Marketing & Big Data Surat Teerakapibal, Ph.D. Lecturer in Marketing Director, Doctor of Philosophy Program in Business Administration Thammasat Business School
What is Marketing?
Anti-Marketing
Marketing The process by which companies create value for customers and build strong customer relationships in order to capture value from customers in return. Reference: Kotler, P. and Armstrong, G. (2014). Principles of Marketing. Pearson: London.
The Big Challenge
The Evolution of Marketing Science Pre-1980s Aggregate data Time delayed data E.g., warehouse withdrawals 1980s The first data revolution Behavior of individual consumers making purchases over many shopping trips and for many product categories E.g., scanner data
The Marketing Information System Reference: Kotler, P. and Armstrong, G. (2014). Principles of Marketing. Pearson: London.
Internal Databases Marketing department Customer characteristics Sales transactions Website visits Customer service department Customer satisfaction Service problems Accounting department Sales Costs Cash flows
Internal Databases (cont.) Operations department Production Shipments Inventories Salesforce reports Competitor activities Reseller reactions Marketing channel partners Point-of-sale transaction
Marketing Intelligence Observe consumers Quizzing company s own employees Benchmarking competitors productions Researching the Internet Monitoring the Internet buzz
Marketing Intelligence Service Sample
Marketing Research Exploratory research preliminary information to help define problems Observational research Ethnographic research Depth interview Focus groups Causal research test hypotheses about cause-andeffect relationships Experimental research
Marketing Research (cont.) Descriptive research describe marketing problems, situations, or markets, such as the market potential for a product or the demographics and attitudes of consumers Survey research Secondary data Consumer panel data
Purchase data Store data Customer data Customer ID Store ID Customer ID Brand bought Products available Income Quantity bought Price Household size Place of purchase (Store ID) Feature
Synergy of Data Store data Purchase data Consumer data
Advancements of Data Analysis Descriptive Statistics Mean, standard deviation, standard errors, etc. E.g., Which type of consumers respond to coupons and deals? Behavioral Models Originate from Transportation Science and Economics Inclusion of time-dependence terms, consumer perceptions, context dependence, etc. E.g., Which consumer segments are more loyal than others?
Random Utility Model (RUM)
What is Big Data? Large amount of data Emerge from decreasing storage costs Soaring interests in collecting data Increase at an unprecedented rate From firms and specialized data suppliers Social network data Consumer shopping and purchase behaviors
What is Big Data? Different data sources Internal External Multiple data types Structured: fixed format Semi-structured Unstructured: undefined format
What is Big Data? Many forms Text Web data Tweets Sensor data Audio Video Click streams
What is Big Data? Data in motion Speed continues to accelerate Data creation Data processing Data analysis
What is Big Data? Real-time data access Enabling models to customize marketing instruments to consumers as consumers search for information, compare prices or make purchases
The Age of Big Data
Video-Based Automated Recommender (VAR) System Capturing shoppers behavior Facial expression Body positions Identification of similar customers Database of shoppers with known: preferences purchasing consideration decisions Making recommendations Suggest preferred garments
Video-Based Automated Recommender (VAR) System
Forecasting TV Program Demand Implications for broadcasting companies and advertisers Prediction of TV ratings Show adjustments Pricing of advertising: dynamic, real-time, etc. Challenges Unstructured data: texts, video, audio, and images Large volume of data Solution
Forecasting TV Program Demand
Forecasting TV Program Demand Twitter is a crucial indicator of TV rating 3 measures Volume Sentiment (positive, neutral, negative) Principal Component Analysis ( tonight, can t wait, etc.)
Visualization of Competitive Market Structures Who are your competitors? Brand switching matrix Consideration sets Perceptual map What if you have 1,000 products in the category? Survey will not work (Cognitive capability is an issue) Solution Clickstream data
Visualization of Competitive Market Structures
Visualization of Competitive Market Structures
Visualization of Competitive Market Structures
Visualization of Competitive Market Structures
Visualization of Competitive Market Structures
Data Science
The Shopper
The Consumer Black Box Product Price Place Promotion... CRM Situation Influencers Choice
Tree of Marketing Reference: Department of Marketing, Thammasat Business School
Econometrics. What is it? Econometrics is the branch of economics that aims to give empirical content to economic relations. The most basic tool for econometrics is the linear regression model
Sales (Million) 120 Advertising Expenditure vs. Sales 100 80 60 40 20 0 0 5 10 15 20 25 30 Advertising Expenditure (Million)
Sales (Million) 120 Advertising Expenditure vs. Sales 100 80 60 40 Sales = 45.75 + 2.58 Advertising Expenditure 20 0 0 5 10 15 20 25 30 Advertising Expenditure (Million)
Implications Capability to determine existence of relationship On average, increasing the advertising expenditure by 1 baht will result in 2.58 baht increase in sales. (More careful statistical analysis could be conducted to test for statistical significance.)
Implications (cont.) Future sales can be forecasted based on knowledge about advertisement expenditure For instance, if the firm spends 10million baht on advertising next month, what would be the expected sales?
Software Packages
An Example: Car or Bus?
Random Utility Models (RUMs) A decision maker, n, faces a choice among J alternatives. The utility that decision maker n obtains from alternative j is: known to the researcher unknown to the researcher
Implications We need 2 equations for our case: 1 for car and 1 for bus. We need to know marketing theories in order to identify variables to be included in the model. We need to find the data for each of the variables for both car and bus We need large amount of data for accuracy Existence of relationship Predication
Based on Traditional Logit Framework For car: For bus:
Based on Traditional Logit Framework (cont.) For car: For bus: known to the researcher
How Consumers Choose?
What Do We Have to Do?
Likelihood Function
Software Packages
Nested Logit Alternatives can be partitioned into subsets, called nest Some information regarding decision making process can be revealed
Implications We can find the maximum value of likelihood for different choice structures to determine which resembles consumers behavior best Black Color?? Pepsi Coke Fanta Mirinda Pepsi Mirinda Coke Fanta
Price Elasticity A measure of responsiveness of the quantity of a product or service demanded to changes in its price:
Reference Price Replace price by reference price Price Previous Price Price Advertised Price Price Price that Friends Bought
Machine Learning Reproduce known patterns and knowledge in order to apply results to decision making and actions Automatically Quickly Sample algorithms Classification Regression Clustering Dimensionality reduction
Implications Businesses should consider transforming themselves to be more data-driven Culture Personnel Structure Speed and automation is the key Data is only the raw material. Success lies in Analytics Incorporation of findings into decisions Never stop improving!