WATCHER
A streaming app that doesn't compete on catalog — it competes on behavior. Redesigning discovery for when the user doesn't know what they want to watch.
Context and problem
The real problem
The streaming market grew 74% in platform supply over the last five years — but the user experience stagnated in the 2008 model: a grid of thumbnails, a search bar, rigid categories.
The problem isn't a lack of content. It's the cognitive cost of choosing. Users consistently report one specific feeling:
“I have 30 minutes to watch something. I spent 28 scrolling and closed the app.”
— Interviewee #04, 27, user of 3+ simultaneous platformsThe core hypothesis: the streaming interface needs to respond to emotional state and temporal context, not just historical preferences.
of users scroll without converting
↗ Nielsen, 2023
average time lost in discovery
↗ Interviews conducted
simultaneous platforms per user
↗ Statista, 2023
leave without watching anything
↗ Deloitte Media Report
How might we…
Reduce the cognitive cost of choosing without removing the user's sense of control?
Research
Hypotheses tested
Before the interviews, I documented three explicit hypotheses. That forced me to look for contrary evidence — not just confirmatory.
Hypothesis 01 · Confirmed
Too much content causes paralysis
More filters don't help. Interfaces with many filters increase decision time by 40%. The solution is to have an opinion, not to expand options.
Hypothesis 02 · Refuted
Users want automatic recommendation
I assumed invisible algorithms would solve the problem. But 68% of interviewees reported distrust of recommendations that felt “too automatic”.
Hypothesis 03 · Partly confirmed
Context matters more than history
“What I watch on a Friday night is completely different from what I want on a Monday morning.” — Interviewee #09
Key insights from the interviews
12 in-depth interviews, 45–60 min each. Profiles: ages 22–38, active users of 2+ platforms.
INS 01
Available time changes the choice criterion
With less than 40 minutes, users prefer short episodes — but most platforms don't filter by duration prominently.
INS 02
Search is used as a last resort
Only 22% start the session with search. Most prefer to be “guided” — but reject the feeling of losing control.
INS 03
Thumbnails sell the wrong content
Most can't infer the emotional tone of content from the thumbnail. “I can't tell if it's heavy or light just from the cover.”
INS 04
Personal lists create inertia
Saved lists pile up content that “seemed interesting before” and become a second source of paralysis. 7 of 12 no longer use their own lists.
Design process
Justified decisions
Every interface decision was documented with the question: “why this choice, and what did we discard?”
Wireframes — documented iteration
The wireframe phase isn't a “rough draft”. It's where information-architecture hypotheses are tested before any visual investment.
What v1 revealed: the “streaming finder” structure as a form flow drove abandonment. v2 replaced the form with a three-choice conversation.
Interface decisions
Documenting decisions turns a portfolio into a record of reasoning — not an album of pretty screens.
Decision
Conversational onboarding, not a form
The “Hi! How are you?” screen uses conversational language and offers 3 intent options — not dozens of filters.
Discarded alternative
A form with genre, price and feature fields — it replicated the very paralysis we were trying to solve.
Decision
Emotional tone above the thumbnail
Each title carries visible mood and duration labels, answering the insight that the cover doesn't convey tone.
High-fidelity prototype — main flows
Four screens cover the critical flow: contextual entry → discovery → detail → profile.
01 · Entry
02 · Discovery
03 · Detail
04 · Profile
Design system
Design system
Not just a style guide — a system with documented tokens, accessibility decisions and component usage criteria.
Palette — a grounded decision
Deep black reduces eye strain in the main context of use (dark environment). Yellow #F5D407 reaches AA contrast on dark backgrounds without losing vibrancy.
Typography — editorial system
Display / Brand
UI / Body
Labels / Meta
Spacing
An 8px base unit; all spacing is a multiple of 4px — ensuring consistency at handoff.
Results and learnings
Measured impact
Results from usability testing with 6 participants (n=6), compared to the reference flow of current platforms. As a concept project, these are prototype results — not production.
time to choice
Reduction in decision time
Participants took 6.8 min on average to choose (vs. 18 min on current platforms). Conversational onboarding was the main factor — removing the “what do I want?” step before discovery.
SUS Score / 100
System Usability Scale
A score of 83/100 on the SUS after a free-task test (reference benchmark: ~70 for large platforms). Above 80 — rated “Excellent” on the Bangor scale.
would return to the flow
Stated intent to use
5 of 6 said they'd prefer to use this flow on their current streaming service. The only “no” cited the lack of integration with an existing account — a scope limitation, not a design one.
What I'd do differently
Seniority includes honesty about limitations and iterations not made.
01
Test before high fidelity
I built high-fidelity wireframes too early. Paper testing would have saved 2 weeks of refinement.
02
Document hypotheses before interviews
I went into the first interviews without explicit hypotheses — slower analysis and more confirmation bias.
03
Accessibility at the start, not the end
WCAG was checked at the end. Next project, contrast and motion enter as criteria from the style guide onward.
Takeaway
When the interface responds to the user's state — and not just to the catalog — discovery stops being a task and becomes the pleasure it should be.
Watcher is a concept study, but the method is real: explicit hypothesis, contrary evidence, documented decision.