With a tight deadline and limited resources, I worked with product to narrow the feature set and build on those existing parts.

I designed the journey across Discovery, Video Detail, and Player. The first release supported browsing, choosing a title, and playing it, including basic season and episode selection. Reuse helped us get the service live, but the mixture of inherited and newly built parts left inconsistencies and made later development difficult.

In mid-2025, a planned move toward a web-based implementation created an opportunity to rebuild the player. The existing version lacked a consistent component structure and layouts suited to different screens. I redesigned the web player with the product manager and frontend engineer, using Cafe Bazaar’s design foundations.

I preserved the familiar small-screen arrangement, with large, central playback controls suited to touch. On desktop, I designed for mouse interaction by moving playback controls to the bottom and putting settings in menus.

Continue Watching’s first release followed a familiar pattern from other video services. Evaluating it showed that more continuation did not necessarily produce more watch time.

The first experiment found continuation in 33% of sessions in the Continue Watching group versus 26% in the control group, with no statistically significant difference in daily watch time. Viewers without the row found unfinished titles through Search or Discovery and resumed manually. The data team found that multi-session journeys accounted for 44% of sessions and half of total watch time.

A product review found missing or duplicate titles, completed films that remained visible, incorrect time information, and unclear entry rules. These were concrete failures in how the feature represented unfinished viewing.

For the second iteration, which followed the player redesign, I worked with product, data, and engineering to define how unfinished content should behave across the viewing journey. Two decisions shaped the requirements and interface.

Entry and exit rules

I specified rules for when a title should enter and leave Continue Watching. Entry required 5% of the content to have actually played; moving the seek bar forward would not qualify it. Exit rules covered reaching 95% of the timeline, 90 days of inactivity, manual removal, and unavailable content.

We did not have reliable intro and credits timing, so the percentage thresholds were provisional rules the service could support. They gave the feature explicit entry and exit conditions that we could revisit when better content timing became available.

Progress across screens

A returning viewer could encounter the same title in Discovery, on its detail page, or inside the player. I defined how unfinished state should carry across those surfaces, including progress indicators and actions such as Continue S2 E3.

Individual episode titles were not always reliable. I used the series name with season and episode numbers to identify the content consistently, accepting less descriptive labels in exchange for dependable information.

Results: second release

After the second iteration shipped, all three outcome measures improved:

• Successful resume rate increased from 33% to 37%.
• Completion of resumed titles increased from 40.5% to 43%.
• Seven-day VOD return increased from 42.6% to 43.2%.

Viewers needed enough information to decide whether a title was worth watching, but the Detail page gave those cues limited space. In my usability sessions, five of six participants used posters or actors to guide their choice, and four read descriptions. Earlier interviews with active viewers showed similar behaviour. I compared the usability findings with heatmaps and a heuristic review.

I prioritised a clearer structure using the content already available: prominent artwork, grouped metadata, and descriptive components in place of informational chips that looked clickable. Trailers and genre exploration remained outside the initial restructuring scope.

The work reached handoff, but company priorities and team ownership changed before launch. Its user and business impact remained untested.

The Continue Watching experiment changed how I interpret engagement results. Viewers were already resuming manually, so a lack of additional watch time did not mean a lack of need. I would now check the accuracy of remembered viewing state before using engagement alone to judge a continuation feature. For redesigns such as the player, I would establish the baseline and follow-up measurement before launch, especially for the users whose experience would change most.