Latest / The Yield Doctor with James Deaker / Paul Lewis - Set Better Prices

Transcript

Paul Lewis - Set Better Prices · The Yield Doctor with James Deaker

0:00
today we are going to be talking about
0:02
activities right at the dead center of
0:05
yield Management in digital media which
0:08
is how to set prices and price flaws in
0:12
both managed and programmatic
0:15
environments my guest today is Paul
0:18
Lewis Paul has over 20 years of
0:21
experience working at various
0:23
incantations of AOL oath Verizon media
0:27
and Yahoo and that's where I first met
0:30
Paul he has had responsibilities in
0:34
functions like marketing and program
0:36
management but for much of the last
0:39
decade he has been firmly in the middle
0:41
of digital management of yield
0:44
management so he has become a vice
0:47
president in y management for the last
0:49
eight years Paul I am so glad that we
0:53
are able to
0:54
connect yeah thanks for having me James
0:56
I've really enjoyed the video series
0:58
today and I'm excited to delve into the
1:00
world of rard setting with you excellent
1:03
well let's get into it Paul the first
1:07
time that you and I met was in the
1:10
process of bringing AOL and Yahoo's
1:12
yield management functions together both
1:16
companies had rigorous approaches to set
1:18
prices a and we had to find a way to
1:21
resolve the differences now I am going
1:24
to be extremely careful to stay away
1:26
from how Yahoo sets prices now or even
1:30
the types of conversations that we had
1:33
six or seven years ago so let me just
1:36
stay at the conceptual level and say
1:40
that there are different Frameworks that
1:42
could be applied and you and I often had
1:45
discussions around the types of
1:47
Frameworks that could be applied and
1:49
when it might be appropriate to use one
1:51
versus another and where I want to start
1:54
is this idea of one framework called a
1:56
tops down approach another that's called
1:58
Bottoms Up and you know when one might
2:01
be appropriate versus another so if you
2:04
could Define those terms for me talk
2:07
about where they might be useful and the
2:10
types of different products and
2:11
different environments where where they
2:13
may be applicable sure thing when we
2:15
think about pric and strategy and yield
2:18
management tactics and operations may
2:20
change from publisher to
2:22
publisher however when we think about
2:24
the overall
2:25
strategy most decisions do boil down to
2:27
those two Frameworks
2:29
for inventory that's
2:32
non-differentiated and more likely to
2:34
transact through a programmatic channel
2:36
a Bottoms Up approach is
2:40
appropriate at its core Bottoms Up is
2:43
taking the cost per product and applying
2:46
against a fixed percentage to mark up
2:48
the cost and in most cases this is a
2:51
fairly straightforward
2:53
process however when we think about
2:54
digital media
2:56
inventory there's a few considerations
3:01
the inventory is
3:03
transient and there's not a hard cost
3:05
per se so in order to create that
3:08
framework the first step is to look at
3:10
the opportunity cost of the inventory
3:13
and in that case we're looking at
3:15
average clearing rates and average fill
3:17
rates calculating a cost By Priority
3:21
level so when we think about the
3:23
standard waterfall
3:25
model at the bottom we have Remnant or
3:28
filler inventory and monetizes at the
3:30
lowest rates so for this framework the
3:34
first step would be to create an
3:36
opportunity cost for that level market
3:38
up which creates a floor for the next
3:41
priority level up which in this case
3:42
would be open the open environment and
3:45
then you replicate that process for each
3:47
priority going up the the
3:49
chain um so what this model does is it
3:53
really protects against calization of
3:55
blocking higher paying
3:57
demand most large Publishers these days
4:00
would use Dynamic flooring to calculate
4:02
the hard cost but then for higher
4:05
priority products like private
4:07
marketplaces typically a yield team
4:10
would monitor that on a regular basis
4:12
and make adjustments based on seasonal
4:15
Trends and other demand Trends so so
4:17
Paul before you go on uh off of um a
4:22
Bottoms Up approach and we can talk
4:23
about a tops down here in a moment but
4:26
just playing that back in terms of the
4:28
type of examples so if I'm understanding
4:31
correctly you would use this if you're a
4:34
publisher and you're selling a
4:36
relatively undifferentiated product
4:39
let's say a run of network type product
4:43
but it's demanded by your advertisers
4:46
who you you have a direct relationship
4:48
with you look at the prices that
4:51
inventory might be getting in an open
4:53
market through some kind of exchange and
4:56
and let's say the open exchange rate is
4:59
is 50 cents you're talking about a
5:01
situation where okay you know you can
5:04
get 50 cents for that that's your your
5:06
Baseline your opportunity cost and so
5:09
you're going to start with that 50 cents
5:11
and think about okay now I may set up a
5:14
private Marketplace it's going to be x%
5:17
higher than that 50 Cent so maybe you're
5:20
going to price it at 70 or 75 cents
5:23
you're going to think about other ways
5:25
you might repackage that inventory maybe
5:28
you're going to sell a product that's a
5:30
manage product run of network that
5:32
delivers at a higher priority than
5:34
programmatic so you're going to go up
5:36
again from 75 cents up to a dollar or
5:39
some higher price so it's like a a
5:41
series of building blocks am I thinking
5:43
about that in the right
5:44
way uh that that's correct so the the
5:48
process would include using multipliers
5:49
to make sure that there are premiums um
5:53
at the different levels of of the
5:54
waterfall and to make sure that if
5:56
targeting is applied or for better
5:58
performing inventory that's being
6:00
packaged that the appropriate premium is
6:02
being applied to account for that that
6:04
better performance got it so so that
6:07
since Bottoms Up firmly anchored in sort
6:10
of the opportunity cost or potentially
6:12
the direct cost although as you
6:14
mentioned much less relevant in in most
6:17
digital media types of environments so
6:20
if that's an approach where there is a
6:22
good programmatic Baseline what what is
6:26
top down approaching and and when might
6:28
that be appropriate
6:30
right so a top sty model would be used
6:32
for inventory that is differentiated in
6:33
market and has more of a premium value
6:35
proposition so in this case think about
6:37
things like sponsorships branded content
6:41
unique formats really any inventory that
6:44
has a higher propensity to sell
6:46
direct and for this framework an
6:49
opportunity cost would still be
6:51
calculated and used internally to set a
6:54
hard floor for direct sales but the
6:57
actual rate card itself would be based
6:58
on other attributes such as elasticity
7:02
looking at average sold rates and fill
7:05
rates in in the past uh as well as other
7:08
feedback such as the willingness to pay
7:10
what types of clients you may be
7:12
interested in that type of product and
7:14
how does it compar to other Solutions in
7:16
the market for the more unique
7:18
Solutions there's less options so the
7:21
premium should be
7:22
higher um so the the main point of this
7:26
type of model is really to protect the
7:29
Val of inventory that is the most
7:32
unique and and the most scarce to make
7:34
sure that uh overall rates aren't being
7:37
deluded so one of the things I'm hearing
7:39
is it's potentially with inventory
7:41
that's that's more valuable maybe where
7:43
there isn't a clear market in terms of
7:46
you know lots of liquidity there isn't
7:49
an auction or exchange that you can
7:51
Benchmark against does does that mean
7:54
you and I are sitting here uh early in
7:56
the year 2025
7:59
um would this apply more to products
8:03
that are more Niche like CTV digital out
8:08
of home things where they're either you
8:12
know less traded programmatically where
8:15
there um and if so how would you think
8:18
about applying the framework to those
8:20
areas well in some cases there may be
8:22
products that don't backfill at all and
8:24
they're only sold direct and in that
8:27
case we'd want to look at the inherent
8:28
value of the inventory
8:29
uh so again going back
8:31
to are there other products that we can
8:33
incor against um what does performance
8:36
look like and again how unique or how
8:38
differentiated is that inventory in
8:40
Market um and has anything similar been
8:42
sold in the past to use as an anchor as
8:44
well um so again in this case the
8:48
process is a little more of a gland of
8:50
Art and Science than pure science where
8:52
we take those qualitative and
8:53
quantitative inputs to kind of calculate
8:56
a starting point for how the inventory
8:58
should be valued so so let me let me um
9:01
push on that a little bit because even
9:03
though I acknowledge it is a little bit
9:05
more art than science I I've talked in
9:08
in other formats or other forums about
9:10
this not idea of using different
9:13
attributes to think about the inherent
9:16
value of products and those attributes
9:18
could be things like the size of the ad
9:21
unit certainly a larger billboard by the
9:24
side of the road is going to be more
9:25
valuable than a smaller one um the type
9:28
or size of audience that is going to see
9:31
a placement uh the location of the
9:34
placement the persistence so you know
9:38
clearly there are attributes that you
9:40
can use to to factor in um how might you
9:44
think about which of those attributes
9:47
might be uh appropriate to use because
9:50
I've just thrown five things at you you
9:52
you mentioned a number of others
9:54
including things like sell through rate
9:56
and presumably you aren't going to try
9:59
to use 10 different attributes in your
10:02
model how if someone's hearing this for
10:05
the first time and they're like yeah
10:06
tops down sounds like the way I might
10:08
need to do this how should they think
10:10
about what are the right attributes for
10:13
them when they're thinking about trying
10:14
to price some
10:17
products I think the starting point is
10:19
the nature of the unit um so if we're
10:21
talking about a larger format or
10:24
something that can be anchored against
10:26
standard formats you'd want to look at
10:28
the size of the unit maybe normalized to
10:30
a price per pixel and use that a
10:32
starting point to um come up
10:35
with where where that that product
10:37
should be gored in other cases there
10:40
there are hen products that are
10:42
completely groundbreaking completely new
10:45
and in those cases the process is a bit
10:47
more convoluted because there's not much
10:49
to go off of um you know so those cases
10:53
that's where I think it becomes more of
10:55
an art um where we pull in some of that
10:58
that qualitative feedback
11:00
um what is the value prop in the market
11:02
or any other Publishers offered
11:03
something
11:05
similar and that's where you know it is
11:08
a plan between the qualitative and
11:11
quantitative got it got it um let
11:15
me you I don't know if what we've been
11:17
discussing is relatively easy stuff but
11:19
let's get to some of the toughest stuff
11:21
or some of the stuff that yield manages
11:23
uh some of the questions they hate being
11:25
asked and I think one of the questions
11:27
that I used to hate being asked asked
11:29
was how to do benchmarking how to
11:32
compare how we as an organization might
11:35
be doing against our competitors or
11:38
against the market how are we doing this
11:40
year versus others so when you think
11:44
about
11:45
benchmarking um how would you think
11:48
about the sources that you might want to
11:51
go to and and and let's focus on on the
11:53
positive because I think you and I could
11:55
spend all day on some of the inherent
11:58
challenges or challeng or problems with
12:00
some of the vendors in the space if we
12:03
if we focus on the positives what are
12:06
your experiences with the best sources
12:08
for competitive benchmarking and in
12:10
particular what are the things that
12:12
people need to be careful to watch out
12:14
for right and I agree there's always
12:17
high demand to to pull in competitive
12:21
benchmarks in the past we've gotone
12:23
requests from multiple groups so yield
12:25
would use as a consideration when when
12:28
pricing but other teams such as BISD may
12:31
use those insights to calculate fair
12:33
market value when working on external
12:35
partnership so these insights are are
12:37
very valuable
12:39
and one
12:42
thing that I want to point out is
12:44
there's really not a silver bullet so
12:46
there's not one source in the market
12:48
that will have everything that a team is
12:52
looking for to to truly understand
12:54
actual rates in the market so the breast
12:56
approach is to use various
12:59
sources or a few sources and
13:03
triangulate um on the programmatic side
13:07
there's a few vendors that work with
13:08
Publishers on reporting
13:10
Solutions ingest the data Aggregate and
13:14
create bench a bench workking Suite um
13:17
I've worked with vendors such as stack
13:19
which is now part of operative bird
13:22
Intelligence on pulling in some of these
13:25
insights the data is extremely reli able
13:29
because it's coming from actual
13:31
campaigns and they allow for a multitude
13:34
of different filtering so you can pull
13:36
in Trends by transaction type our size
13:39
the key variables that would drive that
13:40
would drive rates in in the
13:42
market um however when we use those
13:46
sources it needs to be an a directional
13:48
way because there's other variables that
13:50
are as transparent such as targeting
13:53
layers or the add density on the page
13:56
which also which also affects rate
13:59
and may not be reflected in in the
14:01
aggregation of of that data um but
14:03
overall to look at General trends of the
14:05
market the sources are extremely
14:09
valuable there's another tool called
14:12
pathmatics oh sorry James no keep going
14:14
keep going I say there's another tool
14:16
called
14:17
pathmatics which I've used in the past
14:20
and it's primarily used as a share of
14:22
wallet tool to understand how ads are
14:26
monetizing on on different Publishers
14:28
different
14:29
[Music]
14:30
advertisers but it does include a
14:32
breakout of of channels um to
14:35
differentiate between direct and
14:36
indirect so that's another valuable tool
14:38
to understand what the channel mix looks
14:40
like and how pricing might be Blended
14:42
for some of these
14:45
sites when when you are seeing these
14:49
kinds of Dynamics let's stay with the
14:51
examples you gave first of of stack and
14:54
B where they use real
14:56
campaigns as you mentioned um it has the
15:00
advantage that if you're using real data
15:03
you know it's grounded in reality but
15:06
you may not know which reality uh it's
15:10
part of and by that what I mean is how
15:12
how do you factor in the fact that how
15:15
different individual buyers may be
15:18
swaying the the data
15:21
set right so those tools do break
15:24
out by DSP by SSP however to your point
15:28
the data not weighted um so it's
15:32
difficult to understand which demand
15:34
sources are are driving the average so
15:37
again in those cases the best approach
15:39
is to use directionally to understand
15:42
overall
15:43
movement and when the trends don't
15:47
necessarily in line with other
15:48
sources um
15:52
or there's something behind the scenes
15:54
um that needs to be taken with the grain
15:56
of salt um sometimes
15:59
we've worked with the account managers
16:01
at those vendors to get a little more
16:03
insight but that is one of
16:06
the the issues with with some of those
16:08
tools it's impossible to understand what
16:10
the true drivers are so it needs to be
16:12
used for more of a 500 foot level how
16:15
how do you think about the potential
16:17
ways you can check your sourcing of that
16:21
so I'll give you a couple of examples
16:23
and you can tell me whether you think
16:24
they're useful you might check against
16:28
your own per performance if you're a
16:29
publisher you know exactly how you
16:32
performed over the last few months and
16:34
so you could check whether the vendors
16:36
are seeing that accurately uh another
16:38
potential way you could check them would
16:40
be to check their performance against
16:43
publicly reported financials for the
16:46
biggest companies whether it's a meta a
16:48
Google maybe some of the biggest
16:50
Publishers they have uh their financials
16:54
that they have to report granted they
16:56
don't always break out their advertising
16:59
businesses in the same way but but just
17:02
to sort of put a bow on this using
17:04
different forms of benchmarking how do
17:06
you go about validating th those
17:08
different
17:10
sources you're right so as actually you
17:12
can bring up earnings reports are
17:13
another valuable resource and they tend
17:16
to provide other metrics that can be
17:20
helpful to kind of understand
17:21
performance so Google and meta both
17:24
provide average fa Revenue per user
17:27
which is another way to kind of
17:28
understand
17:29
and engage the health of the
17:31
business
17:34
um but
17:36
to validate the
17:40
overall data um you know one tactic that
17:43
we have used is to normalize based on
17:46
what we're seeing against our own data
17:48
so if ecpm are off by 15% we'd assume a
17:52
certain
17:53
range um you know plus minus percentages
17:58
think and kind of gauge
18:00
um kind of gauge you know that way
18:02
knowing that uh it's not going to be
18:04
100% precise and and and and I don't
18:08
know if it's an answer to the exam
18:09
example you just gave but many of those
18:13
vendors don't have insight into things
18:15
like whether there is bonus being
18:18
applied against a campaign Etc and so is
18:22
that the kind of reason that you might
18:23
just apply an aggregate factor to scale
18:26
up or scale down their performance
18:28
because you'll you accept that they
18:31
either don't apply bonus we know that X
18:34
percentage of bonus is being used or
18:36
what would be the reasons that you would
18:37
think about you know would would
18:39
influence it one way or the other right
18:41
so when we use data from those vendors
18:43
we have to assume that it's it's fully
18:45
diluted so when we compare internally we
18:47
should use a fully diluted metric but
18:49
then also Trend against our our sold
18:52
rates and understand that you know those
18:54
benchmarks are going to be somewhere in
18:56
between um but again I think the main
18:59
use is to understand how a publisher is
19:02
indexing against the overall movement of
19:04
rates so for example if one month or one
19:08
quarter rates are up by 15% across the
19:11
board in the industry based on these
19:13
sources but the Publishers seeing flat
19:15
to negative Trends it'd be the first
19:18
step to kind of dig in has anything
19:20
changed um what do we think we're seeing
19:23
across the different sources that may be
19:25
driving that that that change and in
19:28
some cases is again it could be based on
19:31
you know the way that the
19:32
sources data is being pulled in where we
19:35
may see that blip using One Source but
19:37
then validating that against another um
19:41
you know it may not be the same Trend so
19:43
in that case there's an understanding
19:44
that again it could
19:47
be due to one of the issues with the you
19:50
know not having full transparency with
19:52
how the course are being calculated okay
19:56
um I want to ask you about new product
19:58
ructions which again I'm apparently I'm
20:00
only asking you tough questions today uh
20:03
which again is one of those areas that I
20:05
think makes a lot of pricing y managers
20:08
cringe a little bit because it's a
20:10
different it's a difficult problem and
20:13
I'll I'll differentiate here I think
20:15
often yield managers you know in a
20:18
status quo type of environment you're
20:21
making adjustments to your rate card
20:23
year over year or quarter over quarter
20:25
and the signals that determine whether
20:27
you do that are based on market
20:29
conditions sell through rates is there
20:32
demand for that type of product and so
20:35
pricing your existing products can often
20:38
be um well understood or relatively
20:42
routine and in contrast pricing of a new
20:46
product launch or a new product
20:48
introduction can be relatively difficult
20:51
because you don't have that Baseline
20:53
history you don't have necessarily A
20:56
sense for how well it's going to be
20:59
received by the market and so I think it
21:02
genuinely is one of the most difficult
21:04
problems that pricing managers face how
21:08
have you thought about new product
21:09
introductions historically in your
21:11
career and what advice do you have for
21:13
approaching
21:14
them so I would say the the first step
21:17
is to understand if there's an a way to
21:21
Anchor the new the new product so cases
21:24
of standard IAB formats they're
21:28
typically is
21:29
enough data you know between internal
21:32
and external resources to find that
21:35
anchor um and when we look at that
21:38
initial
21:39
rate we typically want to filter by the
21:41
key variables that drive performance
21:44
such as art
21:45
size page type maybe a Content here so
21:49
the content of the page more similar to
21:52
a different site
21:54
perhaps
21:56
and you know typically performance
22:00
there's a narrow range of performance
22:03
where it's more or less straightforward
22:06
to devised by an initial price point now
22:09
in cases of a brand new format or
22:11
something extremely Innovative in the
22:13
market that's obviously a little tougher
22:15
in the case of a brand new large format
22:19
going back to that tops down you know
22:22
the first step would be to look at okay
22:24
what is the size of the unit can we
22:25
normalize based on the how much of the
22:28
landscape on the on the pages taking up
22:31
then looking at the functionality of the
22:33
unit um is it intrusive do we expect
22:36
higher performance because of you know
22:38
the layout or the the animation um
22:42
different aspects of of of the new unit
22:45
um and again some of it does come down
22:48
to that qualitative input so is there
22:51
demand for that type of unit in in the
22:52
market is it coming from feedback from a
22:56
specific Advertiser agency
22:59
and we would incorporate that I would
23:01
say for those setes of formats sometimes
23:03
it is diff difficult to get initial
23:06
demand in in the market because you
23:08
don't have that
23:09
Baseline so
23:11
actually moving that inventory and and
23:14
getting initial demand is a challenge uh
23:17
in some cases it makes sense to package
23:20
that in with other inventory um in order
23:23
to again get some demand and understand
23:25
what the performance is to build off of
23:27
adjust race going forward and also
23:30
create collateral so that the teams can
23:32
go in with a greater go to market
23:36
strategy when you think about that new
23:39
product and it's launching is your bias
23:41
to price it slightly on the higher side
23:44
or on the lowest
23:46
side um you would always want to buffer
23:48
in and price at a premium and one thing
23:51
that is always consistent it's always
23:53
easier to bring down a rate after it's
23:55
been established in the market then then
23:57
bring it up uh the latter is is almost
24:00
impossible once a rate has been
24:03
shared over a few months I don't know if
24:06
it's because we worked together for so
24:07
long but obviously I agree with you on
24:09
this now I think the only additional
24:10
thing I would would would say on that is
24:13
it's critical to get a couple of sort of
24:15
pilot customers for that launch and you
24:19
may if they are getting early access to
24:22
a product and we acknowledge that we
24:24
don't really know what the price is so
24:26
it's on the high side that they get some
24:29
temporary sort of pricing during the
24:31
pilot program in exchange for something
24:35
that that they might give be it um
24:38
insights about the performance or the
24:41
ability to participate in a case study I
24:44
think um one of the things we often get
24:46
very hung up on when we talk about
24:47
pricing is the number of the price and
24:51
forget about all of the other attributes
24:53
that go into the relationship and
24:55
certainly when there's a new product
24:57
introduction you are you are trying to
24:59
make sure that new product introduction
25:01
is successful and for that you need case
25:03
studies Market Market information Etc
25:07
and all of that has value and maybe
25:08
worth working with an Advertiser where
25:11
they can get effectively a lower price
25:13
in the short term in in exchange for
25:15
some other considerations that be
25:18
fair I would say too from a claim
25:20
relationship standpoint it also helps
25:22
right if we identify you know top
25:24
advertisers ones where we feel like
25:27
budgets can can be increased some
25:28
opportunity there it makes them feel
25:30
like they're getting something that the
25:32
rest of the market isn't um so from that
25:34
standpoint it's it's very helpful for
25:36
the relationship excellent hey Paul
25:39
thank you for for taking the time to do
25:41
this I as I reflect on the questions
25:43
I've asked you I think I've asked you
25:44
harder questions than than anybody else
25:47
I've had on this uh this podcast so I I
25:50
really appreciate it what is the best
25:53
way for people to to find you online or
25:55
or in any other way thanks James that
25:58
can be found on LinkedIn Paul
26:01
d-s and my email is Lewis Paul
26:04
04 and it be great to chat to more
26:08
detail with anyone that's interested in
26:09
the stuff excellent well thank you Paul
26:12
have a great rest of your day thanks
26:14
James