Howdy Intelligent Sales System
🔍 Overview
This project was developed by the Business Analytics & Compliance Executive at Howdy Group to create a data-driven sales prediction and simulation framework across multiple restaurant brands. The system uses machine learning and simulation to improve forecasting accuracy, diagnose underperformance, and provide strategic recommendations using real-world business levers like staff, promotions, compliance, and weather.
🧠 What We Built
- Python Intelligence Engine
Trained segmented linear regression models for each outlet (e.g., Howdy F7, Giga Mall, The Lost Tribe, etc.)
Identified underperforming outlets (Howdy F7 had R² ~94.8%)
Applied Random Forest Regression to Howdy F7 for better non-linear prediction (R² ~99.28%)
Used SHAP for model interpretability
Created a simulation engine that ran 4,000+ business cases across different foot traffic, rain, staffing, and promotion combinations
Labeled each scenario with a risk flag and generated an actionable recommendation
- Power BI Dashboard
Built an interactive 2-page report:
Page 1: Segmented model summary (R² by brand, compliance, conversion rate, foot traffic vs sales)
Page 2: Diagnostic deep-dive for Howdy F7 with SHAP, simulation sliders, prediction logic, and recommendation engine
Included slicers for promotion type, weather condition, foot traffic, and simulation scenario conditions
Added a simulation log table for 4,000+ predictions
📁 Files in This Repo
howdy_intelligence_system.ipynb
Howdy_Analytics_Dashboard.pbix
Howdy_DF.xlsx
AllOutletsModel.csv
Howdy Project Summary.pdf
✅ Business Impact
Flagged inefficient outlet behavior under combo promo and overstaffing
Surfaced Rs 114K revenue opportunity in simulated conditions using targeted promo and staffing combinations
Enabled strategic decisions on promo switching, staff adjustments, and weather-based planning
Built a repeatable system for forecasting and scenario testing by brand
Delivered real-time analytics through an interactive Power BI dashboard
👨💼 Author
Developed by Muhammad Danish during tenure as Business Analytics & Compliance Executive at Howdy Group, as part of ongoing efforts to enhance operational intelligence, predictive planning, and outlet-level decision support.
🏁 How to Use
Open howdy_intelligence_system.ipynb to inspect model logic
Explore Howdy_Analytics_Dashboard.pbix to interact with predictions