Gemini for HR Data Analysts in Logistics: Build a People Analytics Dashboard That Predicts Driver Attrition 90 Days Out

You are a specialist HR data analyst with 11 years of experience building workforce analytics for logistics, transport, and supply chain companies. Help me build a people analytics dashboard so I can predict driver attrition 90 days before it happens and give operations enough time to respond. My situation: - Logistics company type: [e.g., last-mile / long-haul / 3PL / courier] - Fleet size and driver headcount: [NUMBER] - Current data infrastructure: [e.g., Workday + Power BI / SAP + Tableau / spreadsheets only] - Historical attrition rate among drivers: [PERCENTAGE per year] - Known leading indicators from previous exit interviews: [e.g., route dissatisfaction / manager conflict / pay vs competitor / hours creep] - Regulatory data constraints on driver personal data: [e.g., GDPR / CCPA / transport union agreements] Deliver: 1. A feature engineering guide: the 12 driver behaviour and engagement signals that are most predictive of attrition in logistics — with the data source for each signal and the calculation method 2. A 90-day attrition prediction model specification: which algorithm type to use given the data infrastructure specified, the training data requirement, and the expected false positive rate at different probability thresholds 3. A dashboard wireframe with five views: individual driver risk score, depot-level attrition risk, route-assignment correlation with attrition, manager-level attrition rate, and month-over-month trend — with the update frequency for each view 4. An intervention trigger protocol: at what risk score threshold does the system alert the depot manager, what action does the manager take, and how is the intervention outcome recorded for model retraining 5. A data governance framework for driver risk scores: who can see individual scores, how long scores are retained, how drivers are notified of profiling, and how to handle a driver who requests access to their score under GDPR 6. A model bias audit: how to check whether the attrition prediction model is producing systematically different risk scores for drivers by age, ethnicity, or route type — and the threshold at which to pause the model 7. An operations integration guide: how to connect attrition risk alerts to the route-planning system so high-risk drivers are not assigned to new long-haul routes that correlate with increased attrition 8. A 12-month model performance report template showing prediction accuracy, intervention success rate, and net attrition reduction — structured for the COO and CHRO to review jointly **Write every output as something a data analyst can build in Power BI or Tableau without a custom machine learning infrastructure.**