1. Cross-Channel Impacts of Online Advertising,
Salesforce and Product Line Strategies in O2O
Retailing Environments
Yunkun Zhao*, Khim Yong Goh* and Liwen Hou**
* National University of Singapore, ** Shanghai Jiaotong University
Contact Author: mozartkun@gmail.com
Motivation
Brand Ad
Product Line
Product Ad
Research Objectives and Hypotheses
• Evaluating the impacts of salesman attributes on effectiveness of
advertising strategies
• Evaluating the impacts of product line length on effectiveness of
advertising strategies
H1A: Brand Ad + Salesman Train
+
Brand Ad + Salesman Tenure
+
H1B: Product Ad + Salesman Train
+
Product Ad + Salesman Tenure
+
H2A: Brand Ad + Length Across
Different Brand -
H2B: Product Ad + Length Across
Different Brand -
H2C: Brand Ad + Length Within
Same Brand +
H2D: Product Ad + Length Within
Same Brand -
Data Background
Data provided by a multi-brand multi-product O2O automobile manufacturer in China,
selling 3 brands, 35 models in 1,980 dealership stores throughout China. We have (1)
Purchase history, (2) Offline visit records, (3) Automobile information, and (4) Salesman
information
Individual Level Analysis
Aggregate Level Analysis
Sample, Dependent (Y) and Independent Variables (X)
Sample: (1) Customer-product-day level data from Jan, 2014 till Jun, 2016;
(2) 551,056 observations from 524,991 customers
Choiceijt Y1: Whether customer i choose to buy product j on day t (=1 yes, =0 no)
BrandAdijt X1: Whether customer i is lead by brand-oriented advertising for product j
into offline official store to visit at day t
ProductAdijt X2: Whether customer i is lead by product-oriented advertising for
product j into offline official store to visit at day t
SpBrandTenureit X3: The number of months the salesperson who serves customer i at day t
has been responsible for the car product j
SpPassTrainit X4: Whether the salesperson who serves customer i at day t passes sales
training program
CarLineDiffBrandit X5: The number of same-line different-brand car models of the car j that
the customer i intends to buy at day t
CarLineSameBrandit X6: The number of same-line same-brand car models of the car j that the
customer i intends to buy at day t
Model Specification:
Pr( 1) ( )ijt ijtChoice X
0 1 2 3 4 5 6
7 8 9 10
11
=
*
*
ijt ijt ijt ijt it it it
it it it ijt it
ijt it
X BrandAd ProductAd DeciLevel SpMidSchool SpCollege SpGraduate
SpAge SpBrandTenure SpPassTrain BrandAd SpBrandTenure
BrandAd SpPassTrain
12 13
14 15 16 17
18 19
* *
ln( ) ln( )
ln( )
ijt it ijt it
jt jt jt j
j j i j t ijt
ProductAd SpBrandTenure ProductAd SpPassTrain
CarLineDiffBrand CarLineSameBrand CarPrice CarDisplacement
CarFuelEconomy CarSeats
Sample, Dependent (Y) and Independent Variables (X)
Sample: (1) Car-week level data from Jan, 2014 till Jun, 2016;
(2) 2,493 observations from sales of 35 unique car models
Transactionjt Y1: The total number of transactions of car model j on month t
Visitjt Y2: The total number of offline store visits of car model j on month t
TotalBrandAdjt X1: The total number of brand-oriented advertising exposures of car
model j on month t
TotalProductAdjt X2: The total number of product-oriented advertising exposures of
car model j on month t
CarLineDiffBrandjt X3: The number of same-line different-brand car models of the car j
at month t
CarLineSameBrandjt X4: The number of same-line same-brand car models of the car j at
month t
Model Specification:
andln(1 )jt jtTransaction X ln( )jt jtVisit X
0 1 2 3 4
5 6
7 8
=
* *
*
jt jt jt jt jt
jt jt jt jt
jt jt
X TotalBrandAd TotalProductAd CarLineDiffBrand CarLineSameBrand
TotalBrandAd CarLineDiffBrand TotalProductAd CarLineDiffBrand
TotalBrandAd CarLineSameBrand T
9 10 11
12 13 14
*
ln( )
ln( ) ln( )
jt jt
jt jt jt
j j j t jt
otalProductAd CarLineSameBrand
RivalBrandAdSType RivalProductAdSType CarPrice
CarDisplacement CarFuelEconomy CarSeats
Results and Findings
Individual-Level Aggregate-Level
Variables (1)
Logit
Choice
(2)
Probit
Choice
BrandAd 0.743*** 0.362***
ProductAd 0.428*** 0.259***
SpBrandTenure -0.002 -0.002
SpPassTrain 0.153*** 0.080***
BrandAd ×
SpBrandTenure
-0.005 -0.003
BrandAd ×
SpPassTrain
-0.098 -0.067
ProductAd ×
SpBrandTenure
-4.19e-04 -8.59e-04
ProductAd ×
SpPassTrain
0.231*** 0.092***
Controls √ √
Constant -10.21*** -4.177***
BIC 7099.215 8359.542
Observations 551,056 551,056
Note: Standard errors in parentheses: *** p<0.01, ** p<0.05, * p<0.1
Variables (1)
Ln(1+Transac
tion)
FE
(2)
Ln(1+Transa
ction)
RE
(3)
Ln(Visit)
FE
(4)
Ln(Visit)
RE
TotalBrandAd 1.57e-04 9.86e-04*** 1.52e-03*** 3.02e-03***
TotalProductAd 7.57e-05 5.37e-04*** 8.97e-04*** 1.70e-03***
CarLineDiffBrand -0.011 0.156*** 0.014 0.293***
CarLineSameBrand 0.032 0.098*** 0.477*** 0.751***
TotalBrandAd ×
CarLineDiffBrand
-5.02e-05 -2.50e-04*** -2.70e-04*** -5.80e-04***
TotalProductAd ×
CarLineDiffBrand
1.08e-05 7.82e-06 1.03e-05 -1.17e-05
TotalBrandAd ×
CarLineSameBrand
1.24e-04*** 1.64e-04*** 3.38e-05 -8.32e-06
TotalProductAd ×
CarLineSameBrand
1.51e-05 -1.12e-04*** -2.04e-04*** -4.16e-04***
Controls √ √ √ √
Constant 0.286 14.51*** 3.437*** 25.69***
Overall-R2 0.282 0.532 0.319 0.497
Observations 2,493 2,493 2,493 2,493
Identifications and Robustness
Instrument
Variable
Estimation
Non-
randomly
Assigned
Salesman
Endogeneity
Concern of
Product
Line Length
Simultaneity
Alternative
Model
Hierarchical
Linear
Bayesian
MCMC
Estimation
√ √ √ √ √ √ √
Contributions
• Contribute to the literature on the interdependencies between online
advertising and offline salesman interactions by granularity testing
• Contribute to the literature on the interdependencies between online
advertising strategies and product line management strategies
• Managers should evaluate trade-offs of different advertising strategies, offline
salesman training and product line management