Latest / The Yield Doctor with James Deaker / Tim Stephens

Transcript

Tim Stephens · The Yield Doctor with James Deaker

0:00
[Music]
0:04
as part of my ongoing series of
0:06
techniques to make more money through
0:09
better management of digital advertising
0:12
today my guest is Tim Stevens Tim has
0:16
been in the digital media industry for
0:18
decades starting first at ad.com as an
0:22
account manager back in 2001 and then
0:26
doing tours of Duty through AOL and
0:29
through yah
0:30
he has worked on the demand side with
0:33
dsps but for the last few years his main
0:37
area of focused has been on the supply
0:39
side specifically working with
0:41
Publishers on how they can make more
0:44
money through better programmatic demand
0:46
management Tim I am so happy that you're
0:49
able to join me for this
0:51
conversation thanks for having me James
0:54
great hey Tim let's start at the highest
0:57
level can you describe your work over
1:00
the last few years as a supply
1:03
specialist how do you describe your role
1:06
and and what are the problems that
1:08
you're trying to solve sure so my most
1:12
recent role was director of or senior
1:15
director of pricing and yield for the
1:17
Verizon media uh and then what
1:19
ultimately became the Yahoo ad exchange
1:22
U which has since been deprecated but uh
1:25
in pricing and yield for that role
1:26
ultimately my responsibility was to grow
1:29
both Topline Revenue um which is mostly
1:32
fee Revenue share based Revenue that ran
1:35
through the ad exchange and then also to
1:37
increase uh profit which was was the
1:40
yield part so the question that I had to
1:42
answer was uh how do we increase uh
1:45
profitability on the ad exchange got it
1:49
and you know obviously it's a pretty
1:50
broad area can you make it more tangible
1:53
with some sort of specific problems that
1:55
you might see or areas that might might
1:57
need to be fixed on a day-to-day basis
2:01
sure so what we largely became focused
2:05
on um over the years was trying to
2:11
optimize Revenue per auction so in a
2:15
world where you know there's billions of
2:17
auctions coming in uh it became a
2:20
question of you know how do you offset
2:22
the cost of running an ad exchange
2:25
that's AWS cost it's account management
2:28
cost how do you offset the costs in a re
2:31
a revenue share model and maximize uh
2:35
the profitability of the exchange which
2:37
as you go Downstream through the auction
2:39
it actually means that you are
2:42
optimizing uh the amount of uh revenue
2:46
for the publisher obviously in a uh in a
2:49
Model A business model that is largely a
2:51
REV share uh if you aren't helping the
2:54
Publishers uh generate more money uh you
2:57
are not going to be making money by the
2:59
fee that you take out of each uh
3:01
transaction that
3:02
closes so so let's get really tactical
3:06
because you know all of these are
3:07
auctions that are run lots of people
3:10
might be competing for the same
3:12
impression when you're thinking about
3:15
how Publishers can sort of out compete
3:17
each other or put it another a nicer way
3:20
to make more value from the their
3:22
auction experience what are the things
3:25
that they need to be focusing on is it
3:27
viewability is it flaws is it match
3:30
rates a lot of different things get
3:32
thrown around in in your experience if
3:35
when you working with a new publisher
3:37
for the first time where would you start
3:40
when you're trying to understand what
3:42
the areas for opportunity are so I think
3:44
a huge factor that has come to light uh
3:47
and there's a lot of buzzword or a lot
3:49
of acronyms that kind of talk about this
3:51
um but everybody's dancing around the
3:52
same thing so uh in the programmatic
3:55
world uh what has happened over the
3:57
years is that uh behavior that is not
4:01
necessarily the best behavior or that is
4:03
not necessarily profitable for the
4:05
different entities along the chain um
4:07
they've become kind of more and more
4:09
incentivized so when you think about
4:12
things like Auto refresh stacked ads ads
4:16
below the fold uh sending the same uh
4:21
sending the same request for a single
4:23
impression multiple times uh to the same
4:26
exchange which is something that has
4:28
happened even as CTV has has grown or
4:32
started to grow uh as the pandemic hit
4:34
we saw it even in something like CTV
4:37
that's uh you know kind of a more
4:38
refined model but I think the biggest
4:41
thing that I worked on and my team
4:43
worked on over the years is how do you
4:46
kind of uh
4:50
decomod or incentivize good behaviors
4:53
from the Publishers uh so again when I
4:56
talk about Revenue per add auction if
4:59
you can get to that point where uh it's
5:04
one impression one request for one
5:07
impression and you're doing things that
5:09
are good for the ecosystem in terms of
5:12
again getting away for those from those
5:14
Auto refreshes and stack ads and you
5:17
know formats like smaller formats and
5:19
things like that as you move away from
5:22
things that were incentivized throughout
5:24
the years that would potentially make
5:25
you more money um I I will use uh
5:28
sending a uh an impression request
5:30
multiple times so you send it out a
5:32
bunch of times it's not you know it's
5:34
not too hard to understand you send it
5:36
out a bunch of times to increase the
5:38
chances that somebody's going to buy
5:39
that you're G to make money off it um
5:41
what that does just to just to be clear
5:43
on the language say SE it out a bunch of
5:45
times are you talking about the use
5:47
cases sending the same Impressions to
5:50
different exchanges or you talking about
5:51
an example where someone might even send
5:54
the same impression multiple times to
5:56
the same exchange to the same exchange
5:59
so so it's one impression it's
6:00
ultimately going to get served but you
6:03
send that request to bid on that
6:05
impression multiple times hoping that a
6:09
system that's an SSP or an upstream DSP
6:13
uh a programmatic platform that is
6:15
having billion billions of requests come
6:17
in very high QPS you send that multiple
6:20
times in hopes that it increases your
6:24
chances to win if you send the request
6:27
to bid on that impression five times
6:29
versus one time for example so what I
6:32
found myself and my team doing was you
6:36
know creating metrics around uh things
6:40
like you know things like
6:42
viewability uh you know a lot of uh
6:44
traditional metrics um we would look at
6:47
you know like viewability things like
6:48
that um but what we would do is we would
6:51
ultimately back that into a revenue per
6:56
ad request so Revenue per ad requests
7:01
that were lower than what we would
7:02
expect from a certain placement a
7:04
certain creative size a certain format
7:07
is it mobile um even even uh Geo a
7:10
certain uh Geographic you know Europe uh
7:14
Amia versus the us we would learn you
7:16
know what is a good uh what is a good
7:19
Revenue per ad auction what is a
7:21
standard or average revenue per ad
7:23
auction and then a lot of the analysis
7:25
that my team did started with figuring
7:29
out which ones uh were not making the
7:32
grade and then we would double back and
7:35
see what was driving that was it low
7:37
viewability was it high frequency um so
7:42
again we would start with that Revenue
7:44
per add auction metric and then you know
7:46
we would work backwards and we would
7:47
have as much as possible we would be
7:49
capturing the data that would allow us
7:51
to go back and work with Publishers to
7:53
ultimately
7:54
improve uh those numbers and again as a
7:57
publisher or as an addict ex change you
8:00
know if a publisher my team would be
8:02
looking at cost for ad auction
8:04
incremental cost for ad auction we'd be
8:05
looking at things like that but if
8:07
you're a
8:09
publisher I can't speak to the exact
8:11
cost but I have to imagine that it cost
8:14
more to send that request out more times
8:17
um and what H has happened over the
8:18
years is that as things like SPO have
8:21
come along um entities may not tell the
8:26
publisher that this is what they're
8:28
going to do but they will look at things
8:30
like Revenue per at auction and they
8:33
will simply filter that inventory out
8:35
from even being auctioned to save those
8:37
costs talk to me about whether it's in a
8:41
Publisher's interest to send lots of
8:44
signals from their side into the
8:47
exchange in hopes that that's going to
8:50
drive out the value of the impression
8:52
and I think the the subtext here you you
8:55
know Sean quook very well and I remember
8:58
10 years ago Yahoo he and I got into a
9:01
long drawn I won't say disagreements
9:03
because we're probably on the same side
9:05
a lot around we were pushing for more
9:08
signals to be sent to from the publisher
9:11
side to the demand side and obviously
9:14
there's a risk in doing that that you're
9:15
going to find your buyers start to
9:18
cherry-pick your best
9:20
Impressions um but the hope is that if
9:23
you send more information you are going
9:25
to drive up the value or the revenue per
9:28
impression what your experience either
9:30
in terms of the number of signals that
9:33
should be sent from the publisher into
9:36
the exchange or if it's not a volume of
9:39
of signals what are the most valuable
9:42
things that a publisher can
9:44
share so from a standpoint of strategy
9:48
you know you can look at an entity like
9:51
the New York Times that has their own
9:53
Salesforce so you can argue maybe for
9:56
them they shouldn't be sending more
9:57
signals they should keep those signals
9:59
for what they sell uh but the vast
10:01
majority of the Publishers I would say
10:04
you know not even the midtail but maybe
10:07
outside you know top 100 on um it's hard
10:11
to to me and what I've seen over the
10:12
years is that it's hard to make an
10:16
argument against sending signals in
10:19
order to increase the revenue that each
10:22
impression is going to make it's hard uh
10:25
coming at it from The Exchange side it's
10:27
hard to argue
10:29
uh against doing that when you're not
10:33
sure how you're going to sell that
10:35
impression right so you've got that
10:36
eyeball you've got a second to to sell
10:39
that impression for as much money as you
10:42
you can get um you can withhold the the
10:45
this the signals but if you do what is
10:48
the end result is it a back fill where
10:50
you make you know you show a house at is
10:52
it a back fill where you make you know
10:54
10% of what you could have made
10:56
otherwise so in terms of sending signal
10:59
we are into a world where sending
11:03
signals is be going to become more and
11:06
more important you know like uh I'll use
11:09
an example CTV and uh like content
11:13
object um was an example where uh an
11:17
entity telling the programmatic uh
11:21
auction uh what the person's watching it
11:24
doesn't necessarily have to be show what
11:25
type of content is it com you know
11:27
entertainment things like sports
11:28
entertainment things like that um that
11:31
became a big deal in terms of being able
11:33
to sell um and especially when it came
11:36
to you know
11:38
pmps uh deal IDs things like that I
11:42
think when we talk about pmps and we
11:44
talk about D
11:45
IDs what has happened is that people
11:48
have paid premium prices for PMP dlid
11:51
but then they don't get the things
11:53
behind that that would push up the
11:57
ability to sell that impression to an
12:00
Advertiser because of what they know
12:02
about that impression so I think we've
12:04
gotten into a world where uh there isn't
12:08
a lot of value behind uh what's
12:11
happening uh in some of the value like
12:13
the PMP uh some of it is manufactured
12:16
value uh through you know pmps that
12:19
maybe aren't anything special uh and
12:21
things like that or pmps that aren't
12:23
that don't don't have rigor around them
12:25
so you don't add extra Impressions maybe
12:27
you have and above the fold PMP to use
12:30
something very simplistic but then it's
12:32
not making enough revenue and you you
12:34
know dabble in a few below the fold
12:36
Impressions things like that um so again
12:38
like I think that the main thing that my
12:40
team always struggle with is the ability
12:44
um to battle against those dis
12:48
disincentives where Publishers were
12:50
incentivized to do certain things um
12:53
like again below the fold uh sending uh
12:56
requests multiple times that would would
12:59
ultimately make them more money when
13:01
they started to look at their
13:03
metrics um and again just for the
13:05
audience benefit when you say things
13:07
like a PMP which is private Marketplace
13:09
or a deal idea we just talking about the
13:12
ability for The Advertiser to Target
13:14
against some Impressions that have
13:16
certain information associated with them
13:20
exactly yeah yeah so yeah a deal ID in
13:23
and of itself is not it does not have
13:25
any value right a deal ID has value when
13:28
there's something behind that that
13:30
creates Extra Value okay so now the the
13:33
follow on question and I think I created
13:35
a a whole video on my opinion on this
13:38
that I I made link to uh here in in
13:42
minute but there are certain attributes
13:45
of an ad or of an impression that create
13:48
more value rather than less it's the
13:50
size is the rich media associated with
13:52
it is it above the fold is it below the
13:54
fold if if I had you ask what are sort
13:58
of the three things within a
14:00
programmatic exchange environment that
14:03
you've seen that do add more value in
14:06
terms of either the deal ID or just even
14:10
the inherent value of the ad what's top
14:13
of your list of things that sort of add
14:16
value within an
14:18
exchange yeah so if you if you break it
14:20
down into you know kind of format um you
14:23
know I've largely focused on display and
14:27
then later on you know mobile and then
14:29
you know uh uh CTV connected TV or
14:31
Advanced TV ATV um became a big you know
14:34
Revenue driver so you know as you as you
14:38
look at those three things or th say
14:41
those three formats you know if I'm
14:43
talking display it would be a bigger
14:45
Banner size above the fold without a lot
14:48
of clutter without a lot of other ads
14:49
around it uh in the mobile World um
14:52
there was the concept of you know an
14:54
interstitial you know maybe there's a
14:56
brief moment uh in a game or if somebody
14:59
scrolling down like a news feed in an
15:00
app you know maybe there's a brief
15:02
moment where it's just that on their
15:05
screen um and then you know with with
15:08
CTV obviously CTV is an emerging format
15:11
um there's a lot still to be discovered
15:13
I think some of the behaviors within CTV
15:15
are not optimal um but again in in in
15:18
CTV you know you have those uh you have
15:21
those ad breaks and you know I think
15:24
it's um you know in CTV it largely
15:27
becomes you know is the the ad break
15:30
sufficiently long um are there too many
15:32
of them uh but again I think even in
15:35
something like CTV it's giving that user
15:41
a short break with a single ad that
15:44
they're going to watch for some uh
15:48
significant amount of time you know
15:49
whether it's 15 seconds playing that
15:52
back in a slightly different given the
15:54
sheared volume of ads that are out there
15:56
in very clotted environment right now
15:59
I think any kind of AD where it's a
16:02
large ad to a real person and you have
16:06
their attention that spans mobile CTV
16:09
display that the theme is the same and
16:11
it's one one of those things that how do
16:13
you actually represent that you have
16:15
that highquality advertising it's a
16:17
trick that a lot of Publishers are
16:19
dealing with let me um change gears a
16:21
little bit because one of the things I
16:23
find so fascinating about your role or
16:26
the roles that you've had recently is
16:28
that
16:29
um and probably why you're in known
16:31
danger of being taken over by AI anytime
16:34
soon is that the role of a supply
16:36
specialist is this blend of Art and
16:39
Science you have you the science is the
16:41
need to sort of dive deep into the
16:43
auction data and do analytics on it uh
16:46
and then you have maybe the the art or
16:48
I'd call the experience or the or the
16:50
pattern recognition which is the ability
16:53
to sort of root cause why they might be
16:56
underperformance and we've talked
16:58
through a lot of examp already but is
16:59
there something where you have an
17:01
example where you've had to sort of
17:03
either you or your team go in and blend
17:05
that sort of pattern matching experience
17:07
to solve a
17:10
problem yeah I think when you talk about
17:12
clutter and also you talk about the
17:14
programmatic ecosystem and you know a
17:17
lot of these things that I'm going to
17:18
talk about um there's a lot of you know
17:21
the newest kind of uh buzzword is
17:24
curation um if you go to LinkedIn and
17:28
and you follow
17:29
uh you know some of the big adte uh
17:31
adtech God the bigest maybe the bigest
17:34
adtech God so if you follow a lot of um
17:38
you know the experts that are um doing
17:40
us all a service by uh writing what they
17:43
think uh we can never uh deemphasize the
17:46
the contributions of those individuals
17:48
but um you see a new word out there
17:50
called curation right so so essentially
17:52
that's um you know people are calling
17:55
that an ad Network because you know
17:57
essentially instead of you know you
17:59
still use a DL ID but you go all the way
18:02
down the chain to the publisher and you
18:04
make what it means to me is that instead
18:07
of uh telling the publisher you know
18:10
through most of my career the messaging
18:12
to the publisher be is is send
18:15
everything you have and we'll figure it
18:17
out and make money off it um to me
18:19
curation is don't send everything you
18:22
have don't even you know create more um
18:25
let's truly curate and like you said for
18:29
you know the person can see it um it's
18:31
on a site that is not you know we have
18:34
the term MFA made for advertising um you
18:37
know a site that exists solely to drive
18:39
traffic from you know search engines so
18:42
somebody you know sees a minimal amount
18:44
of content but sees a bunch of ads on
18:46
top of each other um that's not what
18:49
curation is all about it's really
18:51
getting down to that level where um you
18:53
give the advertiser what you described
18:56
before a large ad that's somebody's
18:59
focused on without a lot of clutter
19:00
around it um and and that's what uh
19:04
We've really started to to move to um so
19:09
you know at the end of the day I think
19:11
uh in all the roles I've had um at the
19:14
end of day you you need to move back to
19:17
where like an ad is is going to be
19:19
something that that a real person like
19:21
you said actually
19:23
sees um one of the things I think is
19:25
hard for anyone who's coming into adte
19:29
is how complex this whole is and there
19:32
are often things that seem like they
19:34
should be easy and end up being
19:36
difficult so in that vein what is an
19:40
underappreciated area of complexity in
19:43
supply chain optimization or or to put
19:45
it in another way what is something that
19:48
should be easy and should work but
19:51
really is H something that should be
19:54
easy and should work uh but rarely is I
19:58
think you know when you get cross format
20:02
um and uh you have you know CTV versus
20:08
mobile um that doesn't have cookies then
20:10
you have display that has cookies and
20:12
then you have you know Safari which is
20:15
display that that's not going to be
20:17
cookie uh something that should be easy
20:22
but is very complex is knowing what kind
20:27
of information
20:29
is potentially going to be available to
20:31
you so if you come in to attech as an
20:34
analyst and you say okay in my last
20:37
industry you know say I was in auto
20:39
insurance you know you you how many P
20:42
how many accidents uh did the person
20:44
have do they have teenagers in their
20:46
house you know all that is something
20:48
that you can find out right before you
20:50
decide what you're gonna charge somebody
20:53
or you know uh that kind of thing you
20:55
know
20:57
what is what am I getting you know what
20:59
am I what what am I charging for um if
21:02
you're an Advertiser what I'm what am I
21:03
paying for um it's very complex to even
21:06
know you know what is available to you
21:09
to make decisions
21:12
on uh how much you should pay what what
21:15
the Impressions actually were and again
21:18
go to Circle back to this curation I
21:20
I've seen some comments on LinkedIn
21:23
where you know people are talking about
21:25
the fact
21:26
that uh like ads on fari for example um
21:30
have always been you know undervalued
21:33
the sell through rate is is
21:35
minimal however on the flip side um you
21:38
can make an argument that demographics
21:41
of Apple users they're the one Safari
21:44
users they're the ones you you should
21:45
want to Target right so the what has
21:48
happened is that the fact that uh users
21:52
and Safari are are typically not going
21:54
to have the ability to set third-party
21:56
cookies and then nobody buys it because
21:59
they don't know what to look at to
22:00
decide whether it's uh valuable or not
22:03
and that's also where the artc comes in
22:05
right because I think if a lot of buyers
22:08
if a lot of
22:10
advertisers could uh you know be a
22:13
little birdie on the ceiling of that
22:15
Safari user and you know know what kind
22:17
of house they lived in know you know
22:19
know what kind what their discretionary
22:21
income was um I think they would want
22:23
them you know I think they would be
22:24
willing to pay a lot of money but uh but
22:26
that doesn't happen because you know
22:28
again we don't um those signals that we
22:31
we've built out that we need to use to
22:33
determine value aren't there perfect hey
22:37
Tim thank you so much for taking the
22:39
time um both what is the right or the
22:42
best way for people to find you online
22:45
and do you have any closing
22:49
thoughts uh I think my closing thought
22:51
is that I appreciate you doing this um I
22:54
appreciate your videos uh this industry
22:56
is always changing um you know I I I
22:59
came from a I was with a CPA Network to
23:02
programmatic um to pricing and yield so
23:05
it's always been excited it's always
23:06
been a challenge and I appreciate you
23:08
doing these it was fun doing it um as
23:11
far as getting in touch with me uh
23:13
anybody please feel free to reach out to
23:15
you know talk through this stuff uh
23:17
wrestle over this stuff if you don't
23:19
agree with me um Tim Stevens LinkedIn uh
23:22
Stevens with a pH uh you'll see Yahoo at
23:25
the top of my uh previous experience but
23:28
again James I really thank you for
23:30
bringing me on not a problem it's been
23:33
so much fun and uh I'm really enjoying
23:36
doing these videos and so if anybody
23:38
wants to subscribe please do because I'm
23:40
refocusing on how to make more money
23:43
online and I think that's that's going
23:45
to be fun thanks again Tim thank you
23:48
James thanks