Built from what you do in the app rather than a questionnaire. It tracks:
Getting started
- 01
Set it up once
Onboarding is two steps: rate a handful of titles you have already seen, and tick the streaming services you actually pay for. That is the whole setup — everything after this the app learns from what you do.
- 02
Ask for a pick
On Discover, choose Film, Series or Surprise me. Open the mood panel to add how much time you have, what kind of night it is, and whether you are watching alone or with people. You get one main pick with a match score and a written reason, four alternatives beside it, and a longer ranked list underneath. Refresh as often as you like — you get a different set each time.
- 03
Tell it when it is wrong
Thumbs up and down on anything, and Not tonight to hide a title for a month without recording it as a dislike. Explore has a swipe deck for rating a lot of titles quickly. Dislikes teach it as much as likes do.
- 04
Keep the record
Watchlist, watch history down to the episode, your own lists and collections, ratings and reviews. One thing worth knowing: anything on your watchlist is excluded from recommendations, so a watchlist you never clear quietly shrinks what you get offered.
How it works
Your taste profile
Built from what you do in the app rather than a questionnaire. It tracks:
- ·Genres, weighted by how strongly you respond to each one
- ·Actors and directors that keep recurring in what you like
- ·Tone — light or dark, funny or serious, comforting or unsettling
- ·Pace and runtime you actually finish
- ·Themes and keywords your favourites have in common
- ·What you finish, what you abandon, and what you never start
How a pick gets chosen
Candidates are gathered from several sources, then scored on six signals. The weights shift once you pass around 15 rated titles, which is when the collaborative signal has enough to work with:
- Taste match — how close it sits to what you already like. 36%, or 50% before the engine knows you.
- Context — your mood, the time you have, and whether you are watching alone or with people. 18%.
- People with taste like yours — what they watched and rated. 15%, and switched off entirely until you have rated around 15 titles.
- Quality — ratings, and how many people stand behind them. 14%.
- Session and freshness — what you have been reacting to in the last few minutes (12%) and how recent the title is (5%), then a novelty multiplier that stops five near-identical picks in a row.
It keeps learning
Every rating, swipe, watch and review updates your profile, and the next request uses it. Nothing needs saving or refreshing. Settings shows a confidence score for how well it thinks it knows you — expect that to be low for the first week.
Getting better results
Rate more than feels necessary
Rating things you have already seen is the fastest way in, and you do not have to watch anything new to do it. Around 15 ratings is where the collaborative signal switches on and picks noticeably sharpen.
Keep your services current
CHOICE only offers titles you can actually stream, so an out-of-date list in Settings is the single biggest cause of thin results. Update it when you subscribe or cancel, and check your country is right.
Say what the night is for
The mood panel takes ten seconds and changes the answer a lot. Ninety minutes alone on a weeknight and a whole evening with friends should not produce the same pick.
Use Not tonight
When a pick is fine but not right now, Not tonight hides it for a month without teaching the engine you dislike it. Using dislike for that is the quickest way to distort your profile.
Clear your watchlist
Watchlisted titles are excluded from recommendations, so thirty stale entries are thirty things you will never be offered. Explore has a triage view for working through them quickly.
Give it a week
The first few picks are working from very little. It gets noticeably better after 15 or 20 ratings, and better again once there is watch history to read.
Common questions
Still stuck?
Email support@choiceapp.org and you will get a reply from the person who built the app. During the beta that is the fastest route for anything broken, confusing or missing.
CHOICE is in open testing. Bug reports and blunt feedback are genuinely useful — especially when a recommendation is bad, since that is the one thing that cannot be tested from the inside.