Breaking through the
status quo.
A European consumer entertainment company had a strong physical product — 44 curated couple challenges spanning at-home experiences, outdoor adventures, and food-focused dates. Customers bought it, couples did the challenges, and then nothing happened. No feedback loop. No way to record how either partner felt. No system that learned what a specific couple enjoyed and built on it. The product ended exactly when it was supposed to deepen. Alongside this, paid Meta and TikTok campaigns had no reliable way to tie ad spend to post-registration behavior, and their email list had no automated flow connecting it to in-app activity. Every new couple entered a black box. Off-the-shelf tools couldn't solve it either — the core requirement was a two-sided data model where both partners rate the same challenge independently before it counts as complete, feeding an AI that returns a distinct personality profile per person with no archetype overlap. No app builder or SaaS template ships that logic.
Strategic engineering
built for impact.
We built a full-stack web application with React 18 and TypeScript on the front end and Firebase Firestore as the database. Couples register together, enter both partner names, and work through 44 challenges across 5 seasonal categories. After each challenge, both partners independently answer 6 rating questions — covering enjoyment, standout moments, repeatability, creative freedom, comfort zone, and instructions clarity. Each set of answers is stored under that partner's name within the challenge record, keeping data attributable to a specific person rather than a shared account. Once a couple completes 10 challenges, both partners' full rating histories are sent to Google Gemini 2.0 Flash Lite in a single request. The model returns 3 personality archetypes per person from 10 predefined types — with a hard rule enforced in the prompt and validated in code that no archetype can appear in both partners' top 3 simultaneously. If Gemini hits a rate limit or times out, a local fallback algorithm using keyword matching and rating multipliers runs automatically. Profiles regenerate at thresholds of 10, 15, 20, 25, and 30 completed challenges. A bonus challenge tier unlocks through a puzzle mechanic tied to a Klaviyo email opt-in. All four analytics platforms — Meta Pixel, TikTok Pixel, Firebase Analytics, and GA4 — fire from a single cookie consent toggle for GDPR compliance. The app is fully bilingual in German and English.
Measurable success
at scale.
1,200 couples registered within the first 90 days of launch with no changes to ad spend. 68% completed at least 5 challenges in their first month, against a 23% engagement benchmark for comparable consumer apps. 340 couples crossed the 10-challenge threshold and unlocked AI personality analysis — a 28% conversion rate from registered to analyzed. 54% of couples who triggered the bonus unlock flow opted into the Klaviyo email list, building over 600 engaged subscribers with zero incremental ad spend. Average session included 2.3 challenge interactions, with couples returning to rate activities completed offline. The full product shipped from brief to launch in 8 weeks.
Explore more case studies