Hybrid Workplace Intelligence

This project demonstrates an end-to-end Data Analytics and Business Intelligence workflow for a modern SaaS company (Deskbird). It is designed to solve the problem of fragmented sales and usage data, providing a single source of truth for strategic decision-making and automated reporting.

Problem Statement

Sales and product usage data are typically isolated (CRM, web events, product logs), making it impossible to answer high-value questions like, "Which company size is most likely to convert, and how do we optimize our office utilization product?" This project creates a robust data pipeline to unify these disparate sources, establish data integrity, and provide predictive and prescriptive insights for revenue growth and client advisory services.

Architecture & Workflow

The pipeline follows a modern ELT (Extract, Load, Transform) workflow using the dbt (Data Build Tool) framework to manage all transformations, ensuring governance and modularity.

• Extract & Load (EL): Raw, synthetic sales, CRM, web event, and product usage data are generated using Python (Pandas/Faker) and loaded into MySQL (acting as the Data Warehouse).

• Transform (T - dbt): Transformations are executed using dbt and SQL. Models are built in stages:

  • Staging: Isolates and cleans raw source tables.

  • Intermediate: Joins core dimensions (Users, Companies) early to resolve ambiguity and create clean relationships (e.g., int_user_company_map).

  • Marts: Creates final analytical tables (growth_funnel_mart, office_utilization_mart) ready for analysis.

• Analyze & Visualize (Streamlit): The cleaned data marts are connected to a Python analytics layer for advanced modeling and an interactive dashboard built with Streamlit.

Technologies Used

• dbt (Data Build Tool): The core framework for defining, testing, and executing modular SQL transformations (the 'T' in ELT).

• MySQL: Used as the persistent data warehouse for both raw and transformed data.

• Python: Used for generating synthetic raw data, statistical analysis (Propensity Modeling), and building the final application.

• Streamlit: Used to create the interactive, executive-facing dashboard that presents all key insights and recommendations.

• Pandas/Statsmodels: Key Python libraries for data manipulation and building the Logistic Regression propensity model.

Key Insights & Analytics

The unified data marts enable highly strategic analytics that drive direct revenue action:

• Sales Propensity: A Logistic Regression Model revealed that Enterprise leads have the highest statistical propensity to convert, while Mid-Market leads, despite their high Average Deal Value, have a 72% lower chance of closing.

• Funnel Optimization: Analysis identified the Mid-Market funnel as the most critical bottleneck due to its high value ($10,636 ADV) and low conversion rate, demanding immediate sales process intervention.

• Product Utilization: Data revealed highly non-traditional client usage (e.g., Enterprise weekend utilization and high SMB demand for meeting rooms), enabling the creation of a new Utilization Advisory Service as a product differentiator.

• A/B Testing: Statistical analysis confirmed that a new free trial offer generated a 37.5% statistically significant uplift in conversions.

Key Visualization & Application

The final output is an interactive analytics application built using Streamlit, designed for executive stakeholders. It provides a real-time, consolidated view of revenue performance, sales pipeline health, and customer utilization patterns.

How to Run the Project

• Clone the Repository: git clone [repo-url]

• Setup Environment: pip install -r requirements.txt

• Configure Database: Fill in your MySQL credentials in the .env file.

• Run Streamlit App: streamlit run app.py