Preface
Acknowledgements
Contents
Part I Pricing
1 The Retailer's Pricing Challenge
1.1 The Potential of Data-Driven Pricing
1.2 Limitations of Traditional Economic Theory
1.3 The Shifting Objectives Behind Price
1.3.1 Price Strategy
1.3.1.1 Consistency in Pricing
1.3.1.2 Relative Price Position
1.3.1.3 Minimal Margin Rules
1.3.1.4 Competitive Price Position
1.3.1.5 Psychological Pricing Rules
1.3.1.6 Challenging Pricing Rules
1.3.2 Price Tactics During the Product Life Cycle
1.3.2.1 Product Introduction
1.3.2.2 Pricing During the Product Life Cycle
1.3.2.3 Shifting Objective over Time
1.3.2.4 End-of-Life Pricing
1.4 Escaping the Discount Trap
1.5 The Next Chapters
References
2 Understanding Demand and Elasticity
2.1 Price-Response Curve, Not Demand Curve
2.2 Measures of Price Sensitivity
2.3 A Sensible Model of Demand
2.3.1 Linear Price-Response Model
2.3.2 Constant Elasticity Price-Response Model
2.3.3 Logit Price-Response Model
2.4 Fitting Demand Curves Using Data
2.4.1 Demand and Price Indices
2.4.2 Fitting Price-Response Curves to Historical Sales Observations
2.4.2.1 Grouping Products
2.4.2.2 Scaling Product Sales for Combination
2.5 Making Forecasts
2.6 Evaluating Performance
2.7 Conclusion
References
3 Improving the List Price
3.1 Improving List Pricing
3.2 Market Conditions: Direct Versus Indirect Competition
3.3 Obtaining Competitor Price Information
3.3.1 Web Scraping
3.3.2 Transformation and Matching
3.3.3 Using Competitor Price Information
3.4 Dynamic Pricing
3.4.1 Preconditions for Dynamic Pricing
3.4.1.1 Data Availability and Quality
3.4.1.2 Variability of External Conditions
3.4.1.3 Dynamic Prices Are Socially Acceptable
3.4.1.4 Operational Feasibility
3.4.2 Types of Dynamic Pricing
3.4.2.1 Fixed Rule Dynamic Pricing
3.4.2.2 Variable Rule Dynamic Pricing
3.4.2.3 Dynamic Pricing by Means of a Learning Agent
3.4.3 Dynamic Pricing and Price Wars
3.5 Differential Pricing
3.6 Optimizing Long-Term Value
3.7 Conclusion
References
4 Optimizing Markdowns and Promotions
4.1 The Challenges of the Markdown Decision
4.2 The Traditional Markdown Process
4.3 Where the Markdown Process Fails
4.3.1 Not Making Use of Price Elasticity
4.3.2 Contaminating the Objective
4.3.3 No Anticipation of Changes in Demand Patterns
4.3.4 Time-Consuming and Error-Prone Process
4.3.5 Repeating Past Mistakes
4.4 Blueprint of an Improved Markdown Process
4.4.1 Objective
4.4.2 Portfolio Forecast and Price Selection Engine
4.4.3 Product-Level Forecast Model
4.5 Core Components of an Improved Markdown Process
4.5.1 Defining the Right Objective: Transaction Costs and Residual Value
4.5.1.1 Estimating Transaction Costs
4.5.1.2 Estimating Residual Value
4.5.2 Estimating Rotation Speed
4.5.3 Estimating Elasticity
4.5.4 Updating Elasticity
4.5.5 Satisfying Business Rules and Other Constraints
4.6 Complicating Factors
4.6.1 Operating in Multiple Markets
4.6.2 Demand Erosion
4.6.3 Combined Discount Types
4.6.4 Substitution and Cross-Price Elasticity
4.6.5 Virtual Stockouts and Low Inventory
4.7 Running Markdown Experiments
4.7.1 Single and Fixed Objective
4.7.2 A Good Split of Test and Control Groups
4.7.3 Avoid Contamination of the Control Group
4.7.4 Big Differences
4.7.5 Do Not Continue Testing Indefinitely
4.8 Promotional Discounts
4.8.1 The Purpose of Price Promotions
4.8.2 Estimating Promo Effects
4.8.3 Selecting Products for Promotional Discounts
4.9 Conclusion
4.10 Markdown Terms Glossary
References
Part II Inventory Management
5 Product (Re-)Distribution and Replenishment
5.1 Inventory Management as a Profit Driver
5.2 The Traditional Retailer's Perspective on InventoryManagement
5.3 Data-Driven Inventory Management Framework
5.4 Correcting Demand to Account for Lost Sales
5.4.1 Regular and High Sales Volumes
5.4.1.1 Traditional Time Series Models
5.4.1.2 Analyst in the Loop
5.4.1.3 Causally Related Time Series
5.4.2 Low Sales Volumes
5.5 Demand Forecasting Models
5.5.1 Forecasting Without Observed Sales
5.5.2 With Limited Historical Data
5.5.3 Improved Time Series Forecasting
5.6 Evaluating Forecast Accuracy
5.6.1 Basic Forecast Performance Measures
5.6.1.1 Use of Unseen Data
5.6.1.2 Evaluating a Point Estimate
5.6.1.3 Evaluating a Time Series Forecast
5.7 Optimizing Allocation
5.7.1 Initial Distribution of Inventory
5.7.2 Redistribution of Inventory
5.7.2.1 Identification of Sources and Destinations
5.7.2.2 Solving the Allocation Problem
5.7.2.3 Possible Extensions
5.7.3 Continuous Replenishment
5.8 Inventory Management When Selling on Third-PartyPlatforms
5.9 Conclusion
References
6 Managing Product Returns
6.1 The Challenges Created by Returns
6.2 How to Measure the Impact of Returns
6.3 Investigating Patterns in Return Behavior
6.3.1 Estimating Return Likelihood Based on Product Properties
6.3.2 Estimating Return Likelihood Based on Product Performance
6.3.3 Estimating Return Likelihood Based on Customer Behavior
6.3.4 Estimating Return Likelihood Based on OrderProperties
6.4 Taking Action to Prevent or Reduce Returns
6.4.1 Product-Based Actions
6.4.1.1 Addressing Product-Specific Causes for Returns
6.4.1.2 Adjusting the Product Assortment
6.4.2 Transaction-Based Actions
6.4.3 Customer-Based Actions
6.5 Conclusion
References
Part III Marketing
7 The Case for Algorithmic Marketing
7.1 What Is Algorithmic Marketing?
7.2 Why Algorithmic Marketing Systems Fail to Take Off
7.3 Precision Bombing, Not Carpet Bombing
7.4 Should You Focus on High-Value Customers?
7.5 Do Not Try to Beat Big Marketplaces at Their Own Game
7.6 The Low-Hanging Fruit: Get Started Without the Need for Complex Algorithms
7.7 Measuring and Experimenting
7.8 Conclusion
References
8 Better Customer Segmentation
8.1 The Purpose of Segmentation
8.2 The Problem with Traditional Segmentation
8.2.1 Segments Based on Descriptive Properties
8.2.2 RFM Segmentation
8.3 What Makes a Segment Actionable?
8.4 Customer Value Done Right
8.4.1 The Traditional RFM Approach
8.4.2 CLV-Based Customer Segmentation
8.5 From Lifetime Value to Customer Segments
8.5.1 A Simple Approximation Using Customer Groups
8.5.2 Causal Model for Variable Selection
8.5.3 From Variables to Segments
8.5.3.1 Creating Segments Using Unsupervised Clustering Techniques
8.5.3.2 Creating Segments Using Supervised Prediction Models
8.6 You Have Your Segments, Now What?
8.6.1 Product-Specific Nudges
8.6.2 Measured Incentives
8.6.3 Creating Customer Journeys
8.7 Conclusion
References
9 Anticipate What Customers Will Do
9.1 Propensity Modeling 101
9.2 The Basic Principles of Scoring Models
9.3 Using the Outputs of Scoring Models to Experiment
9.4 Pitfall: Models That Are Too Generic to Perform Badly
9.5 What Can Be Predicted Using Propensity Models?
9.5.1 Will a Customer Buy Something?
9.5.2 The Bigger Picture: Actions During Life Cycle Stages
9.5.3 Will This Customer Act on This Promotion, Action, Event, etc. ?
9.6 Using Nudges to Influence Customers
9.7 What About Recommendation Engines?
9.8 Conclusion
References
10 Anticipate When CustomersWill Do Something
10.1 Getting the Timing Right
10.2 Survival Modeling Basics
10.3 Churn Prediction Using Survival Models
10.4 Find the Rhythm: Predicting Renewal Purchases
10.5 Putting Models to Work
10.6 Conclusion
Part IV Conclusion
11 Conclusion
11.1 Where Is Retail Headed Next?
11.2 Three Big Forces
11.2.1 David and Goliath Will Keep Fighting
11.2.2 Environmental Impact Will Continue to Become More Important
11.2.3 Intelligent Models Will Learn to Cooperate
11.3 Being a Retailer
A Experimenting the Right Way
A.1 The Need for Experiments
A.2 The Basics of a Good Experiment
A.2.1 A Reasonable Path for Cause and Effect
A.2.2 A Good Hypothesis
A.2.3 Defining Success Measures
A.2.4 Actionable Results
A.3 Power: Estimate the Chance of a Successful Experiment
A.4 Selecting an Audience
A.4.1 Avoiding Experiment Contamination
A.4.2 To Stratify or Not to Stratify?
A.5 When Not to Experiment
A.5.1 Volatile Environment
A.5.2 When the Proof Has Already Been Delivered
References