Every streaming service markets itself on what is in the catalogue. Almost nobody chooses what to watch by browsing a catalogue. They open an application, look at a screen of large images, and pick something from what is presented.
Which means the interface is doing more to determine what gets watched than the library does, and the design decisions inside it are worth looking at directly.
The home row is the whole business
The first row of the first screen is the most valuable real estate in the service. It is seen by everyone who opens the application, on every device, every time.
What goes there is a decision, made partly by recommendation systems and partly by commercial arrangement, and the two are not distinguished for the viewer.
Anything not in the first two or three rows is effectively in the back of the shop. It exists, it is searchable, and it will be found by people who came looking for it specifically.
This is why the size of a catalogue is a poor guide to anything. A service with twenty thousand titles and a service with four thousand present roughly the same number of options to a viewer at any moment.
Autoplay changes what a viewer is
The decision to play the next episode automatically after a few seconds is the single most consequential interface choice any of these services made.
It converts a decision into a non-decision. Continuing requires nothing. Stopping requires an action.
The consequences on viewing time are large and well understood by the people who implement it. The consequences on how programmes are made are less discussed and are real — endings designed to be arrived at rather than chosen, episode structures that assume continuation, less need for a closing that satisfies.
Several services now let you switch it off. The setting is generally several menus deep and off by default.
Thumbnails are personalised and most people do not know
Some services select the image shown for a title based on what they know about the viewer. The same programme presents with a different image to different people.
The stated rationale is that different aspects of a work appeal to different viewers, which is true. The effect is that the image is not a description of the programme, it is a prediction about you.
This has produced genuine complaints — images that foreground a minor character, or imply a genre the work is not, because that framing tests better with a particular segment.
It is a small deception, arguably, and it is the interface making a claim on the content's behalf that the content does not support.
Continue watching is a commitment device
The row of partially watched things does something psychologically specific. An unfinished item creates a mild pull toward completion.
Which is useful when you actually want to finish something and slightly manipulative when the thing was abandoned on purpose. Some services make removing an item difficult, which suggests they know what it is doing.
The related design is progress indication — a bar showing how far through you are. It converts watching into completion, which is a different activity with a different motivation.
Search is deliberately unrewarding
An observation that took me a while to be confident about. Search on most streaming services is worse than search on almost any other kind of application.
Partial matches, poor handling of misspellings, results that include things not in the catalogue as a route to upselling, and no useful filtering by the criteria people actually care about — year, country, length, whether it is finished.
I do not think this is incompetence. Search is how a viewer exercises intent, and intent bypasses the promotional surface. A viewer browsing sees what the service wants to show. A viewer searching sees what they came for.
The services with the best search tend to be the ones whose business is not built on directing attention within the catalogue.
What good interface design would look like
Some of this exists in places and is worth naming.
Filtering that actually works — by length, by year, by whether a series has concluded, by whether it is leaving soon.
Honest thumbnails and honest genre labels.
A visible, easily accessible list of everything, browsable in a way that does not depend on recommendation.
Autoplay off by default, or at least a setting that is easy to find.
And a leaving-soon list, prominently, since the viewer's interest in that is obvious and the service's is not.
The version I use
I keep a written list of things I intend to watch, outside any service, and I search for them directly.
It sounds unnecessarily deliberate and it has changed what I watch considerably. Almost nothing on that list would have been surfaced to me by a home screen, and I have enjoyed more of it than I did of what the home screen used to give me.
Which suggests the recommendation is optimising for something other than my satisfaction, or is optimising for a version of me that is tired at nine in the evening and will accept whatever is offered.