Vitality × Spotify Rewards — Product Analytics

Problem Rewards programmes that integrate lifestyle products face a fundamental measurement question — do the rewards change behaviour, or do they simply attract members who were already active? Without a structured pilot framework and bias controls, programme ROI claims rest on shaky ground.

Focus The project designs a 12-week pilot measurement framework to evaluate whether a Spotify integration drives measurable improvements in physical activity and retention. The analysis includes bias documentation, cohort comparison, and a warehouse-style data model built for Power BI.

Outcome Product and partnerships teams can review engagement and retention deltas with appropriate context — understanding not just what the data shows, but where the measurement limitations sit and what a stronger experimental design would require.

// analyst.signals

Pilot measurement design with pre and post cohort analysis Retention and engagement analytics for rewards programmes Selection bias documentation and transparent limitation framing

Overview

Vitality × Spotify Rewards is a product analytics case study simulating a wellness-music partnership. The analytical focus is pilot evaluation — measuring behavioural change, documenting selection bias risks, and translating findings into actionable product recommendations.

Intelligence Layer

Opt-in reward designs are structurally prone to selection bias. Members who choose to activate rewards are rarely representative of the full base. A credible analytics layer surfaces this risk explicitly rather than papering over it.

Problem Statement

Rewards programmes that integrate third-party lifestyle products face a core measurement challenge — do rewards change behaviour, or do they attract members who were already active? Without a structured pilot framework and bias controls, programme ROI claims rest on shaky ground.

Analyst Objective

Design a 12-week pilot measurement framework to evaluate whether a Spotify Africa integration drives measurable improvements in physical activity and retention. Document bias risks and translate findings into concrete programme design recommendations.

Stakeholders

Product owner (engagement and habit loop design)
Partnerships lead (commercial ROI)
Analytics lead (measurement integrity)
Marketing (segment targeting)

Key Questions

Did reward activation correlate with activity score improvements above baseline?
Was retention materially higher in the reward-engaged cohort at 8 weeks?
How much of the observed difference is explained by pre-existing activity level differences?
Which member segments respond most strongly to the reward?

KPI Framework

Spotify streams pre vs post activation
Weekly activity score delta from baseline
Monthly cohort retention rate by engagement tier
Baseline equivalence test between opt-in and control cohorts
Cost-per-active-member by reward tier

Approach

Designed a 12-week pilot with opt-in cohort vs control group structure.
Built a warehouse-style star schema data model (dim_member, dim_week, dim_month, dim_reward, dim_content_category; fact tables for activity, Spotify, campaign exposure, reward events, retention).
Generated analysis-ready outputs: member_week_pilot.csv and member_summary.csv.
Ran baseline equivalence tests to document selection bias before comparing cohort outcomes.
Produced Python pipeline covering data generation, cleaning, and feature engineering.

Insights

Reward-engaged members showed higher post-pilot activity scores relative to their own baseline. Retention was notably higher at 8 weeks in the engaged cohort. Baseline testing confirmed opt-in members were already more active — selection bias is a documented confounding factor. The reward performs best as a retention tool, not an activation lever.

Implication

A tiered reward structure targeting different engagement segments would sharpen programme impact. Inactive segments require a different intervention. A randomised design in a future pilot would allow causal rather than directional attribution.

Deliverables

Warehouse-style star schema data model for Power BI
Python pipeline covering data generation, cleaning, and feature engineering
12-week pilot evaluation framework with cohort comparison
Interactive HTML dashboard for stakeholder review

Results

The project demonstrates rigorous product analytics thinking — building a measurement framework that surfaces its own limitations and translates findings into concrete programme design recommendations.