Upload
others
View
0
Download
0
Embed Size (px)
Citation preview
Mobile Targeting
Michelle AndrewsTemple University
Chee Wei PhangFudan University
Xueming LuoTemple University
Zhang FangSichuan University
Background
Uniqueness of Mobile Technology
• Mobile Portability = Real-time Targeting• GPS, Wi-Fi, Bluetooth, iBeacon = Geo-Targeting
• Geo-Targeting + Temporal Targeting
Research Questions
(1) How do timing and location in combination affect mobile sales?
(2) What are the underlying mechanisms for these effects?
Contextual Marketing Theory
• Efforts to influence purchases must be context dependent
– Portability enables ubiquitous reach and time-sensitive offerings
? Temporal & spatial boundaries may interactively impact behavior
• Decision to attend event is function of event time and place
• Ex: web usage contexts affect revisit intentions
Kenny and Marshall 2000; Johnson 2013; Galletta et al. 2006
Field Experiment
• Large, randomized field experiment
• Text messages promoting movie tickets
• Users had not previously purchased mobile tickets
• Large city in China
• Single movie promoted
• Sent to 12,265 mobiles
SMS Message
To enjoy a movie showing this
Saturday at 4:00 pm for a reduced price, download this online ticket app to purchase
your movie tickets and select your
seat.
Variables
• Dependent
• Mobile targeting effectiveness: ticket purchase via new app
• Independent
• Temporal targeting
• Geo-targeting
Defining Temporal Targeting
• Messages sent at 2 pm
• 2 hours (Sat.), 26 hours (Fri.), 50 hours (Thurs.) before movie
• Movie time: 4 pm, Saturday
• Messages sent to mobiles located at
• Near distances: < 200 meters (from the movie theater)
• Medium distances: 200 meters < x > 500 meters
• Far distances: 500 meters < x > 2km
Defining Geo-Targeting
200m
350m
1km
Random Sampling by Location
Oversamplingof Near distance
Undersamplingof Far Distance
Cinema
Control Variables
• Theater (A, B, C, D)
• Rate plan types
• MOU (minutes used monthly)
• ARPU (monthly bill)
• SMS (amount of text messages sent and received)
• Traffic (amount of data usage)
Response Rate
• 901 of 12,265 users downloaded app and bought tickets= 7.35%
• Mobile click rates in Asia:= 0.42%
eMarketer 2012
Time0.00
0.02
0.04
0.06
0.08
0.10
0.12
Estim
ated
Mar
gina
l Mea
ns
Sam
e-da
y
Evidence: Temporal Targeting
One
-day
prio
r
Two-
day
prio
r
χ² = 9.53, p < .01
χ² = 14.68, p < .01
Mean purchase for same-day messages:• Higher than one-day prior
• Higher than two-day prior
0.00
0.02
0.04
0.06
0.08
0.10
0.12
near medium far
Estim
ated
Mar
gina
l Mea
ns
Far
Medium
Near
Evidence: Geo-Targeting
Mean purchase for proximal distances:• Higher than moderate distances
χ² = 9.20, p < .01
• Higher than far distancesχ² = 18.33, p < .01
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
0.18
0.2
Time
Estim
ated
Mar
gina
l Mea
ns
Sam
e-da
y
One
-day
prio
r
Two-
day
prio
rFar MediumNear
Sam
e-da
y
One
-day
prio
r
Two-
day
prio
r
Sam
e-da
y
One
-day
prio
r
Two-
day
prio
r
Geo- and Temporal Targeting Combined
Additional Results: Customer Scenarios• Messages sent to mobiles located in
• Residential districts
• Shopping districts
• Financial districts
0
0.05
0.1
0.15
0.2
0.25
Near Medium Far Near Medium Far
Estim
ated
Mar
gina
l Mea
ns
Sam
e-da
y
Sam
e-da
y
Sam
e-da
y
Sam
e-da
y
Sam
e-da
y
Sam
e-da
y
Sam
e-da
y
Sam
e-da
y
Sam
e-da
y
One
-day
prio
r
One
-day
prio
r
One
-day
prio
r
One
-day
prio
r
One
-day
prio
r
One
-day
prio
r
One
-day
prio
r
One
-day
prio
r
One
-day
prio
r
Two-
day
prio
r
Two-
day
prio
r
Two-
day
prio
r
Two-
day
prio
r
Two-
day
prio
r
Two-
day
prio
r
Two-
day
prio
r
Two-
day
prio
r
Two-
day
prio
r
Time
Near Medium Far
At Mall At Home At Work
Distance x Time x Customer Scenario
Geo-Targeting on same day is most effective for shoppers vs. others (χ²=5.12 & 19.07, p = .01)
Geo-Targeting’s effect diminishes over time
U-shape for far distances is robust across segments
• Mobile targeting by location and time
• Customer context matters
Geo- and Temporal Targeting Takeaway
Effectiveness Effectiveness
Time TimeNear Distance
Far Distance