Upload
others
View
8
Download
0
Embed Size (px)
Citation preview
Geo-conque s t ing : Com p e t it ive Locat ional Targe t ing of Mobile P rom ot ions
Nathan FongZhe ng Fang
Xue m ing Luo
Geo-conque s t ing
• Mobile p lat form s have e nab le d a new form of com p e t it ive t arge t ing calle d ge o-conque s t ing
• Se nding coup ons or in form at ion to p ote n t ial cus tom e rs ne ar a com p e t itor ’s b usine ss
3
Targe t ing cus tom e rs
ne ar a com p e t it ive
locat ionFocal firm ’s
locat ion
Grow ing Popularity
4
Sou rce : xAd Mob ile Locat ion In s ig h t s Q2 2013
Highly Ef fect ive?
• Ge o-conque s t ing d e live re d h igh coup on re d e m p t ion :– Dunkin’: 3.6%– De p t . s tore : 2%
• Ne e d an exp e rim e nt to isolate t arge t ing e ffe ct s
5
6
p 0
th re shold e ffe ct ssatu rat ion e ffe ct s low targe t ing value
Focal Com p e t it iveNe u t ral
Predict ions
• Com p e t it ive locat ional t arge t ing w orks !– Show by varying locat ion and t im ing
• Sw itch ing cos t s sh ift re se rvat ion value s– Concave discount se nsit ivity ne ar focal firm– Convex discount se nsit ivity ne ar com p e t itor
7
Our Empir ical Set t ing: Movie Tickets
• De m and is context -se nsit ive• Diffe re n t iate d m ain ly by locat ion• No re sale or s torage• Se nsit ive to p rom ot ions• High fixe d cos t s
8
P h oto: Th om as Haw k
Experim ental Design
• Targe t ing 3 locat ions– Ne ar the focal re t aile r– Ne ar a com p e t itor ’s locat ion– Sim ilarly t rafficke d ne u t ral locat ion
• Offe r t im ing : now or sam e t im e next w e e k
• Rand om assignm e nt of p rom ot ion d e p th– e .g . 20% , 40% , 60% discount
9
3
10
Com p e t it ive Locat ion
Ne u t ral Locat ion
Se le ct ion on othe r factors?Sim ilar cus tom e r characte ris t ics
2.4 m ile s (4 km )
Focal Locat ion
Experim ental Design
• Targe t ing 3 locat ions– Ne ar the focal re t aile r– Ne ar a com p e t itor ’s locat ion– Sim ilarly t rafficke d ne u t ral locat ion
• Offe r t im ing : now or sam e t im e next w e e k
• Rand om assignm e nt of p rom ot ion d e p th– e .g . 20% , 40% , 60% discount
11
× 2
3 Targe t ing e ffe ct to b e m e asu re d agains t w ith in -sam ple b ase line
Rand om ize d non-targe te d g roup id e nt ifie s re al-t im e e ffe ct
Experim ental Design
• Targe t ing 3 locat ions– Ne ar the focal re t aile r– Ne ar a com p e t itor ’s locat ion– Sim ilarly t rafficke d ne u t ral locat ion
• Offe r t im ing : now or sam e t im e next w e e k
• Rand om assignm e nt of p rom ot ion d e p th– e .g . 20% , 40% , 60% discount
12
× 3
× 2
3
13
Experim ental Design
• Targe t ing 3 locat ions– Ne ar the focal re t aile r– Ne ar a com p e t itor ’s locat ion– Sim ilarly t rafficke d ne u t ral locat ion
• Offe r t im ing : now or sam e t im e next w e e k
• Rand om assignm e nt of p rom ot ion d e p th– e .g . 20% , 40% , 60% discount
14
× 3
× 2
3 N = 1,000 p e r ce ll (18,000 total)
15
0.0
1.0
2.0
3.0
4.0
5.0
6.0
7.0
8.0
9.1
Purc
hase
Rat
e
Low Medium High
Non-targeted (control) Targeted (treatment)
0.0
1.0
2.0
3.0
4.0
5.0
6.0
7.0
8.0
9.1
Low Medium High
Non-targeted (control) Targeted (treatment)
0.0
1.0
2.0
3.0
4.0
5.0
6.0
7.0
8.0
9.1
Low Medium High
Non-targeted (control) Targeted (treatment)
Focal Com p e t it iveNe u t ral
J-shap e d Re sp onse to Locat ion
16
0.0
1.0
2.0
3.0
4.0
5.0
6.0
7.0
8.0
9.1
Purc
hase
Rat
e
Focal Neutral Competitive
Low Medium High
-.01
0.0
1.0
2.0
3.0
4.0
5.0
6.0
7.0
8.0
9
Targ
etin
g Ef
fect
on
Purc
hase
Rat
e
Focal Neutral Competitive
Low Medium High
-.01
0.0
1.0
2.0
3.0
4.0
5.0
6.0
7.0
8.0
9
Targ
etin
g Ef
fect
on
Purc
hase
Rat
e
Low Medium High
Focal Competitive Neutral
0.0
1.0
2.0
3.0
4.0
5.0
6.0
7.0
8.0
9.1
Purc
hase
Rat
e
Low Medium High
Focal Competitive Neutral
Discount Response Curves
17
Compet it ive target ing: higher m arginal ef fects for deep discounts
0.3
-.01
0.0
1.0
2.0
3.0
4.0
5.0
6
Targ
etin
g Ef
fect
on
Rev
enue
Low Medium High
Focal Competitive Neutral
0.0
1.0
2.0
3.0
4.0
5.0
6
Rev
enue
(per
cus
t. as
% o
f reg
. pric
e)
Low Medium High
Focal Competitive Neutral
Revenue Maxim izing Discounts
18
Est im ates m onetary value of locat ional target ing
-.01
0.0
1.0
2.0
3.0
4.0
5.0
6.0
7.0
8.0
9
Targ
etin
g Ef
fect
on
Purc
hase
Rat
e
Low Medium High
Focal Competitive Neutral
0.0
1.0
2.0
3.0
4.0
5.0
6.0
7.0
8.0
9.1
Purc
hase
Rat
e
Low Medium High
Focal Competitive Neutral
Prom ot ional Elast icit ies
19
–3.2*
–1.7
–2.3*
–0.2
–2.2*
–2.0
Test ing for Concavity/Convexity
20
Focal Com p e t it ive Ne u t ralHigh 8.9% 5.1% 3.2%Me dium 8.2% 2.0% 1.4%Low 3.1% 1.2% 0.7%
Diff-in -d iff –4.4% ** 2.3% * 1.1%Diff-in -d iff-in -d iff 6.7% **
High – Me dium 0.7% 3.1% ** 1.8% **Me dium – Low 5.1% ** 0.8% 0.7%
Lim itat ions and Extensions
• St rate g ic re sp onse– Com p e t it ive re sp onse– Custom e r re sp onse
• Effe ct ive ne ss in othe r cate gorie s?– Context ve ry im p or tan t ( re t ail, re s t au ran t s )– Ne e d m ult ivariate d e sign
21
Conclusions
• Inve s t igate s an e m e rg ing t act ic using m obile– At t rib u t ion to locat ional t arge t ing– Effe ct ive th ird -d e g re e p rice d iscrim inat ion
• Ind icat ion of m obile ’s im p act on com p e t it ion– Ince n t ive to go on offe nsive– Te s t ing and m e asu re m e nt offline
22