USER ACTIVATION & subscription
USER ACTIVATION
& subscription
USER ACTIVATION & subscription
Logistic Regression + Survival Analysis | 45% Lift in Conversion Velocity
Logistic Regression + Survival Analysis | 45% Lift in Conversion Velocity
Analyzed over 600K user events from a leading SaaS platform to identify behaviors that drive activation and subscription. This two-part project moves from predicting "if" a user converts to understanding "when" they take action .
// Part One: Finding the Conversion Baseline
Built a logistic regression model using newly engineered time-based features such as. meeting cadence, integration speed, and engagement density, I arrived at a model with strong prediction performance (AUC = 0.82), accurately spotting early subscription signals.
Key findings :
A second meeting within 7 days significantly increased subscription likelihood ( β = 0.54)
Integration within 14 days emerged as one of the strongest predictors ( β = 0.76)
High 30-day meeting volume strongly correlated with paid adoption ( β = 0.93)
However, scheduling the first meeting within 7 days showed a confusing signal with β = -0.17; this I called the "First Meeting" paradox.
Analyzed over 600K user events from a leading SaaS platform to uncover the behavioral levers that drive activation and subscription. This two-part project moves from predicting if a user converts to understanding when they take action .
// Part One: The Probability Baseline (Logistic Regression)
Built a logistic regression model using newly engineered time-based features such as. meeting cadence, integration speed, and engagement density, I arrived at a model with strong prediction performance (AUC = 0.82), accurately spotting early subscription signals.
Key findings :
A second meeting within 7 days significantly increased subscription likelihood ( β = 0.54)
Integration within 14 days emerged as one of the strongest predictors ( β = 0.76)
High 30-day meeting volume strongly correlated with paid adoption ( β = 0.93)
However, scheduling the first meeting within 7 days showed a confusing signal with β = -0.17; this I called the "First Meeting" paradox.
Analyzed 600K+ user events from a leading SaaS platform to uncover the behavioral levers that drive activation and subscription. This two-part project moves from predicting if a user converts to understanding when they take action .
//Part One: The Probability Baseline (Logistic Regression)
Built a logistic regression model using newly engineered time-based features such as. meeting cadence, integration speed, and engagement density, I arrived at a model with strong prediction performance (AUC = 0.82), accurately spotting early subscription signals.
Key findings :
A second meeting within 7 days significantly increased subscription likelihood ( β = 0.54)
Integration within 14 days emerged as one of the strongest predictors ( β = 0.76)
High 30-day meeting volume strongly correlated with paid adoption ( β = 0.93)
However, scheduling the first meeting within 7 days showed a confusing signal with β = -0.17; this I called the "First Meeting" paradox.
// Part Two: Time to Convert (Survival Analysis)
Extended the analysis using survival modeling (Kaplan-Meier + Cox proportional hazards) to solve the "First Meeting" paradox coming out from Part One.
The "Aha!" moment :
Users who waited 7+ days before their first meeting were associated with materially higher conversion velocity (HR = 14.08) compared to immediate meetings (HR = 9.69), a ~45% relative lift. This suggests that early self exploration strengthens downstream subscription intent.
// Part Two: The Velocity Deep-Dive (Survival Analysis)
Extended the analysis using survival modeling (Kaplan-Meier + Cox proportional hazards) to solve the "First Meeting" paradox coming out from Part One.
The "Aha!" moment :
Users who waited 7+ days before their first meeting were associated with materially higher conversion velocity (HR = 14.08) compared to immediate meetings (HR = 9.69), a ~45% relative lift. This suggests that early self exploration strengthens downstream subscription intent.
//Part Two: The Velocity Deep-Dive (Survival Analysis)
Extended the analysis using survival modeling (Kaplan-Meier + Cox proportional hazards) to solve the "First Meeting" paradox coming out from Part One.
The "Aha!" moment :
Users who waited 7+ days before their first meeting were associated with materially higher conversion velocity (HR = 14.08) compared to immediate meetings (HR = 9.69), a ~45% relative lift. This suggests that early self exploration strengthens downstream subscription intent.
View in GitHub