Turn data into decisions about people.

Available now

Product Analytics

Applied Data Science Techniques for Actionable Consumer Insights

Addison-Wesley Data & Analytics Series · 2020

Product Analytics is a complete, hands-on guide to generating actionable business insights from customer data. It introduces practical statistical techniques for determining why things happen and how to change what people do at scale, paired with social science methods for building better theories and metrics and driving faster, more sustained behavior change.

Written for entrepreneurs, product managers, marketers, and other business practitioners, it teaches through intuitive examples from both web and offline environments. Light on math, it walks you step by step through choosing the right technique for each application, running the analysis in R, and getting answers you can trust.

• Develop testable theories of individual and social behavior in web products
• Build better metrics and KPIs for any product or system
• Run more informative A/B tests, then go beyond them with difference-in-differences, regression discontinuity, propensity score matching, and uplift modeling
• Project user populations and business needs with demographic techniques

• A/B testing
• Causal inference
• Metrics & KPIs
• Behavior change
• Uplift modeling
• R

Coming December 2026

Demographic Techniques for Data Scientists

Population thinking for metrics, models, and fairness in machine learning.

Springer Nature · Forthcoming, December 2026

Demographic Techniques for Data Scientists brings over 500 years of demographic insight into modern data science practice. For centuries, demographers have rigorously studied how populations evolve over time, developing sophisticated approaches to measurement, forecasting, and the statistical examination of population. These tools, deeply grounded in the interaction between population and time, offer powerful, underutilized perspectives for today’s data scientists.

The book starts with the foundations of robust metrics, then applies them to one of the field’s most pressing problems: finding and addressing bias and discrimination in machine learning. It culminates in the SHIFT Framework, a step-by-step methodology grounded in the patterns of behavioral data, with code in Python and R.

• Build robust, consistent metrics across period, age, and cohort, with principled exposures and denominators
• Reinterpret everyday data science metrics, like precision and recall, through population and time
• Evaluate bias in machine learning with subgroup analysis, causal inference, and algorithmic impact assessments
• Integrate census and survey data with behavioral datasets to measure who your models really serve

• Demography
• Metrics
• Census & survey data
• Algorithmic bias
• Causal inference
• Python
• R

Inside the book: chapter overview

1. Developing Robust and Consistent MetricsThe foundational concepts of period, age, and cohort, and how to build simple metrics across them. Essential for understanding the interplay of time and population in metrics and models.
2. The Craft of Complex Metric ConstructionPeriod-based metrics, exposure variables, and principled denominators and rates, plus path dependence and funnel metrics for population flows. Implemented in Python and R.
3. Demographic Techniques for Data Science MetricsApplying population and time to common data science metrics such as precision and recall, with theory and code.
4. Metrics to Assess DiscriminationEvaluating bias in machine learning through population representativeness: subgroup analysis, causal inference with observational data, and structured evaluation across demographic groups.
5. The World of Census DataThe history, methodology, and global context of census and survey data, including the United States, Canada, India, and Brazil.
6. Census and Survey Data IntegrationBuilding metrics from U.S. American Community Survey and international census data, and integrating them with behavioral datasets. Python and R examples.
7. Mapping Conceptual and Demographic Variables in ModelsTurning preferences, heuristics, and decision-making patterns into measurable variables, the limits of conventional ML approaches to bias, and a framework for human heuristics of bias and discrimination.
8. Algorithmic Impact Assessments and BenchmarkingEvaluating model performance in real-world contexts, including bias, coverage, and representativeness, with frameworks for external benchmarking.
9. The SHIFT FrameworkA step-by-step methodology for addressing discrimination in machine learning, built on the underlying patterns of behavioral data, with applied case studies.

Video course · 7h 16m

Product Analytics for Data-Driven Decisions

Derive Insights from Web Analytics Data

O’Reilly · Pearson · June 2022 · Beginner to intermediate

Eight video lessons on the core concepts for working with data: spotting bias in datasets, telling good data from bad, and turning findings into actionable business decisions. Real-world examples show how to use qualitative and quantitative measures to reach the outcomes you want. No prerequisites.

• Identify and create effective metrics and KPIs that drive growth
• Avoid common mistakes when interpreting data
• Progress from raw data to inference and strategic decisions

• Metrics & KPIs
• Theory building
• Behavior change
• Survey design
• Index creation
• A/B testing

Inside the course: lesson outline

1. Explore the Data-Generating ProcessPart 1: Theory building techniques in product analytics
2. Theory BuildingPart 1: Theory building techniques in product analytics
3. Behavior ChangePart 1: Theory building techniques in product analytics
4. Basic Statistical Techniques and Common PitfallsPart 2: Testing theories, feature and metric development
5. Conceptualization, Operationalization and Metric DevelopmentPart 2: Testing theories, feature and metric development
6. Metric Development ProcessPart 2: Testing theories, feature and metric development
7. Index CreationPart 2: Testing theories, feature and metric development
8. A/B TestingPart 2: Testing theories, feature and metric development