Estimating the Retail Profit Effects of Retailer Targeted Coupons

Transcription

1 Estimating the Retail Profit Effects of Retailer Targeted Coupons Any mention of specific brands are for educational purposes only Seethu Seetharaman W. Patrick McGinnis Professor of Marketing December 2, 2016 co-authored with 1. Alex Chaudhry, Assistant Professor of Marketing, Texas Tech, and 2. Carrie Heilman, Associate Professor of Marketing, Virginia Slide #1

2 Coupons >> Research Questions >> Data >> Model >> Findings We are teetering on the brink of an era of mass personalisation. - Phillip Clarke, Tesco CEO ( The Emporium Strikes Back, The Economist, 13 July 2013) 2

3 Coupons >> Research Questions >> Data >> Model >> Findings Institutional Background One-to-One Themed Event Solo Mailing M 1 M 2 M 400 M 1 M 2 M 30 M 1 Impact of Sponsored Coupons on Brand Sales/Share R R R Retailer Loyalty/Patronage and Overall Store Profitability C 1 C 2 C n C 1 C 2 C 3 C C 4 C 1 C 2 C 3 5 Relevant Coupons Save Time, Effort and Money entertainment.accessatlanta.com 3

4 Coupons >> Research Questions >> Data >> Model >> Findings One-to-One Coupon coupon bundle Customer s shopping history 2179 campaigns 4

5 Coupons >> Research Questions >> Data >> Model >> Findings Themed Event Coupon Back to school, Healthy living, Baby 10 coupons Customer s lifestyle o Finer foods, healthy, convenience 1672 campaigns 5

6 Coupons >> Research Questions >> Data >> Model >> Findings Solo Mailing Coupon 1-6 coupons Customer s lifestyle & purchase history 362 campaigns 6

7 Coupons >> Research Questions >> Data >> Model >> Findings Research Questions 1. Assess Coupon Personalization Profitability - Analyze the impact of varying levels of personalized coupons on expected store profitability. 2. Capture Cross-Category Demand Spillovers - Estimate the + and - demand spillover effects within a large market basket (28 categories), i.e., Big Data context. 3. Prescribe Optimal Coupon Bundles - Across all possible combinations of all types of coupons, identify profit maximizing bundles. 7

8 Data National Supermarket Retail Chain Retailer s Loyalty Card Database Purchase Incidences from 800 Households in 28 Categories over 102 Weeks. Marketing Mix Variables (Displays, Features). Coupon Types. Demographics (Household Income, Age, Family Size). Gross Margins (Retailer Profits). Store Visits. Big Data Solution Coupons >> Research Questions >> Data >> Model >> Findings 800 Households, 42 Covariates, 102 Weeks (Sample) ~1 million observations 55 Million Customers in Database (Population) ~ 70 billion observations 8

9 Basket-level Purchase Incidence Model Multivariate Logit (MVL) Allows for cross-category dependencies across 28 categories. Targeted retail coupons impact consumers basket choices. Allows for differences across households (750 unique segments) Store Choice Model Nested Logit Coupons >> Research Questions >> Data >> Model >> Findings Proposed Model Allows store choice to depend on basket attractiveness. Targeted retail coupons impact consumers store choices. 9

10 Coupons >> Research Questions >> Data >> Model >> Findings Basket-Level Purchase Incidence Model U hct = α hc + X ct β 1hc + X hct β 2hc + Inv hct β 3hc + W h β 4hc + Z hct β 5hc + ε hct α hc - Household h s intrinsic preference β 1hc -Sensitivities to retailer marketing variables β 2hc -Sensitivities to retailer targeted coupons β 3hc -Household inventory sensitivity β 4hc -Demographic effects β 5hc -Cross-category purchase spillovers 10

11 Coupons >> Research Questions >> Data >> Model >> Findings Store Choice Model u ht = δ h + StoreAttrac ht γ h + ߟ ht δ h -Household h s intrinsic preference for the store γ h -Sensitivities to store attractiveness StoreAttrac ht = ln C c=1 e V ht, where V ht = α hc + X ct β 1hc + X hct β 2hc + Inv hct β 3hc + W h β 4hc + Z hct β 5hc 11

12 Consumer Segments Houston Chicago Support 1 Support 2 Alcohol Babyfood Bread Candy Chicken Chips Cleaning Condiments Diapers Fresh Fruit Fresh Seafood Frozen Meat Ice Cream Laundry Tea Coffee Tobacco Water Possible Segments 750 Unique Segments back Houston Chicago Alcohol Baby Food Banking Needs Bread Water Seg 2 Seg 2 Seg 1 Seg 1 Seg 2 Seg 1 Seg 1 Seg 1 Seg 2 Seg 1 12

13 Expected Store Profits in Dollars Coupons >> Research Questions >> Data >> Model >> Findings $3.66 Fresh Vegetables Each curve represents how the expected store profit changes as the retailer increases the number of retailer targeted coupons within a single category. For example: The Expected Store Profit yielded by a Customer who receives 5 Themed coupons on Fresh Vegetables is $3.66. Number of Themed Event Coupons 13

14 Expected Category Profits in Dollars Coupons >> Research Questions >> Data >> Model >> Findings 0.5 FreshVegetables 0.4 Expected Category Profit in Dollars Water Bread FreshFruit Cheese IceCream TeaCoffee Cleaning Deli Nuts $3.66 Condiments DriedFruits FrozenMeat Cookies 0.1 Candy BabyFood Alcohol Diapers FloralShop 0 Photo Number of Type B Coupons for FRESH VEGETABLES Chips Chicken BakingNeeds Butter Cereal Laundry FreshSeafood Tobacco Number of Type B (Themed Events) Coupons for FRESH VEGETABLES 14

15 210 Coupons >> Research Questions >> Data >> Model >> Findings Assess Coupon Personalization Profitability 200 Fresh Vegetables Themed Event Type B Expected Firm Profit in Dollars (millions) Butter Cereal Solo Mailing Type C One-to-One Type A Assuming all 55 million households get each coupon type, Themed Event coupons generate the highest per-customer Expected Store Profit Number of Coupons 15

16 210 Coupons >> Research Questions >> Data >> Model >> Findings Assess Coupon Personalization Profitability 200 Expected Firm Profit in Dollars (millions) Cleaning Water Deli Deli Deli Cereal Fresh Vegetables One-to-One Type A One-to-One profits are less concave than Themed coupon profits. Themed Event Type B Number of Coupons Themed Event coupon targeted segments (15-20%). 16

17 Coupons >> Research Questions >> Data >> Model >> Findings Prescribe Optimal Coupon Bundles Each Coupon Type 20 Category Coupon Bundle Simulation One coupon per category Big Data Solution (Optimization Burden) 3,108,105 possible combinations 17

18 Coupons >> Research Questions >> Data >> Model >> Findings Baseline One-to-One Themed Solo Expected Store Profit $2.37 $3.00 $2.74 $2.69 Increase in Profit N/A 26.6% 15.6% 13.5% Alcohol Babyfood BakingNeeds Bread Butter Candy Cereal Cheese Chicken Chips Cleaning Condiments Cookies Deli Diapers DriedFruits FloralShop Fresh Fruit Fresh Seafood Fresh Vegetables Frozen Meat Ice Cream Laundry Nuts Photo TeaCoffee Tobacco Water Category Coupon Bundle Simulation: one coupon per category More degrees of freedom, One-to-One Coupon can generate higher per-customer Expected Store Profits. 18

19 Expected Store Profits in Dollars per-representative customer Coupons >> Research Questions >> Data >> Model >> Findings Why measure the separate impact of each retailer targeted coupon type? Number of Coupons Themed Type Event B Fresh Veg Veg Type One-to-One A Fresh Veg Fresh Veg Type Solo C Mailing Fresh Fresh Veg Veg 19

20 Expected Category Profits in Dollars Coupons >> Research Questions >> Data >> Model >> Findings Why capture cross-category dependencies and correctly account for spillover effects? Largess Effect Fresh Vegetables FreshVegetables Focal Category Expected Category Profit in Dollars Water Bread FreshFruit Cheese IceCream TeaCoffee Cleaning Deli Nuts Condiments DriedFruits FrozenMeat Other Categories Cookies 0.1 Candy BabyFood Alcohol Diapers FloralShop 0 Photo Number of Type B Coupons for FRESH VEGETABLES Chips Chicken BakingNeeds Butter Cereal Laundry FreshSeafood Tobacco Number of Themed Event Coupons for FRESH VEGETABLES 20

21 0.3 Coupons >> Research Questions >> Data >> Model >> Findings Nuanced Cross-Category Impacts Expected Category Profit in Dollars Water IceCream Cleaning Candy TeaCoffee Cookies FrozenMeat Chips FreshFruit 0.05 Butter Ice Cream Candy Cookies Bread Bread Fresh Vegetables Cheese FreshVegetables Cheese Chicken Chicken Tobacco FreshSeafood Alcohol Type A for FRUIT 21 Number of One-to-One Coupons for FRESH FRUIT

22

23 Estimating the Retail Profit Effects of Retailer Targeted Coupons Seethu Seetharaman W. Patrick McGinnis Professor of Marketing November 11, 2016 co-authored with 1. Alex Chaudhry, Assistant Professor of Marketing, Texas Tech, and 2. Carrie Heilman, Associate Professor of Marketing, Virginia Slide #23