2. Real
Time
Bidding
-‐
Breakdown
• What
is
Real
Time
Bidding?
• Is
it
really
Real
Time?
• What
is
caching?
• What
elements
determine
ad
decisioning?
• Tree
+
Bayesian
• Bid,
Ask
Call
–
Second
Price
AucLon
3. Real
Time
bidding
(RTB)Defined
eMarketer:
Real-time bidding (RTB) is a digital advertising technology that lets
marketers buy and publishers sell display ads dynamically, in real time,
on an impression-by-impression basis.
Scott Spencer, Director of Product Management, Ad Exchange,
Google
"Real time bidding (n);
1) A technology introduced in 2009 to make it easier for ad networks to
buy only the inventory they want.
2) A common buzzword that is part of everyone's business plan in 2013.
3) The act of buying digital inventory from multiple publishers on an
impression by impression basis, typically involving an auction pricing
mechanism.
4) One of the biggest opportunities for publishers to participate in new
media spend, but not a panacea if done in isolation.
4. Is
it
really
Real
Time?
Real
Time:
of
or
relaLng
to
a
system
in
which
input
data
is
processed
within
milliseconds
so
that
it
is
available
virtually
immediately
as
feedback,
e.g.,
in
a
missile
guidance
or
airline
booking
system.
Source:
Google
Real
Time
AdverLsing
Systems
rely
on
historical
data
that
is
cached
as
well
as
secondary
or
primary
test
variables.
Cache
update
cycles
maUer.
The
newest
data
will
help
you
make
the
best
decision.
5. What
is
a
Cache
and
Decision
Flow
Cache
(compuLng)
-‐
Wikipedia,
the
free
encyclopedia
en.wikipedia.org/wiki/Cache_(compuLng)
Wikipedia
In
compuLng,
a
cache
(/ˈkæʃ/
KASH)
is
a
component
that
stores
data
so
future
requests
for
that
data
can
be
served
faster;
the
data
stored
in
a
cache
might
be
the
results
of
an
earlier
computaLon,
or
the
duplicates
of
data
stored
elsewhere.
Decision
Flow:
Impression
-‐>
Bid
Engine
EvaluaLon
-‐>
Decision
(Bid
or
no
bid,
price
to
bid
based
on
historical
data
or
price
floor
+
potenLally
ad
quality)
6. What
Elements
Determine
Ad
Decisioning?
• TargeLng
• Publisher
• Audience
Segment
• CreaLve
• AdverLser
• Floors
• ProbabiliLes
and
Likelihoods
(click,
conversion
otherwise)
• Learning
• And
more!
7. Tree
and
Bayesian
Network
Tree:
A
tree
structure
or
tree
diagram
is
a
way
of
represenLng
the
hierarchical
nature
of
a
structure
in
a
graphical
form.
It
is
named
a
"tree
structure"
because
the
classic
representaLon
resembles
a
tree,
even
though
the
chart
is
generally
upside
down
compared
to
an
actual
tree,
with
the
"root"
at
the
top
and
the
"leaves"
at
the
boUom.
Source:
Wikipedia
Bayesian
Network:
is
a
probabilisLc
graphical
model
(a
type
of
staLsLcal
model)
that
represents
a
set
of
random
variables
and
their
condiLonal
dependencies.
Source:
Wikipedia
A
simple
Bayesian
network.
Rain
influences
whether
the
sprinkler
is
acLvated,
and
both
rain
and
the
sprinkler
influence
whether
the
grass
is
wet.
8. Example
of
a
Tree
–
Source:
Wikipedia
A
tree
structure
showing
the
possible
hierarchical
organizaLon
of
an
encyclopedia.
9. AucLon
Mechanics
• Vickrey
AucLon:
A
Vickrey
aucLon
is
a
type
of
sealed-‐bid
aucLon.
Bidders
submit
wriUen
bids
without
knowing
the
bid
of
the
other
people
in
the
aucLon.
The
highest
bidder
wins
but
the
price
paid
is
the
second-‐highest
bid.
Source:
Wikipedia
• Modern
adverLsing
aucLons:
adverLsers
are
charged
only
a
penny
more
than
the
next
lowest
bid,
with
some
adjustments
that
depend
on
the
quality
of
the
ad.
Source:
hUp://www.gsb.stanford.edu/insights/high-‐price-‐internet-‐
keyword-‐aucLons
10. Four
Key
RTB
Publisher
Takeaways
• Real
Time
Bidding
is
a
result
of
mathemaLcal
equaLons
derived
from
defined
compuLng
and
math
principals.
It
is
not
magic.
• Inventory
changes
and
shils
can
change
learning
and
valuaLon.
You
are
changing
the
inputs.
• Understand
how
your
exchange
determines
the
winning
adverLser
bid.
It
may
affect
your
pricing
strategy.
Price
Floors:
Price
floors
may
block
algorithms
from
bidding
“true
value”,
especially
when
dynamic
CPM
is
employed.
• ProgrammaLc
is
just
the
method
by
which
RTB
algorithms
are
deployed
across
the
exchanges
and
inventory
pools,
usually
by
Demand
Side
Plaporms.