Early accessPilots open

Personalisation that explains itself.

The Boxlist Taste Engine scores your catalogue on how titles feel, learns each user in two minutes, and returns a match and the reasons why on every title. No viewing history needed.

  • Works from day one: no cold start
  • No names or emails
  • You stay data controller

Your streaming app · “Because it’s you”

  • Anatomy of a Fall
  • Past Lives
  • Oppenheimer
  • Dune: Part Two

Why Anatomy of a Fall? Intricate and stirring: the strongest matches for this viewer.

POST /v1/taste-match

{ "user": "u_8f2c",
  "titles": [ …your row… ],
  "sort": "match" }

Pilot API · not public yet

Already live, at consumer scale

15

characteristics scored per title

5 formats

film, TV, books, games, podcasts

35,310

titles on boxlist.com

MCP

works in Claude, ChatGPT, Meta Muse and Claude Code

Why it matters

Your users don’t lack choice. They lack a reason to press play.

  • Cold start

    New users and new titles have no history, so the first week is generic and many people never come back.

    A two-minute taste test, or a few liked titles, gives every user a profile before their first play.

  • Black-box picks

    “Because you watched X” doesn’t say why. Users don’t trust what they can’t understand.

    Every match comes with plain-English reasons you can show on screen.

  • Siloed formats

    A viewer’s taste in TV says a lot about the books and games they’d love, but your systems can’t connect them.

    One profile matches across film, TV, books, games and podcasts.

Live demo

Same row. Three viewers. Three different front pages.

Pick a viewer, switch on Taste mode, and watch one “Popular films” row re-rank, with the API call that does it.

Viewer

Popular films, re-ranked for a Dark Horse

Best match first, with a reason on every title

  1. 1
    Anatomy of a FallPerfect

    ✓ intricate, stirring

  2. 2
    Past LivesPerfect

    ✓ stirring, heavy! wholesome

  3. 3
    OppenheimerPerfect

    ✓ stirring, heavy! informative

  4. 4
    Dune: Part TwoStrong

    ✓ heavy, fresh! big & varied

  5. 5
    BarbieStrong

    ✓ fresh, entertaining! big & varied

Sample viewers. No visitor data is used.

View the API call

The call behind it

Pilot API · not public yet

POST /v1/taste-match

{
  "user": "sample_dark_horse",
  "fingerprint": {
    "pace": 2.5, "originality": 4,
    "wholesomeness": 2.5, "weightiness": 4,
    "realism": 3, "intricacy": 4.5,
    "linearity": 3.5, "styleAndVisuals": 3,
    "scale": 2.5, "emotionalPunch": 4.5,
    "structuredness": 3, "serialisation": 3,
    "informativeness": 2, "timelessness": 3,
    "personalityPull": 3
  },
  "titles": [
    "tt15398776", "tt15239678", "tt1517268",
    "tt1745960", "tt12584954", "tt13238346",
    "tt6166392", "tt17009710"
  ],
  "sort": "match"
}

200 OK

{
  "tasteType": "Dark Horse",
  "results": [
    {
      "title": "tt17009710", "band": 1,
      "matches": ["intricate", "stirring"],
      "clashes": []
    },
    {
      "title": "tt13238346", "band": 1,
      "matches": ["stirring", "heavy"],
      "clashes": ["wholesome"]
    },
    {
      "title": "tt15398776", "band": 1,
      "matches": ["stirring", "heavy"],
      "clashes": ["informative"]
    },
    // + 5 more, best match first
  ]
}

The product

One Taste Engine. Three building blocks.

Start with one, add the others when you’re ready. They all share the same profile and the same scored catalogue.

People are profiled on 15 scales, and every title is scored on the same 15 characteristics, so the two can be matched directly.

  1. 1Learn

    Taste Profiles

    Know every user from their first session.

    • Taste test widget

      An embeddable two-minute test, themed to your brand

    • Seed from liked titles

      Already know a few favourites? Skip the test

    You get

    A Taste Fingerprint™ on 15 scales, plus a Taste Type

    • POST /v1/profiles (proposed)
  2. 2MatchStart here

    Taste Ranking

    Re-rank anything you already show, and say why.

    • Taste Match

      Any row, search result or email, best match first

    • Similar in feel

      “More like this” by feel, not genre or cast

    • Watch together

      Picks for two or a household. Each side sees bands only

    You get

    A band (Perfect match to Not for you) and reasons on every title

    • POST /v1/taste-match (pilot)
    • /similar (proposed)
    • /together (proposed)
  3. 3Understand

    Taste Intelligence

    See your catalogue and audience the way your users feel them.

    • Catalogue enrichment

      Every title scored on 15 characteristics and moods, from your own licensed metadata

    • Audience insights

      Aggregated, anonymous taste profiles, and the gaps in your catalogue

    You get

    An enriched metadata feed and an insights dashboard

    • GET /v1/catalogue/{id}/scores (proposed)

Ship it your way

  • Drop-in widgets

    Taste test, Taste mode switch and match badges.

  • REST API

    JSON in, ranked titles out. You choose where results appear.

  • AI agents (MCP)

    Give your assistant or chatbot a sense of taste in one config line.

Use cases

Built for anyone with a catalogue to sell.

Fix cold start and blank rows.

  • A taste test in onboarding gives every new user a profile before their first play
  • Personal rows and moods, with a reason on every title
  • Watch-together picks for shared and household accounts

Pilot KPIs

  • Onboarding completion
  • Row click-through
  • Time to first play

Row: “Because it’s you”

  • Anatomy of a Fall

    Film · intricate, stirring

    Perfect
  • Past Lives

    Film · stirring, heavy

    Perfect
  • Oppenheimer

    Film · stirring, heavy

    Perfect

Real titles, matched for a sample Dark Horse.

Taste Intelligence, in public

Here’s what our scores already say about the big services.

Imagine this for your own catalogue, and your own audience.

All Boxlist analyses →

Source: Boxlist data analyses, updated November 2026.

How a pilot works

From first call to a measured result.

  1. Step 1

    Scope call

    Agree the surface (one row, onboarding or email) and the KPI that defines success.

  2. Step 2

    We score your catalogue

    Send your licensed metadata. We score it and send back a taste audit.

  3. Step 3

    A/B test live

    Widget or API on a share of your users, against your current ranking.

  4. Step 4

    Readout

    Results against the agreed KPI. Roll out, extend or walk away.

What you need: A metadata feed · Pseudonymous user IDs · A developer for one surface

Pricing

Start small. Pay as it proves itself.

Every pilot starts with a free taste audit of your catalogue.

  • Most teams start here

    Pilot

    Fixed price, agreed up front

    One surface, one KPI, a clear answer.

    • Free taste audit of your catalogue
    • Catalogue scoring
    • Taste Match and the widget
    • Results readout
    Book a pilot call
  • Growth

    Priced per monthly active profile

    For rolling out what the pilot proved.

    • All Learn and Match blocks
    • Catalogue kept up to date
    • Widgets, REST API or MCP
    Talk to us
  • Enterprise

    Custom

    For multi-app bundles and large catalogues.

    • Everything in Growth
    • Taste Intelligence dashboard
    • One profile across apps
    • Custom data processing terms
    Contact us

For developers

Try it before you call us.

The free Boxlist MCP server runs on the same engine. Add it to any MCP client in one line:

claude mcp add --transport http boxlist https://mcp.boxlist.com/mcp

Privacy by design

Your users stay yours.

  • We never sell personal data. Your users’ taste profiles stay yours.
  • UK and EU GDPR: you are the controller and Boxlist acts as your processor, under a data processing agreement.
  • Fingerprints need no name or email: a pseudonymous id is enough.
  • Joint taste shares bands only. Neither person sees the other’s answers or scores.
  • Audience insights are aggregated and anonymous.

Questions buyers ask

How is this different from collaborative filtering?

Collaborative filtering needs lots of viewing history and learns “people who watched X also watched Y”. The Taste Engine matches on how a title feels across 15 characteristics, so it can explain every pick and works without viewing history. Run it alongside your current recommendation API, or use it to re-rank its output.

Does it solve cold start?

Yes, for users and titles. A two-minute taste test or a few liked titles gives a new user a profile straight away, and new titles are scored from their metadata, so they can be matched from the day they land.

What data do you need from us?

Your catalogue metadata and a pseudonymous user ID. No viewing history, names or emails. You stay the controller of your users’ data, with Boxlist as your processor.

Can it use our own catalogue?

Yes. We score it from your own licensed metadata and keep the scores up to date as titles are added. Films, TV, books, games and podcasts can share one profile.

How long does integration take?

It depends on the surface. The widgets drop into a page, and the API returns JSON for one surface at a time. We agree the scope, and the developer time it needs, on the first call.

See it on your catalogue.

Tell us a little about your service. We’ll come back with a pilot plan and a free taste audit of your titles.

Opens your email app with a message to team@boxlist.com. We’ll only use it to reply. Privacy policy

Pilot it on your catalogue

Free taste audit included

Book a call