A Brief Introduction of Real-time Bidding Display Advertising and Evaluation Datasets
1. A
Brief
Introduc/on
to
Real-‐/me
Bidding
and
Test
Datasets
Dr.
Jun
Wang
Senior
Lecturer,
Computer
Science,
UCL
Co-‐founder,
MediaGamma
The
first
event
on
ComputaAonal
AdverAsing
and
Behavior
TargeAng
Meetup
4. When
online
adver/sing
goes
wrong
hVp://mashable.com/2008/06/19/contextual-‐adverAsing/
Web
users
were
unlikely
to
click
a
shoes
ad
that
appeared
along
side
an
arAcle
about
the
rather
gruesome
story
about
severed
feet
washing
up
on
shore
7. Real-‐/me
bidding
“This
is
Lawrence
from
India.
I
was
searching
Recommender
model
in
web
and
found
your
webpage
in
search
engine.
Then,
I
visited
your
webpage
searching
relevant
contents
and
saw
unrelevant
Google
add
in
"Research
Team"
page
(aWached
screenshot).
This
add
might
vary
from
country
to
country.
But
I
feel
it
will
mislead
and
give
wrong
opinion
to
users
who
visit
your
webpage.”
-‐
Lawrence
from
India
8. Life
of
a
display
ad
in
the
RTB
environment:
0.36
seconds
8
Ad
Exchange
Demand-Side
Platform
Advertiser
Data
Management
Platform
0.
Ad
Request
1.
Bid
Request
(user,
context)
2.
Bid
Response
(ad,
bid)
3.
Ad
AucAon
4.
Win
NoAce
(paying
price)
5.
Ad
(with
tracking)
6.
User
Feedback
(click,
conversion,
etc.)
User
InformaAon
User
Demography:
Male,
25,
Student,
etc.
User
SegmentaAons:
Ad
science,
London,
etc.
Webpage
User
11. Op/mal
Bidder:
Problem
Defini/on
Bid
Engine
Bid
Request
Bid
Price
11
Input:
bid
request
include
Cookie
informaAon
(anonymous
profile),
website
category
&
page,
user
terminal,
locaAon
etc
Output:
bid
price
Considera/ons:
Historic
data,
CRM
(first
party
data),
DMP
(3rd
party
data
from
Data
Management
Placorm)
What
is
the
op/mal
bidder
given
a
budget
constraint?
e.g.,
Maximise
Subject
to
the
budget
constraint
𝑅=∑( 𝐶𝑙𝑘+ 𝐶𝑜𝑛𝑣∗ 𝑤𝑒𝑖𝑔ℎ𝑡)
12. 12
The
General
Process
for
Bidding
Op/misa/on
Red:
hard
constraints
Green:
features
Blue:
models
Note
that
“Frequency
&
recency
rules”
are
also
used
as
features
13. Op/mal
bidder:
the
formula/on
• FuncAonal
OpAmisaAon
Problem
– Dependency
assumpAon:
• SoluAon:
Calculus
of
variaAons
context+ad
features
CTR
esAmaAon
winning
funcAon
bidding
funcAon
Weinan
Zhang,
Shuai
Yuan,
Jun
Wang,
OpAmal
Real-‐Time
Bidding
for
Display
AdverAsing,
KDD’14
14. Op/mal
bidder:
the
solu/on
Weinan
Zhang,
Shuai
Yuan,
Jun
Wang,
OpAmal
Real-‐Time
Bidding
for
Display
AdverAsing,
KDD’14
15. Experiments
Offline
Online
Winner
of
the
first
global
Real-‐Ame
Bidding
algorithm
contest
2013-‐2014
Weinan
Zhang,
Shuai
Yuan,
Jun
Wang,
OpAmal
Real-‐Time
Bidding
for
Display
AdverAsing,
KDD’14
17. iPinYou
Large-‐Scale
RTB
Dataset
• The
first
published
real-‐world
RTB
dataset
• From
iPinYou,
the
largest
DSP
in
China
• 9
adverAsers,
10
days
in
2013,
64M
bids,
35GB
• Web
link:
hVp://data.computaAonal-‐adverAsing.org