Saved for Later

The problem

Watch Later has barely changed since 2010.

It is still one flat list. No search, no tags, no reminders. On a product with 2.7 billion monthly users, that makes the gap a positioning choice, not an unknown problem.

Bookmarking research has said the same thing since the early 2000s. People create far more saves than they ever revisit, and intentions fail without a cue at the right moment.

A UX audit across six stages showed Discover and Save are already easy. Revisit barely exists.

Research

People save for a reason. They can’t find or finish it.

• 63 survey responses, 56 of them savers

• 16 / 56 watch less than 10% of what they save

• 55% say too many items pile together

• 3.5 / 5 average ease of finding a saved video

45% want search inside saved videos; 43% want a note on why they saved something.

Seven interviews, one pattern

6 of 7 interviewees described forgetting something they had saved. Three of them, separately, suggested the same fix: put Watch Later on the homepage.

I went in assuming organisation was the main problem. The interviews showed that visibility and retrieval cues mattered just as much.

“Our attention span is very less nowadays. Scrolling 3,000 videos would be very cumbersome.” P5, with 3,639 unwatched videos

Who it’s for

Two saving styles. One system that supports both.

• Dina, the Intentional Archivist. Curates playlists on purpose and deletes what she has watched. Still forgets. Needs search and structure inside one collection.

• Wilford, the Passive Accumulator. Saves on impulse and never organises. His Watch Later balloons. Needs saved videos to come back to him on their own.

Approach

Six directions explored. Three carried forward.

The chosen system is contextual retrieval, built from three parts that work together:

• Structure. Search and categories inside saved content.

• Resurfacing. Saved videos show up at useful moments.

• Context. A note or status, so you remember why it mattered.

I cut manual folders (no external cue), swipe-to-triage (clears the pile but doesn’t stop it forming) and AI resurfacing (takes the person out of the loop). AI moved to future scope, once people trust the manual system.

Solution

From wireframe to shippable, in YouTube’s own visual language.

Global search shows already-saved videos first, marked with a bookmark.

Reminders are contextual, never scheduled. They trigger on save age plus inactivity, and the copy states the reason: “saved 3 weeks ago, still unwatched”. One per session at most, so it reads as a nudge rather than a notification.

Validation

One drag model, one confirmation pattern, on every device.

• 5 moderated walkthroughs

• 4.6 / 5 ease of use, against a 3.5 baseline from the survey (directional)

• 4 of 5 hit drag or swipe friction, each in a different way

• Drag-and-drop failed differently every time. Fix: one hold-to-lift drag model across devices, with a visible drop target.

• Swipe doesn’t exist on desktop. Fix: hover-revealed dismiss and watch buttons replace swipe on non-touch devices.

• Saves didn’t reliably confirm. Fix: one confirmation pattern everywhere: a toast, visible placement and a live count.

“The prototype is good, simple to the eye. It is not too cumbersome unlike the regular YouTube.” U3, usability test

Reflection

A number tells you where to look. A session tells you why it hurts.

16 of 56 people watching less than 10% of their saves pointed at the problem. It only felt real when P5 described scrolling past 3,639 unwatched videos without complaint. Resignation doesn’t show up in a percentage.

What I’d pilot first: the homepage resurfacing card on its own. It is the cheapest piece to test and the one most likely to move revisit rate. If it doesn’t, the rest of the system needs rethinking before it’s built.