Gold standard models across user engagement, revenue and profit.
Objective
Implement gold standard modelling across revenue, retention and user engagement.
Obstacle
Rapid growth had left Smitten reliant on inconsistent reporting from siloed sources.
Outcome
A modern, automated analytics stack and confidence in every decision.
Background
Smitten is a fast-growing dating app on a mission to make online dating more meaningful by focusing on shared experiences and authentic connection. With a unique product offering and a rapidly scaling user base, the team faced increasing pressure to make faster, smarter decisions. While Smitten already had strong analytical talent in-house, including a dedicated data analyst and a growth lead, they lacked the technical infrastructure to support scalable reporting and trustworthy, consistent insights. That is why they brought in 173tech: to build a modern data stack that would empower their team to move faster, go deeper, and rely less on manual reporting work.
Challenges
Ad-hoc Analysis: Despite having a data-savvy team, Smitten struggled with fragmented reporting tools and a lack of a centralised, reliable view of performance. Metrics such as revenue, churn, and retention lived in multiple systems; Amplitude, Adapty, Postgres, each with its own inconsistencies and definitions. Teams frequently lost time reconciling figures or working with outdated snapshots.
Thinking Ahead: With increasing user complexity, features like Duo Mode, multiple subscription tiers, and in-app behaviour metrics, it was clear the team needed more than just better dashboards. They needed a foundational rethink: how data was modelled, defined, and operationalised across teams.
Solution
Data Dictionary: The project began by aligning all stakeholders on a shared KPI framework. 173tech collaborated with Smitten to produce a robust data dictionary, covering definitions across revenue, subscriptions, retention, DAU/WAU/MAU and app usage metrics. This served as the blueprint for everything that followed.
Data Pipeline: Our team built a modern data stack on Snowflake, provisioned with Terraform and connected to all key sources. dbt was used to develop layered, modular data models with reusable macros, incremental logic, and robust CI. Staging models were optimised to reduce compute load, and event data from Amplitude and the server was reconciled to create a unified source of behavioural truth. Everything we did during implementation was aimed at long-term cost reduction of analytics and utmost scalability, as we knew from our bumble days just how quickly dating data could scale.
The Proof Is In The Numbers...
8
90%
10
Implementation
Clarity in Revenue & Retention: Smitten’s first major win was visibility into how their product was monetising. Before the project, revenue performance was difficult to track across plan types and cohorts. With a clean, consistent foundation in place, the team could finally see how different subscription tiers were converting, what retention looked like over time, and where potential drop-offs were occurring. This empowered smarter pricing decisions, cohort-specific strategies, and a deeper understanding of lifetime value.
User Engagement: The second breakthrough came from behavioural data. Smitten’s app features both solo and duo experiences, and user activity was scattered across server and third-party sources, making it nearly impossible to understand the full picture. Through a unified and deduplicated view of in-app behaviour, the team uncovered rich insights into how users were exploring features, how usage patterns varied across cohorts, and what engagement loops drove retention.
Cross-Team Visibility: All of this data came to life through intuitive, journey-based dashboards built in Metabase, covering acquisition, activation, engagement, and monetisation. With training and documentation in place, Smitten’s team can now self-serve insights without relying on manual reporting or engineering, but we are still supporting them with strategy sessions to ensure that all of their future analytics initiatives create value.
