Heat Pump Installation Website Analytics and Conversion Prediction Model

Overview

This project focuses on analysing website traffic patterns, user behavior, and conversion rates for a heat pump installation service. It includes exploratory data analysis (EDA), data preprocessing, and the development of a machine-learning model to predict conversion rates. The goal is to identify key factors influencing customer engagement and optimise the conversion funnel.

Table of Contents

• Introduction
• Data Sources
• Libraries Used
• Data Wrangling
• Exploratory Data Analysis (EDA)
• Machine Learning Model Development
• Model Evaluation
• Results
• Challenges Addressed
• Recommendations
• Further Investigation
• Contact

Introduction

The project aims to provide insights into how website visitors interact with a heat pump installation service's online platform. By analysing various metrics, such as website visits over time, time taken to book a design consultation, and conversion rates, the project seeks to identify opportunities for improving customer engagement and increasing the likelihood of conversion.

Data Sources

The project utilises the following datasets:

• funnel_data.csv: Contains data related to user interactions and the various stages of the conversion funnel, including first search, design consultation payments, and completion.
• installer_locations.csv: Contains location data of installers, which can be used to calculate distances and analyse regional trends.
• property_estimates.csv: Contains property estimates that can be linked to user data for further analysis.

Libraries Used

The following Python libraries were used in this project:

• Data Wrangling:
  • pandas: For data manipulation and analysis.
  • calendar: For date-related functionalities.
• Data Preprocessing:
  • LabelEncoder: For encoding categorical variables.
• Geospatial Calculations:
  • math: For calculating distances between geographical coordinates.
• Visualisation:
  • seaborn: For creating statistical visualisations.
  • matplotlib: For creating plots and charts.
  • IPython.display: For displaying images.
• Machine Learning:
  • scikit-learn: For model development, evaluation, and preprocessing.
  • imblearn: For handling imbalanced datasets using techniques like SMOTE.
  • xgboost: For training gradient boosting models.
• Model Evaluation:
  • sklearn.metrics: For evaluating model performance using metrics like accuracy, confusion matrix, and classification report.

Data Wrangling

The data wrangling process involved the following steps:

1. Loading Data: Reading the datasets (funnel_data.csv, installer_locations.csv, property_estimates.csv) using pandas.
2. Data Type Conversion: Converting date columns to datetime format using pd.to_datetime.
3. Handling Missing Values: Addressing missing data in relevant columns.
4. Feature Engineering: Creating new features such as time deltas between first visit and design consultation payment.

Exploratory Data Analysis (EDA)

The EDA process involved the following steps:

1. Understanding Data: Examining the structure, data types, and missing values in the datasets.
2. Visualising Website Visits Over Time: Creating time series plots to identify patterns and trends in website traffic.
3. Analysing Time Deltas: Calculating and visualising the time taken from the first visit to design consultation payment.
4. Determining Conversion Rates: Calculating the percentage of website visitors who booked a design consultation.

Machine Learning Model Development

The machine learning model development process involved the following steps:

1. Feature Selection: Selecting relevant features for the model.
2. Data Preprocessing: Encoding categorical variables using LabelEncoder and handling imbalanced datasets using SMOTE.
3. Model Training: Training various machine learning models, including Support Vector Machines (SVM), XGBoost, and Decision Trees.
4. Pipeline Creation: Implementing a machine learning pipeline using scikit-learn and imblearn to streamline the process.
5. Cross-Validation: Employing cross-validation techniques to ensure model robustness.

Model Evaluation

The model evaluation process involved the following steps:

1. Performance Metrics: Evaluating model performance using metrics such as accuracy, confusion matrix, and classification report.
2. Comparative Analysis: Comparing the performance of different models to identify the most effective one.
3. Visualisation: Visualising the results using confusion matrices and other relevant plots.

Results

Key findings from the analysis include:

• Website Traffic Patterns: Peak first visits occurred at the beginning of December 2023, followed by a progressive decrease.
• Time to Book Consultation: The average time from the first visit to design consultation payment is approximately 17.71 days.
• Conversion Rate: Approximately 0.98% of website visitors booked a design consultation.

Challenges Addressed

The project addressed the following challenges:

• Analysing website traffic patterns over time.
• Determining the time taken for users to pay for design consultations after their first visit.
• Calculating the percentage of website visitors who booked a design consultation.
• Understanding how conversion rates vary based on eligibility for the Boiler Upgrade Scheme grant.

Recommendations

Based on the analysis, the following recommendations are made:

• Track website visits before installation deposit: Monitor the number of website visits before clients make an installation deposit.
• Track proposal downloads: If proposal downloads are an option, track these to understand client engagement.
• Further investigation: Investigate the reasons for the decline in conversion rates after the initial peak in website visits.

Further Investigation

Further areas for investigation include:

• Impact of Boiler Upgrade Scheme grant eligibility on conversion rates.
• Reasons for the decline in conversion rates after the initial peak in website visits.
• Analysis of installer locations and their impact on service delivery.

Contact

For any questions or further information, please contact:

[Einetein]

[[email protected]]

[https://www.linkedin.com/in/einstein-ebereonwu/]