Data Analytics for Marketing: A practical guide to analyzing marketing data using Python

Most marketing professionals are familiar with various sources of customer data that promise insights for success. There are extensive sources of data, from customer surveys to digital marketing data. Moreover, there is an increasing variety of tools and techniques to shape data, from small to big data. However, having the right knowledge and understanding the context of how to use data and tools is crucial. In this book, you'll learn how to give context to your data and turn it into useful information. You'll understand how and where to use a tool or dataset for a specific question, exploring the "what and why questions" to provide real value to your stakeholders. Using Python, this book will delve into the basics of analytics and causal inference. Then, you'll focus on visualization and presentation, followed by understanding guidelines on how to present and condense large amounts of information into KPIs. After learning how to plan ahead and forecast, you'll delve into customer analytics and insights. Finally, you'll measure the effectiveness of your marketing efforts and derive insights for data-driven decision-making. By the end of this book, you'll understand the tools you need to use on specific datasets to provide context and shape your data, as well as to gain information to boost your marketing efforts. Cover Title Page Copyright Dedication Contributors Table of Contents Preface Part 1: Fundamentals of Analytics Chapter 1: What is Marketing Analytics? What is analytics? An overview of marketing analytics Why should we bother with marketing analytics? Exploring different types of analytics Descriptive analytics Diagnostic analytics Predictive analytics Prescriptive analytics Walking through the maze of tools and techniques Beyond simple pivot tables Why Python? Modern challenges in the world of privacy-centric marketing The importance of data engineering and tracking Don’t moonlight as a data engineer Web tracking is hard, and it is becoming harder Summary References Chapter 2: Extracting and Exploring Data with Singer and pandas Technical requirements What is ETL, and why should you care? Data pipelines What is Singer? Summarizing data and EDA Primer on descriptive statistics Percentiles, quantiles, and distributions Measures of central tendency Measures of variability Dealing with common data issues Bill Gates walks into a bar Missing values and data imputation Digging deeper into variable transformations Data standardization or scaling Power transformations Summary Further reading Chapter 3: Design Principles and Presenting Results with Streamlit Technical requirements Types of dashboards and their design Understanding the design concepts of a dashboard Thinking about how to best present data Thinking a bit about processing information Generating effective filters, dimensions, and metrics Filters Dimensions Metrics Getting your data into Streamlit and generating a basic dashboard Starting out with Streamlit Creating a marketing data dashboard with Streamlit Summary Further reading Chapter 4: Econometrics and Causal Inference with Statsmodels and PyMC Technical requirements What is a linear regression? What is a model? What are the assumptions of a linear regression? Exploring different types of regression models What we can do when the assumptions break down How to do a linear regression What is logistic regression? Objectives of logistic regression models Odds of an event What is causal inference? Correlation, causation, and key drivers A more practical application A small detour through the backdoor Watch out for colliders Summary Further reading Part 2: Planning Ahead Chapter 5: Forecasting with Prophet, ARIMA, and Other Models Using StatsForecast Technical requirements What is forecasting? Why forecasting is important Types of times series data Exploratory data analysis What to forecast Weekly, daily, and sub-daily data Time series of counts Prediction intervals for aggregates Long and short time series Transformations What types of patterns are present? Time series decomposition Time series features Basics of time series forecasting Simple methods Fitted values and residuals Correlation and forecasting Variable selection in time series regression models Advanced forecasting methods Extending regression models to time series ETS models ARIMA models The Prophet model Which model to use Summary Further reading Chapter 6: Anomaly Detection with StatsForecast and PyMC Technical requirements What is an anomaly? Techniques to detect anomalies Anomaly detection with STL decomposition Twitter’s t-ESD algorithm for anomaly detection Isolation forests for anomaly detection Forecasting as an anomaly detection tool Practical implementation with StatsForecast Using rates of arrival to identify change points Pros and cons of using rates of arrival for change point detection Summary Further reading Part 3: Who and What to Target Chapter 8: Customer Insights – Segmentation and RFM Technical requirements Understanding the sources of customer dynamics Analyzing customer dynamics – unveiling segmentation and RFM Delving deeper into what segmentation is Clustering Classification Discriminant analysis and classification Exploring RFM Approaches and techniques – independent versus sequential sorting A practical example of RFM analysis Profitability evaluation ROMI after RFM Results of using RFM for targeting Summary Further reading Chapter 8: Customer Lifetime Value with PyMC Marketing Technical requirements Diving deeper into CLV CLV in practice Using CLV to calculate acquisition costs CLV and prospects CLV and incremental value What’s wrong with the CLV formula? Issue 1 Issue 2 Issue 3 Issue 4 Issue 5 Beyond the CLV formula The BTYD model The Pareto/NBD model The BG/NBD model Implementing the BTYD model using PyMC Marketing Predicting the expected number of purchases for a new customer Estimating the CLV Summary Further reading Chapter 9: Customer Survey Analysis Technical requirements Steps in customer survey analysis Questionnaire construction Principles of questionnaire design Types of questions Asking questions Questionnaire design—layout Response formats Reliability and validity Reliability and classical measurement theory Standard error of measurement Using scales with high reliability How to do sampling Types of sampling Probability versus quota sampling Sample size for estimating population mean Response rate Control charts Customer loyalty and NPS methodology Issues with NPS Potential loss of revenue Advocacy, purchasing, and retention loyalty Factor analysis Summary Further reading Chapter 10: Conjoint Analysis with pandas and Statsmodels Technical requirements An introduction to conjoint analysis The fundamentals of conjoint analysis Setting up a conjoint study Step 1 – select the product attributes to be included Step 2 – select the product attribute levels Step 3 – create product profiles Step 4 – collect data from target customers Step 5 – estimate the utility of each product attribute and levels using regression analysis Conducting conjoint analysis in Python Determining the value of a product attribute Choice-based conjoint analysis Reporting findings Summary Further reading Part 4: Measuring Effectiveness Chapter 11: Multi-Touch Digital Attribution Technical requirements An introduction to attribution models Heuristic attribution models The implementation of different heuristic attribution models Algorithmic attribution models Shapley value attribution Fractribution Summary References Chapter 12: Media Mix Modeling with PyMC Marketing Technical requirements Understanding MMM MMM versus MTA versus lift analysis and A/B testing Steps toward implementing MMM Data collection How much data to collect Modeling How to measure the adstock effect Saturation and diminishing returns Which comes first? Selecting a model Experimenting and calibrating A synthetic data example of MMM Synthetic data generation Modeling Model results Summary References Chapter 13: Running Experiments with PyMC Technical requirements What makes a good experiment? A/A testing Type I and Type II errors p-values Common pitfalls Delving deeper into some pitfalls Conversion rate Uplift modeling Experimentation Observational studies Quasi-experiments Difference in differences Synthetic control and causal impact Summary Further reading Index About PACKT Other Books You May Enjoy