Paramount+ asks for your birthday on a spinning dial before it shows you a single frame of television. Thirteen screens stand between a cold visitor and their first play, and the step that is supposed to make the product feel personal is a grid of posters with a counter that reads NEXT (0). We read the whole flow, then rebuilt it as seven screens with one AI step that asks what you love in your own words.
Before anything else: Paramount+ is not a client of ours and we are not affiliated with them. This is a self-initiated rebuild. The teardown works from thirteen screenshots of the shipped flow, and every reimagined screen is our own hi-fi prototype. We do these because the fastest way to show what AI actually changes about a product is to take a flow millions of people already know and rebuild it in the open.
Personalization That Starts at Zero
Here is the whole problem in one screen. Near the end of signup, after the plan sell and the account form and the payment, Paramount+ presents a personalization step: pick at least three shows to get recommendations. It is a poster grid. The button at the bottom reads NEXT (0).
That zero is doing a lot of work. It tells you the product knows nothing about you, that the burden of teaching it is yours, and that the teaching happens by tapping pictures. You have just answered thirteen screens of questions, including your birthday and your ZIP code, and the system still starts from nothing.
This is the gap between collecting data and understanding a person. Forms extract. They do not learn. A signup can ask you for six fields and still have no idea what you want to watch tonight, because none of those fields are about watching.
What the Teardown Found
We had Claude read all thirteen production screens end to end and map the friction step by step, so the brief came from evidence rather than a hunch. Four patterns came out of it.
The anchor and the override. The welcome screen promises plans from $8.99 a month. The very next screen preselects the $13.99 Premium plan and badges it MOST POPULAR. The benefits list itemizes Premium only; Essential is never spelled out, and its most material limitation lives in a 12px footnote. Selection is signalled by color alone, with no radio or checkmark.
Six fields to watch TV. Full name, email, password, birthday, gender, ZIP code. The optionality legend is inverted: a gray dot marks optional fields, reversing the asterisk convention every user already knows, and the legend sits below the form. Marketing consent arrives pre-checked. Password requirements are never stated.
A birthday on a dial. The date of birth is a three-column wheel picker. Most adults have to scroll between twenty-five and sixty year positions to enter a number they could have typed in four keystrokes.
A checkout with no order summary. The pre-purchase screen has no plan name, no billing cadence, no renewal date, and no payment method in the interface. The $13.99 figure appears only inside a paragraph of auto-renew legal copy.
The Rebuild: Read, Ask, Prove
The concept is seven screens, and the AI does the work the form used to. We think about it in three moves.
Read. The teardown itself was the first AI job. Thirteen screens in, a mapped friction list out, with each problem tied to the screen that causes it. That is what set the brief, and it is the same pass we run on a client's funnel before we touch a pixel.
Ask. One adaptive step replaces the poster grid. You type something like "slow-burn crime with great writing" and the system answers in kind: got it, crime dramas, character first, no filler. It is one question, in plain language, and it is skippable. Nothing about watching television should be gated behind a form.
Prove. This is the part most personalization skips. The first home screen names its reasons out loud: because you picked Caldera, because you asked for slow-burn crime. The subscriber can see why a row is there, and can correct it. Personalization you can audit beats personalization that quietly guesses, because the moment a recommendation is wrong you know exactly which input to change.
The rest of the flow follows the same principle: value before the ask. Trailers play before an account exists, so the catalog sells itself. The plan screen makes a symmetric argument and recommends one with a stated reason instead of a badge. Accounts take one tap with Apple or Google, with no birthday and no ZIP code. And the checkout states the price, the first charge, the renewal date, and the cancel path in the place where money changes hands.
Why This Matters Outside of Streaming
Almost every business we work with has a version of NEXT (0). It is the intake form that asks nine questions and tells the customer nothing. It is the quote request that disappears into an inbox. It is the onboarding that collects a profile and then behaves identically for everyone who fills it out.
The fix is rarely fewer fields. It is moving the intelligence to the front, so the first question can be open-ended and the system can actually act on the answer. That is a different architecture, not a cosmetic change: something has to read the reply, reason about it, and carry that reasoning into what the customer sees next.
Value first, forms last, and an experience that proves it listened. That is the whole thesis, and it applies to a streaming signup exactly as well as it applies to a service department or a listing intake.
The Full Case Study The Full Case Study
See all six moments, shipped beside reimagined
The case study puts every screen side by side: the welcome, the plan sell, the account, the checkout, the taste step, and the first home. Tap any comparison to open it full screen.
