← All case studies
Founder-Led Product Development · 2014–2026

pagarii

A verified, video-first dating platform built on one honest introduction a day — in active development and pre-launch, currently in a waitlist phase ahead of the Pittsburgh launch.

Danita Walton — founder, product manager, and delivery lead, building pagarii independently end-to-end: she owns the spec, the review pipeline, and the agents building it, not just a supporting role on someone else's product. Marketing site · conversational onboarding · daily-loop product.

Every other dating app optimizes for more — more swipes, more matches, more time in-app. pagarii is built on the opposite bet: one verified, chance-based intro a day, closed out by a real meeting.


The research

I started by measuring the market, not guessing at it.

I analyzed 242,000 user reviews across 40+ dating apps using cluster analysis and topic modeling, and ran surveys, interviews, and focus groups with 350+ prospective users to test concepts and develop pricing assumptions.

3.04 / 5

Average satisfaction across every dating app in the study. The category isn't losing on features — it's losing on outcomes. That gap is the entire product thesis.

iOS App Store & Google Play reviews · November 2022 · n = 242,000

242k
Reviews analyzed
40+
Apps in the study
350+
Prospective users researched
400+
Event participants, two cities
The bet

One verified intro a day, closed out by a real meeting.

The pagarii Discover feed — member polaroids in colour after the daily awakening
The daily intro reveal, introducing one verified member

The Discover feed awakening · the daily intro reveal. Recordings of the live pre-launch build.

The approach

Research first, then structure, then build.

01

Measured the category before designing anything

242,000 reviews and 350+ user conversations established where the market actually fails people, and what they would pay for instead.

02

Made documentation the source of truth

378 versioned documents govern what gets built; nothing ships that contradicts canon.

03

Put a review gate in front of every build item

An independent technical review and a separate design lane, both before implementation.

04

Governed research reuse with a documented consent gate

A 2016 research archive re-verified in 2026 across 25 subjects; 18 cleared for reuse, 7 excluded pending re-consent.

What's built

From marketing site to daily loop, one continuous thread.

  • Concierge-first marketing site

    With a working chatbot hero.
  • A 20+ screen conversational onboarding

    Application, identity verification, video profile, payment — as one continuous chat thread.
  • Daily-loop prototype

    An Awakening Discover feed, a daily intro reveal, and the RSVP flow that closes it.
  • Screening before any intro

    Identity, liveness, and criminal-history/watchlist screening before any member receives an intro.
  • Governed backend

    58 versioned database migrations and six serverless edge-function services, with CI security checks.
Conversational onboarding chat
Daily intro reveal
Awakened Discover feed
pagarii homepage on mobile

Conversational onboarding · the daily intro reveal · the awakened Discover feed · the homepage.

pagarii homepage on desktop — concierge-first marketing site with chatbot hero

The live pre-launch marketing site on desktop.

The brand film

The letters of the wordmark act out the product.

A walkthrough of the whole promise: verification as a welcome, an introduction arriving like a letter, and an ending where nobody swipes away — they leave together. The two dots of the pagarii wordmark become the two people.

Directed end to end across seven scenes, in more than a dozen rounds of review. Silent — no audio track.

How it's built

A spec repo, a review pipeline, and agents in parallel.

Source of truth

A 378-document versioned spec repo as single source of truth.

Review pipeline

A Claude + OpenAI cross-model review pipeline gating every build item.

Delivery

Claude Code agents running in parallel across four repositories, merged back into canon under her direction.

Research stack

Twelve Labs and Whisper in the cloud; fully local py-feat/MediaPipe pipelines so no footage leaves her machine.

The brand film above was produced the same way. She art-directed a character rig through one generative model and generated the scenes through a second, then held every generation against a written QC checklist and rejected takes until they met spec — more than a dozen rounds. Where the tools couldn't hold the standard, she routed around them: the letterform scenes are produced deterministically in code, because generative video can't keep typography stable. Direction, QC gates, and picking the right tool for each shot — the same operating model as the rest of the build.

pagarii is currently in progress and not yet launched — the site shown is a live pre-launch waitlist build, not a finished public product.