2. Dr Patrick Déglon
Senior Manager, EU Finance
eBay International AG – Switzerland
Customer Analytics
is in eBay’s DNA
Teradata User Group Schweiz – June 26th 2009
pdeglon@ebay.com
3. Agenda
• Introduction to eBay
• Analytics at eBay
• Few examples:
• Measuring impact of initiatives
• Word of mouth and marketing
• Onsite Marketing Keyword Targeting
• Acquisition & Retention Analysis
• Customers Behavior and Internet Marketing
3
4. 1995 – AuctionWeb launches
“What I wanted to do
was create an efficient
market, where regular
people could compete
with big business on a
level playing field. It
was a little bit of an
experiment.”
Pierre Omidyar,
founder
4
9. eBay: The World's Online Marketplace®
every
every
every
26
2
4
min. min. sec.
a Ford Mustang is sold
a major appliance is sold
a pair of shoes is sold
9
10. Sold or Not Sold?
a lunch with
Warren Buffett?
richest man in the world in 2008 (Forbes)
with net worth of $62 billion
Sold! for
$ 2.1 mio
bought by a Chinese hedge-fund manager in 2008
10
12. Sold or Not Sold?
a corn flake
shaped like Illinois?
Sold!
for
$1,350
12
13. Agenda
• Introduction to eBay
• Analytics at eBay
• Few examples:
• Measuring impact of initiatives
• Word of mouth and marketing
• Onsite Marketing Keyword Targeting
• Acquisition & Retention Analysis
• Customers Behavior and Internet Marketing
13
14. Analytics is in eBay DNA: Area for Analytics at eBay
example
example
Marketing
Customer
Insights
Finance
Trust &
Safety
Loyalty
Customer
Service
Search
Engine
Testing
Infrastructure
Technology
Operations
Information
Security
and many other
areas...
Finance is owning most of the areas – assuring objectivity and optimal allocation of recourses 14
15. eBay Analytics DW Infrastructure
Data Access
MicroStrategy
Unica
Crystal
SAS
Primary
Relational Data
SQL
SOA/DAL
Secondary
MPP
Relational Data
MPP
2.5 PB
Teradata
MAX
2.2 PB
Linux
Linux
Teradata
Wide Area
Interconnect
1000 miles
Sun Fire 4xxx
Solaris
Solaris
Sun Fire 4xxx
2.2 PB
XML, name/value, raw
6.6 PB
MPP/HPC/Grid
MPP/HPC/Grid
XML, name/value, raw
Data Integration
Ab Initio
Informatica
GoldenGate
UC4
BES
MAX
SOA
15
16. eBay Analytics Technology Highlights
>50 TB/day of new, incremental data
>50 PB/day
Processed
>100k data elements
>150^10 new records/day
>50k chains of logic
>5000 business users & analysts
Active/Active
turning over a TB every
x365
24x7
Millions of queries/day
Always online
99.9+% Availability
Teradata system
5 seconds
Near-Real-time
16
17. DW Sandbox enables agile analytics
Analytics teams have access
to sandboxes within eBay
Teradata data warehouses
(~ 100 GB per sandbox):
• Enable to keep the “Single
analyst’s
sandbox
Teradata Data Warehouse
Point of Truth” philosophy
• Improved Time To Market – Days / Weeks vs Months
• Enable the business to do agile prototyping
• Enable the users to “Fail
Fast” – Make it easy to try out new ideas
• Eliminate isolated Data Marts
17
26. Agenda
• Introduction to eBay
• Analytics at eBay
• Few examples:
• Measuring impact of initiatives
• Word of mouth and marketing
• Onsite Marketing Keyword Targeting
• Acquisition & Retention Analysis
• Customers Behavior and Internet Marketing
26
27. Levers for a village’s marketplace
As a market inspector in charge
of your local market, what
would be your levers?
•
•
•
•
•
•
Placement of shops
Propose services to shop owners
Pricing (fees)
– Entrance fee for shop owners
– Commission on sales
– Services
– Entrance fee for visitors
Regulations
Marketing & CRM
– Shop owners
– Visitors
…
Jour de marché à Dreux – Frank Will
27
28. Measuring impact of initiatives
A/B test
Pre/Post analysis
illustrative example (Simulation)
illustrative example (Simulation)
Number
of purchases
Number
of listings
35,000
Initiative
launched
450
400
Impact of the
initiative
350
300
test group
200
150
50
0
Aug 1st
pre
Impact of the
initiative
2008
post
D
25,000
20,000
250
100
30,000
Initiative
launched
15,000
B
2007
C
10,000
control
group
Sep 1st
5,000
Oct 1st
• Randomized Test/Control group
methodology is a golden standard in
customer insights research
0
Aug 1st
A
Sep 1st
Oct 1st
• Used to measure the impact of an
initiative in a full market or a market
segment
28
29. Agenda
• Introduction to eBay
• Analytics at eBay
• Few examples:
• Measuring impact of initiatives
• Word of mouth and marketing
• Onsite Marketing Keyword Targeting
• Acquisition & Retention Analysis
• Customers Behavior and Internet Marketing
29
30. Market growth: Simple diffusion model of new products
eBay evolution in a market
Illustrative Example
The number of New Users is proportional
to the number of Existing Users and
to the number of potential Users left
in a market.
i.e. the more people are registered…
… the more people can encourage
their surrounding to join eBay
Exponential Adoption
(network effect,
word-of-mouth)
New Users
Existing Users
Potential New Users
Mathematically:
∆N = A ⋅ (NMAX − N) + B ⋅ N ⋅ (NMAX − N)
marketing
… the less there is people
left to join eBay
Market Size Limitation
(saturation)
word-of-mouth
⇒ ∆N = a + b ⋅ N − c ⋅ N2
i.e. New Users (∆N) is a 2nd order polynomial of
Existing Users (N), i.e. inversed U shape
30
31. Market growth: Mathematical model
Mathematically:
∆N = (A + B ⋅ N) ⋅ (NMAX − N)
Dummy example with
NMAX = 1 mio (all customers)
A = 10-3 (Marketing)
B = 10-7 (Word-of-mouth)
i.e. (2nd order polynomial)
∆N = A ⋅ (NMAX − N) + B ⋅ N ⋅ (NMAX − N)
marketing
dN (New Customers)
word-of-mouth
N (Total Customers in mio)
30,000
1.00
dN (New Customers)
30,000
dN
2nd order polynomial
N
25,000
S curve
25,000
0.75
20,000
20,000
Bell curve
15,000
0.50
10,000
15,000
10,000
0.25
5,000
0
Jan 08
5,000
Jan 10
Jan 12
Jan 14
0.00
Jan 16
0
0.00
0.25
0.50
0.75
1.00
N (Total Customers in mio)
31
32. Agenda
• Introduction to eBay
• Analytics at eBay
• Few examples:
• Measuring impact of initiatives
• Word of mouth and marketing
• Onsite Marketing Keyword Targeting
• Acquisition & Retention Analysis
• Customers Behavior and Internet Marketing
32
33. Example of Internet Marketing: Portals
Customers start from their homepage, ...
... click on
an IM ad/banner
... and land on our site
33
34. Example of Internet Marketing: Paid Search
Customers search on Internet, ...
... click on
an IM ad
... and land on our site
34
35. Example of Internet Marketing: Onsite Marketing
Example of keyword targeting tests
BILLBOARD CTR BY KEYWORD TARGETED CONTENT
NikeAir
JordanXX3
Tokidoki
Handbag
Blackberry
Pearl
Apple
iPhone
Apple
No
MacBookAir Keyword
• Tested top 100 keywords targeting on homepage with dynamically generated
merchandising graphic
• Results showed up to 7x click through rate on homepage billboard (varies by
keyword), demonstrated lift between 100%-400% across multiple markets and
multiple placements
35
36. Agenda
• Introduction to eBay
• Analytics at eBay
• Few examples:
• Measuring impact of initiatives
• Word of mouth and marketing
• Onsite Marketing Keyword Targeting
• Acquisition & Retention Analysis
• Customers Behavior and Internet Marketing
36
37. Return On Investment (ROI) is a 2-D problem
Class
(Month of
registration)
Activity Month
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Jan Feb Mar Apr May Jun Jul Aug
06 06 06 06 06 06 06 06 06 06 06 06 07 07 07 07 07 07 07 07
Jan 06
Feb 06
Acquisition
Revenue
Retention
Revenue
Mar 06
Apr 06
May 06
Jun 06
Jul 06
Aug 06
Sep 06
Oct 06
Nov 06
Dec 06
Jan 07
37
38. Internet Marketing: cost to acquire a new user
Amount
(Cost & Revenues)
Cumulative monthly revenue
from new user since acquisition
Cost to acquire one new user
Cost per
new active user = Y
Revenue per new
active user = Z/month
Pay back after
X months
Months since Registration
38
39. Agenda
• Introduction to eBay
• Analytics at eBay
• Few examples:
• Measuring impact of initiatives
• Word of mouth and marketing
• Onsite Marketing Keyword Targeting
• Acquisition & Retention Analysis
• Customers Behavior and Internet Marketing
39
40. Bidding behaviors and Internet Marketing Investment
Which customer purchases are
triggered by a marketing campaign?
2 bids
missing
Behavioral bid
Uncorrelated to IM
X days
X days
bid
bid
Jan 1st
bid
Feb 1st
IM bid
Correlated to IM
bid bid
click
bid
bid bid
bid
click
Y days
1 bid is
uncorrelated
bid
bid
Mar 1st
Y days
all bids
are incremental
40
41. Latency time for each pair click - bid
Negative Latency
Bid before Click (no causality)
Behavior only
Positive Latency
Bid after Click (potential causality)
Behavior & Internet Marketing impact
Number of events
(pairs click-bid)
Real IM
increment
(correlated bids)
Level of
behavioral bids
-14
-12
Level of
behavioral bids
-10
-8
-6
-4
-2
0
2
User click on an
ad-banner at time=0
4
6
8
User bid on an eBay
item X days later
10
12
14
Latency (days)
41
42. Customer Analytics is in eBay’s DNA
Question?
Contact details:
Patrick Déglon
eBay International AG
Helvetiastrasse 15/17
3005 Berne
pdeglon@ebay.com
http://global.ebay.com
42