presented by leman akoglu march 2010
DESCRIPTION
The Dynamics of Viral Marketing Jure Leskovec Lada Adamic Bernardo A. Huberman Stanford University University of MichiganHP Labs. Presented by Leman Akoglu March 2010. Targeted marketing. Why need Viral Marketing?. Personalized recommendations Cross-selling - PowerPoint PPT PresentationTRANSCRIPT
The Dynamics of Viral Marketing
Jure Leskovec Lada Adamic Bernardo A. Huberman Stanford University University of Michigan HP Labs
Presented by Leman AkogluMarch 2010
Targeted marketing
Personalized recommendations
Cross-selling“people who bought x also bought y”
Collaborative filtering“based on ratings of users like you…”
Viral marketing
We are more influenced by our friends than strangers.
68% of consumers consult friends and family before purchasing home electronics (Burke 2003)
Our friends know about our needs/tastes better.
Why need Viral Marketing?
April 20, 2023 2
The paper in a nutshell• Analysis of a person-to-person recommendation network
(June 2001 to May 2003)– 4 million people– 0.5 million products– 16 million recommendations
Contributions:• Data statistics• Propagation, cascade sizes• Network effects• Effectiveness of viral marketing on product and
pricing categoriesApril 20, 2023 3
products customers recommenda-tions
edges buy + get
discount
buy + no discount
Book 103,161 2,863,977 5,741,611 2,097,809 65,344 17,769
DVD 19,829 805,285 8,180,393 962,341 17,232 58,189
Music 393,598 794,148 1,443,847 585,738 7,837 2,739
Video 26,131 239,583 280,270 160,683 909 467
Full 542,719 3,943,084 15,646,121 3,153,676 91,322 79,164
people recommendations
I. Music CDs and DVDs have the most/least number of items, respectively.II. Still, DVDs account for than half of all recommendations.III. Number of unique edges for Books, Music and Videos is less than number of
customers –suggests many disconnected components
April 20, 20234
1. Largest connected component at the end contains ~2.5% of the nodes.2. Total number of nodes grow linearly over time.
The service itself was not spreading epidemically.
April 20, 20235
products customers recommenda-tions
edges buy + get
discount
buy + no discount
Book 103,161 2,863,977 5,741,611 2,097,809 65,344 17,769
DVD 19,829 805,285 8,180,393 962,341 17,232 58,189
Music 393,598 794,148 1,443,847 585,738 7,837 2,739
Video 26,131 239,583 280,270 160,683 909 467
Full 542,719 3,943,084 15,646,121 3,153,676 91,322 79,164
people recommendations
IV. Influence: 1) Books (1/69) 2) DVDs (1/108) 3) Music (1/136) 4) Video (1/203)
buy+get discount
… buy+no discount
V. People tend to buy books when they can get a discount whereas for DVDsdiscount does not matter much.
April 20, 20236
7
Lag between time of recommendation and time of purchase
1 2 3 4 5 6 7 > 70
0.1
0.2
0.3
0.4
0.5
Lag [day]
Pro
po
rtio
n o
f P
urc
ha
ses
0 24 48 72 96 120 144 1680
100
200
300
400
500
600
Lag [hours]
Co
un
t
1 2 3 4 5 6 7 > 70
0.05
0.1
0.15
0.2
0.25
0.3
0.35
Lag [day]
Pro
po
rtio
n o
f P
urc
ha
ses
0 24 48 72 96 120 144 1680
500
1000
1500
2000
2500
Lag [hours]
Co
un
t
Book DVD
40% of those who buybuy within a day
but > 15% wait morethan a week
daily periodicity
April 20, 2023
Contributions of the paper:Data statistics• Propagation, cascade sizes• Network effects• Effectiveness of viral marketing on product and
pricing categories
April 20, 20238
Identifying cascades
…
t
t
tt
+t’
t+t’’
t’’’ > t’’ > t’Cascade size:
6
t+t’’’
steep drop-off
very few large
cascades
shallow drop off
DVD cascades can grow large
April 20, 20239
10
Propagation model (produces power-law cascade-size distribution)
• Each individual will have pt successful recommendations.
– pt:[0,1]
• At time t+1, the total number of people in the cascade,
Nt+1 = Nt * (1+pt)
April 20, 2023
11
• Summing over long time periods
– The right hand side is a sum of random variables and hence normally distributed. (Central Limit Theorem)
• Integrating both sides, N is log-normally distributed
if large resembles power-law
Propagation model (produces power-law cascade-size distribution)
April 20, 2023
Contributions of the paper:Data statisticsPropagation, cascade sizes• Network effects• Effectiveness of viral marketing on product and
pricing categories
April 20, 202312
2 4 6 8 100
0.01
0.02
0.03
0.04
0.05
0.06
Incoming Recommendations
Pro
ba
bili
ty o
f B
uyi
ng
10 20 30 40 50 600
0.02
0.04
0.06
0.08
Incoming Recommendations
Pro
ba
bili
ty o
f B
uyi
ng
13
Question: Does receiving more recommendations increase the likelihood of buying? (receiver’s perspective)
BOOKS DVDs
Book recommendations are rarely followed. A peak at 2, and then a slow drop (!) For DVDs, saturation is reached at 10 –diminishing returns
April 20, 2023
14
Question: Does sending more recommendations yield more purchases? (sender’s perspective)
10 20 30 40 50 600
0.1
0.2
0.3
0.4
0.5
Outgoing Recommendations
Nu
mb
er
of
Pu
rch
ase
s
20 40 60 80 100 120 1400
1
2
3
4
5
6
7
Outgoing RecommendationsN
um
be
r o
f P
urc
ha
ses
BOOKS DVDs
To too few –changes of success is low versus to everyone –spam effect For Books, the number of purchases soon saturates. For DVDs, the number of purchases increases throughout.
April 20, 2023
15
Question: Do multiple recommendations between two individuals weaken the impact of the bond on purchases?
5 10 15 20 25 30 35 404
6
8
10
12x 10
-3
Exchanged recommendations
Pro
ba
bili
ty o
f b
uyi
ng
5 10 15 20 25 30 35 400.02
0.03
0.04
0.05
0.06
0.07
Exchanged recommendations
Pro
ba
bili
ty o
f b
uyi
ng
BOOKS DVDs
YES! --Less is more…
April 20, 2023
Contributions of the paper:Data statisticsPropagation, cascade sizesNetwork effects• Effectiveness of viral marketing on product and
pricing categories
April 20, 202316
17
Recommendation success by book category
• Success rate: # of purchases following a recommendation / # recommenders
• Books overall have a 3% success rate
• Lower than average success rate– Fiction
• romance (1.78), horror (1.81)• teen (1.94), children’s books (2.06)• comics (2.30), sci-fi (2.34), mystery and thrillers (2.40)
– Nonfiction (personal & leisure)• sports (2.26)• home & garden (2.26)• travel (2.39)
• Higher than average success rate– professional & technical
• medicine (5.68)• professional & technical (4.54)• engineering (4.10), science (3.90), computers & internet (3.61)• law (3.66), business & investing (3.62)
April 20, 2023
18
What determines a product’s viral marketing success?
Modeling recommendation success-- by linear regression
# recommendations
# senders
# recipients
product price
# reviews
avg. rating
xi :βi : Coefficients : success
Over 50K products
April 20, 2023
19
Modeling recommendation successVariable transformation Coefficient βi
const -0.940 ***
# recommendations ln(r) 0.426 ***
# senders ln(ns) -0.782 ***
# recipients ln(nr) -1.307 ***
product price ln(p) 0.128 ***
# reviews ln(v) -0.011 ***
avg. rating ln(t) -0.027 *
R2 0.74
# senders and receivers have negative coefficients, showing that successfully recommended products are actually more likely to be not so widely popular more expensive and more recommended products have a higher success rate avg. rating does not affect success much
April 20, 2023
significance at the 0.01 (***), 0.05 (**) and 0.1 (*) levels
Contributions of the paper: Data statistics Propagation, cascade sizes Network effects Effectiveness of viral marketing on product and pricing categories
Questions & Comments
April 20, 202320