Comprehensive Reports Allow an EdTech Startup to Analyze User Behavior and Refine Its Product Strategy: MentorShow Case

The Challenge

MentorShow, a French EdTech startup offering exclusive online masterclasses by renowned speakers, had already established data processing and analytics processes. However, the quality and accuracy of their calculations left much to be desired. KPI definitions were inconsistent, data pipelines had gaps that produced unreliable numbers, and the analytics team couldn’t confidently answer critical product questions.

The most pressing need was understanding user behavior — specifically, why subscribers were canceling and what content engagement patterns predicted long-term retention. Without accurate analytics, the product team was making content investment decisions (which masterclasses to produce, which speakers to sign) based on incomplete data. Given that each masterclass represents a significant production investment, these decisions had major financial implications.

Each masterclass represented a production investment of tens of thousands of euros — studio time, speaker fees, post-production, and marketing. Making content investment decisions without reliable data about what resonated with subscribers was like navigating without a map. The product team had theories about which content categories drove subscriptions, but couldn’t validate them. Meanwhile, the marketing team was promoting content based on intuition rather than performance data, potentially pushing low-engagement masterclasses that actually accelerated churn rather than preventing it.

Our Approach

We restructured MentorShow’s entire analytics layer from the ground up:

• KPI Audit & Redefinition: We reviewed every existing KPI calculation, identified inaccuracies (several metrics were double-counting or excluding relevant data), and established precise, documented definitions. This alone resolved weeks of conflicting reports between teams.
• Data Pipeline Rebuild: We rewrote the Python and SQL scripts powering MentorShow’s data processing. The new pipelines included data quality checks at each stage, error handling, and logging — replacing the fragile scripts that would silently fail and produce incorrect numbers.
• User Behavior Analytics: We built a comprehensive user behavior analytics framework that tracked the full subscriber lifecycle: acquisition, activation (first masterclass viewed), engagement patterns (viewing frequency, completion rates, content preferences), and churn signals. This gave the product team visibility into exactly why users were leaving.
• Content Performance Dashboard: We created dashboards specifically designed for content investment decisions — showing which masterclasses drove the most subscriptions, which retained viewers best, and which content categories showed growing demand.

A critical deliverable was the cancellation analysis. By correlating viewing behavior with subscription cancellations, we identified specific patterns (e.g., subscribers who didn’t complete at least 2 masterclasses in their first month had 3x higher churn risk) that informed targeted retention campaigns.

We also developed an “early engagement predictor” model that identified which behavioral signals in a subscriber’s first two weeks predicted long-term retention. This allowed the marketing team to create targeted onboarding sequences that encouraged the specific behaviors (viewing multiple masterclasses, exploring different categories, sharing content) correlated with retention. The analytics system was designed to be self-service for common questions while maintaining the flexibility for the data team to run ad-hoc analyses for strategic decisions about content investment and market expansion.

Results

• Refined data processing with accurate, documented KPI calculations trusted across all teams.
• User behavior analytics revealing the root causes of subscription cancellations — actionable insights that directly informed retention strategy.
• Content performance framework enabling data-driven decisions about which masterclasses to produce and promote.
• Engagement patterns correlated with retention, enabling 2x improvement in targeted intervention effectiveness.
• Reliable, automated data pipelines replacing error-prone manual processes.

Technologies Used

Python, SQL, data warehousing, BI dashboards, user behavior analytics frameworks.

Project Screenshots