The design tested well by feel. The data told a different story.

Where it started

"How might we help learners stay focused, without policing them?"

YouTube is great for learning but easy to drift on. Our course project asked: how might we help learners stay focused on a single video session without policing them or breaking the YouTube mental model? I worked on this as a four-person team with Utkarsha Nagtode, Sanjith Kumar, and Priyanka Choudhary, across three phases: building and testing the prototype, then a retrospective that shaped two new proposed experiments.

• Reduce distraction: lower "off-topic" thumbnail clicks and premature exits.
• Clarify intent: make it obvious what Focus Mode is and when it's on.
• Keep control: keep necessary actions visible without clutter.

Concept in one breath

FocusMode+, a lightweight, opt-in layer that you toggle on, which cleans the UI, highlights "Focus-approved" elements, shows a brief onboarding nudge, and adds a gentle, persuasive exit dialog so leaving is a considered choice, not a reflex.

"We set our success metrics before we designed anything."

• Observed

  Problem framing came from listing our own distraction moments plus peer reports: the impulse to scroll, an irresistible side-rail, vague task cues going in.

• Designed

  We defined success metrics up front, before any screens existed: completion rate up, "off-topic" click-through down, early exits down, and a basic comprehension check ("What does Focus Mode do?").

Prototyping for A/B testing

Two prototypes, one deliberately minimal

We built two Figma prototypes to test the concept cleanly against a baseline.

• A. Control case: an exact replica of YouTube, unchanged.
• B. Target case: the same interface with three additions, a Focus toggle, in-focus labeling, and an exit check.

• Focus toggle only: a simple Focus Mode switch added to the YouTube UI, nothing else changed.
• In-focus labeling: while in Focus Mode, the target video (and videos expected to be watched) carried a small "Focus mode" badge.
• Exit check: clicking out to a video that wasn't marked, or out of Focus Mode itself, surfaced a single confirmation dialog nudging the person to reconsider.

How it works

Interaction model

Click a card below to see each stage of the flow.

• Enter +
• During +
• Exit +

A/B testing

Maze setup

Flow tested: browse → enable Focus → watch → encounter distractions → decide to stay or exit.

Captured: completion, "off-topic" clicks, exit rate and decision time, plus a one-question understanding check.

Our hypothesis: users interacting with the FocusMode+ variant would complete 30% more goal-oriented watch sessions and show 40% fewer off-topic clicks than those on the standard YouTube interface.

• Targeted: 20–30 students and professionals who regularly consume YouTube tutorials and educational content, recruited through in-class participation and peer circles.
• Recorded: ages 25–34, mixed genders, frequent users (most several times a day), watching tutorial/educational, entertainment/vlog, and music content, a good match for the "learning on YouTube but getting sidetracked" problem.

Bookended by a pre-task survey (usage and platform familiarity) and a post-task survey (experience interacting with the prototype).

What the Maze A/B test told us

• Inferred

  Interest exists, clarity doesn't, yet. Concept was strong, distinctiveness was weak, most wanted the feature, few could find it.

• Observed

  Behavior signals confusion. Success hovering around 55%, with a high misclick rate, points to unclear affordances and underpowered visual cues, not a bad idea.

• Observed

  Relevance is fine, temptation is strong. 63% felt suggested videos were relevant, and the side-rail remains a powerful distraction unless Focus Mode actively counteracts it.

What we heard

• Helpful: the Focus badge made "which video to pick next" clearer for some, having the toggle visible felt empowering.
• Missing or unclear: several people didn't notice the mode, kept toggling without feedback, or expected the side rails to shrink or hide.
• Requested: a bigger player area, de-emphasized related videos, and variants like a "Today Mode" showing only freshly uploaded content.

Phase 3, based on the Maze results

Two experiments we'd run next

Once the numbers were in, we ran a team retrospective before designing anything new.

• Instructions were too vague, people never learned the video was meant to be educational, which made "off-topic" hard to even define.
• Focus Mode started already turned on, so some people never registered it existed at all, and without a plain YouTube baseline to compare against, our conclusions stayed weak.
• The design wasn't visually distinct enough, which is a large part of why 71% of people exited it during the task.

That critique is what shaped both experiments below, letting people turn the toggle on themselves instead of starting it silently, and replacing a flat warning with something more persuasive at the exact moment someone's about to leave.

• A: Onboarding after a user-initiated toggle +
• B: Social-proof exit dialog +

Review Detailed Report →

What I learned

• Clarity plus persuasion. Make the state unmistakable, and nudge at the decision moment. A tiny onboarding burst and clear visual distinctiveness reduce confusion and cognitive load, then, when someone tries to exit, a polite, evidence-based nudge beats a generic warning.
• Design for drift. The side-rail is powerful. Making the "stay with this video" path obvious, badges, reduced visual noise, stable progress cues, helps people commit without feeling restricted.
• Test, don't guess. We set success metrics before design: completion, off-topic CTR, exits. For the follow-up experiments, that means declaring primary metrics, estimating required sample size, running long enough to smooth weekly cycles, and not peeking early to call a winner.

Maze Live Study → Maze Report →