00:00:01
we're here today with Wharton marketing
00:00:03
professor peter fader and wharton
00:00:05
professor of operations information and
00:00:07
decisions senthil Vera Gavin to talk
00:00:09
about their new paper which discusses
00:00:11
dynamic pricing of major league baseball
00:00:12
tickets thanks for being here great to
00:00:15
be with you Rachel so first off could
00:00:17
you guys talk to us a little bit give us
00:00:18
a brief summary of the paper what did
00:00:20
you study and what were you looking at
00:00:21
so in a major league baseball in
00:00:24
professional sports in the entertainment
00:00:26
world in general there's been this this
00:00:28
Great Awakening that the action take
00:00:30
care about the business side of the
00:00:31
business it's not just a matter putting
00:00:33
the best players on the field or the
00:00:35
best performers on stage and of course a
00:00:37
big part of that is pricing so for years
00:00:40
and years again all these different
00:00:41
kinds of companies have just come up
00:00:44
with arbitrary prices and they've kind
00:00:46
of relied on secondary markets in order
00:00:49
to kind of reach the white the right
00:00:51
equilibrium well it's great to see that
00:00:53
a lot of major league baseball clubs
00:00:54
among others in this this general area
00:00:56
are finally getting smart and saying we
00:00:59
want to take control of this we want to
00:01:00
set the right prices and part of that
00:01:02
means dynamic pricing part of that means
00:01:05
adjusting the price is over time some
00:01:07
charging different prices different
00:01:09
people depending on the nature of the
00:01:11
game and so on so a lot of clubs have
00:01:13
been just trying it out but not a lot of
00:01:16
them have actually stepped back to say
00:01:18
is it working can we do it better what's
00:01:22
the incremental impact of one kind of
00:01:24
pricing policy or another so we were
00:01:26
very fortunate to be able to work with a
00:01:27
club that was asking those kinds of
00:01:29
questions and through it just a really
00:01:31
clever data set and some pretty clever
00:01:33
modeling I think we came up with some
00:01:35
pretty good answers and some tell you
00:01:37
everything dad it's a it's a very
00:01:38
interesting problem it's a confluence of
00:01:41
exciting cutting-edge research with a
00:01:44
practical application that goes directly
00:01:46
onto the field literally in this case
00:01:48
right so and now so can you guys talk to
00:01:51
us a little bit about you found some
00:01:52
interesting key takeaways from this
00:01:53
paper a bit might surprise a lot of
00:01:55
people who are out there buying baseball
00:01:57
tickets or maybe not one of the things
00:02:00
that we found out is people talk a lot
00:02:03
about dynamic pricing it has a lot of
00:02:05
customer response to dynamic pricing a
00:02:09
well-chosen price even if it's static
00:02:11
just as well that's one thing we learned
00:02:13
from this project
00:02:15
which we had some inkling about but we
00:02:17
were really surprised how well a
00:02:19
well-chosen static price dead that's one
00:02:23
of our surprising findings and of course
00:02:25
there's the flip side to it as well
00:02:26
which is dynamic pricing isn't a panacea
00:02:29
just because you're varying the prices
00:02:31
doesn't mean you're necessarily making
00:02:33
more money and in fact in this
00:02:34
particular case if we look at the
00:02:36
dynamic pricing policy at this one club
00:02:38
followed over this one portion of a
00:02:40
season they actually they actually lost
00:02:44
money I relative to to the the static
00:02:47
policy that they had at the beginning so
00:02:49
it's it's funny that just right balance
00:02:50
between what should be dynamic and
00:02:53
essential said in some cases we can pick
00:02:55
the just right prices why even bother
00:02:57
changing things at all now but i would
00:02:59
think that picking the just right price
00:03:01
is easier said than done that is true if
00:03:05
if there is a magic bullet so to speak
00:03:09
that's distributable we would be able to
00:03:11
do that right I think context specific
00:03:15
application is very important here it's
00:03:18
it's a relationship as bit often talks
00:03:21
about between the customers you want to
00:03:23
serve and the team you wanna run and the
00:03:25
organization one around and in this case
00:03:28
that data is useful to understand what
00:03:30
kind of policies would work yeah we can
00:03:32
improve but it's very specific data has
00:03:36
that information they can convert the
00:03:37
data into good pricing policies and I
00:03:40
think that's why it's such a great
00:03:41
collaboration over here because I do
00:03:43
spend my time thinking about the
00:03:45
relationships and thinking about all the
00:03:47
nuances of where demand comes from and
00:03:50
sentio and his colleagues in our oh I d
00:03:53
department it's been about I'm thinking
00:03:54
about optimizing and very often each of
00:03:57
us doesn't do justice to the other side
00:03:59
so our build these really great
00:04:00
descriptive models but I fall short on
00:04:02
the so what and very often folks in the
00:04:05
optimization world will build overly
00:04:07
simplistic models because it enhances
00:04:09
the optimization part this is that just
00:04:12
right combination we have a really rich
00:04:14
description of how people are buying
00:04:16
tickets when and for what sections and
00:04:19
what are they willing to pay for so it's
00:04:20
a nice story about customer behavior but
00:04:22
it lends itself pretty well to the
00:04:24
optimization as well that's right as I
00:04:26
said it's a confluence of how people
00:04:27
behave
00:04:28
and what models you can run in the
00:04:30
background and how we can bring it to
00:04:32
business applications today yeah and now
00:04:35
you guys had some pretty interesting
00:04:36
ways like one thing that I think really
00:04:38
comes out in the papers there's lots of
00:04:39
different ways that a baseball team in
00:04:41
particular and probably really anyone
00:04:43
that wants to do this good dynamic price
00:04:45
I mean so you talked about the seat and
00:04:47
you talked about how well the team is
00:04:49
doing and a bunch of different ways to
00:04:51
do this so we talked about it a couple
00:04:53
of different levels so as you just
00:04:55
mentioned Rachel a lot of it would be
00:04:56
the factors that should be taken into
00:04:58
account when deriving a dynamic pricing
00:05:00
policy and you just mentioned a few of
00:05:02
them there and it's great that and
00:05:04
that's not an academic exercise that a
00:05:06
lot of professional sports teams and and
00:05:09
and other kinds of businesses are
00:05:10
starting to take those factors into
00:05:12
account so one is on the input side what
00:05:15
factors should we be looking at and how
00:05:16
to adjust for them but then there's on
00:05:18
the kind of on the output side in terms
00:05:20
of setting the policies and again this
00:05:22
is sent those expertise about you know
00:05:24
should we be looking ahead or not well
00:05:26
anyway you can talk more about the
00:05:28
nature of the different kinds of pricing
00:05:30
policies and to me it's been an
00:05:31
education just to think about the
00:05:33
different ways that we can go out there
00:05:34
with policies I mean we're we're all
00:05:36
learning different things from this
00:05:37
problem as we talked about it's very
00:05:39
class disciplinary and one of the things
00:05:42
is how far do you look ahead when you
00:05:44
said your dynamic pricing policies do
00:05:47
you look ahead ten games three games how
00:05:49
often do you change how do you
00:05:51
communicate that and these things matter
00:05:55
and these things are just customer
00:05:57
responses and that's going to feed back
00:05:59
into the policies that you're going to
00:06:01
come up with so people think of dynamic
00:06:04
pricing as an evil or a panacea we
00:06:07
talked about it the truth is somewhere
00:06:09
in the middle it's you can go to the
00:06:12
same solution and do it poorly because
00:06:14
of how you use the information so in
00:06:17
this case before example found out if
00:06:19
there is more people in the group that's
00:06:22
buying tickets they are more likely to
00:06:24
buy from certain sections of the stadium
00:06:27
and in my batter sections of this cherry
00:06:30
that are surprising to me as a marketing
00:06:32
person maybe that's not immediately
00:06:34
surprising but that's not what the
00:06:37
models usually assume so it becomes
00:06:39
relevant to understand your customers
00:06:42
to take the data from your from your
00:06:45
understanding of the problem and apply
00:06:48
it and start thinking about dynamic
00:06:50
pricing so how many games you should
00:06:52
look at should you look at your
00:06:53
opponents should you look at the day of
00:06:55
the game the weather all these things
00:06:57
models did you revisit if they go into a
00:06:59
bad streak yeah when you go into a bad
00:07:01
streak then you cannot be charging too
00:07:04
high so you have to think about revising
00:07:05
prices downloads I mean just kind of
00:07:07
think about sometimes there's a team
00:07:08
that could be you know they could tell
00:07:10
they could say in April well this is the
00:07:11
team that's going to the World Series or
00:07:12
the Super Bowl or whatever you want to
00:07:14
say but then by September they've tanked
00:07:17
well that's actually one of the issues
00:07:19
that happened in this case so actually
00:07:21
the whole setting of the model is kind
00:07:23
of interesting that this particular
00:07:24
major league baseball team didn't use
00:07:27
dynamic pricing for the first half of
00:07:29
the season so we were able to calibrate
00:07:30
all of our demand and response models
00:07:33
based on what we see for you know the
00:07:35
first 40 games or so and then we can
00:07:38
project what would happen in the latter
00:07:40
part of the season if they didn't
00:07:42
introduce dynamic pricing so we have
00:07:44
this really good baseline about what
00:07:45
sales would have been and we look at the
00:07:48
difference in order to say how well did
00:07:49
the dynamic pricing work well we didn't
00:07:51
know it at the time but the team did not
00:07:53
do well in the second part of the season
00:07:55
they they went way way way down and so
00:07:57
it's not so much that we can blame the
00:07:58
dynamic pricing policy but the team went
00:08:01
into an area that we hadn't observed in
00:08:04
the first part of the data set so you
00:08:06
know that's just the nature of building
00:08:08
a model and kind of rolling with it that
00:08:10
sometimes things will happen out there
00:08:11
that you can control you can anticipate
00:08:13
right now tell me a little bit so if im
00:08:16
a im another major league baseball team
00:08:18
or even a minor league baseball team or
00:08:19
anyone where this could be anytime soon
00:08:21
this could be applied any number of ways
00:08:22
i know you've done something with some
00:08:24
things with like concert tickets things
00:08:26
like that Peter my boys are now
00:08:28
nonprofits which I looked at it too
00:08:31
that's exactly I mean you you are a
00:08:33
baseball team you know mindedly clean
00:08:35
you're an international team you know a
00:08:37
soccer team cricket team or or even a
00:08:39
non-profit arts organization you have
00:08:42
already the data that can tell you what
00:08:46
how you can think about this problem
00:08:47
where we come in and tell you is this
00:08:50
information is very key it's important
00:08:53
and this is how we can use this
00:08:54
information
00:08:55
and I just shared policies so on one
00:08:58
hand the I think the the framework that
00:09:01
we've built is fairly general because
00:09:02
we're using pretty common set of factors
00:09:04
and data structures and so on but the
00:09:07
specific results I'm not sure that we'd
00:09:09
want to generalize from those I think
00:09:12
the particular weight that different
00:09:14
factors might have the relative
00:09:16
differences between different kinds of
00:09:17
pricing policies that is going to be
00:09:19
context specific but I am pretty
00:09:21
confident that if we took the overall
00:09:23
model and then recalibrate it in another
00:09:25
setting it would continue to work well
00:09:27
maybe with just slightly different
00:09:29
results what I'm wondering so if I'm REE
00:09:30
if I'm an organization i'm looking at
00:09:32
this how might i practically apply the
00:09:34
research part of it is just gaining the
00:09:37
overall understanding so before we even
00:09:38
worry about specific estimated
00:09:41
coefficients or any of the numbers that
00:09:43
are in the paper as we've said before
00:09:45
just thinking about what are the factors
00:09:47
that you'd want to take into account and
00:09:49
operationally can you create a situation
00:09:52
again that we were fortunate to have
00:09:54
here where we have this control
00:09:56
condition where there was no dynamic
00:09:57
pricing going on so we get a clean read
00:09:59
so once we kind of press that button and
00:10:02
start it we can understand what's
00:10:05
happening what's attributable to it so
00:10:07
there's things both in terms of what we
00:10:08
put into the model as well as just the
00:10:10
overall way that it played out in
00:10:11
practice that that can be generalized
00:10:13
and I hope that folks would focus more
00:10:15
on on those kind of operational factors
00:10:19
more than specific results arising from
00:10:21
it that's very important point actually
00:10:22
a lot of times people say yeah you could
00:10:26
do this but what if it happens what if
00:10:28
something else happens and by doing this
00:10:31
kind of experimental data base study
00:10:33
you're able to answer those questions
00:10:36
using counterfactual ti what if the team
00:10:38
got on a hot streak this is what you
00:10:40
should be doing what if the show is so
00:10:42
successful or what if the lead actor
00:10:45
leaves what do you do and we can have
00:10:48
answers to these kind of questions using
00:10:50
the model applications and as pitted the
00:10:53
actual results may vary from case to
00:10:56
case but the fundamental application
00:10:58
remains the same now is there a story in
00:11:01
the media that you guys think really
00:11:02
relates to this research that really
00:11:04
kind of puts it into a new light door
00:11:06
relates to it in general I think there's
00:11:09
a
00:11:09
not only a current story but an ongoing
00:11:11
story as I said at the outset justice
00:11:14
all this this consternation among
00:11:17
professional sports organizations about
00:11:19
the existence of and the profitability
00:11:21
of secondary markets so a lot of this
00:11:24
project arose from a large-scale project
00:11:26
I had done with Major League Baseball I
00:11:28
was very pleased that one of the clubs
00:11:30
wanted to go further with it in that
00:11:32
project with MLB the questioners should
00:11:35
they continue to use stubhub as their
00:11:37
exclusive secondary ticket provider and
00:11:41
most of the teams agree to do so some of
00:11:44
the teams opted out including the New
00:11:46
York Yankees and just this week the
00:11:48
Yankees change their policy on that and
00:11:51
they're they're coming back to step up
00:11:52
albeit with conditions but it's but the
00:11:55
point is that if this this whole
00:11:57
situation is far from settled so you
00:11:59
have a lot of teams that are doing the
00:12:01
kinds of modeling experiments like we
00:12:02
see going on over here and as a whole
00:12:05
it's it's going to be just a lot more
00:12:08
changes taking place we hope that it's
00:12:10
not just trial and error we hope that
00:12:12
it's going to be the club's include and
00:12:13
the secondary ticket providers are just
00:12:15
getting smarter and and and actually in
00:12:18
some sense reducing the opportunities
00:12:21
for arbitrage and some of the game
00:12:23
playing that goes on just by
00:12:24
anticipating what the right prices will
00:12:26
be and hallo very under different kinds
00:12:28
of circumstances yeah pretty much an
00:12:30
ugly bunted that I mean do you feel like
00:12:32
the secondary markets have really put a
00:12:33
lot of pressure on the teams or on
00:12:35
concert I Ranh Bay Aaron concert
00:12:37
promoters or anybody just to really look
00:12:39
more closely at this from their
00:12:40
perspective just because they know the
00:12:42
secondary market is out there and it's
00:12:44
something that's very relevant to anyone
00:12:47
who's buying tickets in a way the
00:12:49
existence of the secondary markets has
00:12:51
and the success of them has been a great
00:12:53
wake-up call to these organizations the
00:12:55
fact that they're now putting so many
00:12:57
resources internally into doing the kind
00:13:00
of modeling and experimentation that
00:13:02
were endorsing here it used to be that
00:13:03
they would take all the analytic talent
00:13:05
and save it for the money ball people
00:13:06
let's go out there and just find the
00:13:08
best players but now they want to find
00:13:09
the best analysts who can help them sell
00:13:12
tickets at the right prices to the right
00:13:14
people it's good to see that there dare
00:13:16
I say leveling the playing field exactly
00:13:18
i mean at some level they owned the
00:13:20
content as we speak right
00:13:23
they own the content they can control
00:13:25
the content they can they understood
00:13:26
that concert better so who else right
00:13:29
can benefit more from these kind of
00:13:32
pricing adjustments than the content
00:13:34
providers themselves right i mean you
00:13:36
can you can get the best players but you
00:13:37
still have to get butts on the seat now
00:13:39
do you feel like there are i think i
00:13:41
mean i think there's a lot of views
00:13:43
about Dana epidemic pricing out there
00:13:44
you've got people to think it's the
00:13:45
worst thing in the world and people
00:13:46
think that's the best thing ever I mean
00:13:48
do you think there's other
00:13:48
misconceptions that this research kind
00:13:50
of dispels one of the things that again
00:13:55
we going back to what we got surprised
00:13:57
by is you could do very well with the
00:14:00
well set static pricing again well set
00:14:03
that is a hard problem so dynamic
00:14:05
pricing gives you flexibility to adjust
00:14:09
and this is very important it'll
00:14:11
actually can help a team to reach out as
00:14:14
fans and consumers to it's not
00:14:16
necessarily the bad thing for both the
00:14:19
team and the facts under consider that
00:14:21
might be the biggest revelation not just
00:14:23
from this one study but from this this
00:14:24
broader project that I was involved with
00:14:26
and when I first started talking to
00:14:29
Major League Baseball they were very
00:14:31
reluctant a lot of clouds are very
00:14:33
reluctant to try dynamic pricing it
00:14:35
seemed first of all scary it's unknown a
00:14:37
second they worried about some kind of
00:14:39
backlash that fans would feel that
00:14:41
they're being gouged and so on it's been
00:14:44
great to see how well-received dynamic
00:14:48
pricing has been the clubs have been
00:14:49
careful about it and so they have an
00:14:52
overreach there's been very few stories
00:14:54
about any kind of backlash and from the
00:14:56
fan side we're getting used to it we're
00:14:59
understanding that the what we see
00:15:01
whether it's with with airplanes now
00:15:03
with professional sports when it comes
00:15:05
to a ride-sharing services that that
00:15:09
dynamic pricing is here to stay and as
00:15:11
long as companies are doing it in a
00:15:13
smart way and as long as they have the
00:15:14
long run in mind and not trying to
00:15:16
squeeze as many dollars they can right
00:15:18
now it really can be in everywhere in
00:15:20
everybody's best interest now what's
00:15:22
next for this research where are you
00:15:24
going to go next with it lots of
00:15:26
direction is actually I mean we we have
00:15:28
interest from many places so we're
00:15:31
actually we just hope there is more time
00:15:34
to do all this stuff and
00:15:36
and better data too we were very
00:15:38
fortunate to get that the data that this
00:15:40
one club provided but there's so much
00:15:41
more than they could provide so there's
00:15:43
other sources of revenue such as
00:15:45
concessions and merchandise that they're
00:15:47
selling there can be other sources of
00:15:49
customer behavior in this case we didn't
00:15:52
have full granularity to track exactly
00:15:55
which household was buying which ticket
00:15:56
at which time so we could link that
00:15:58
together to come up with a better notion
00:16:00
of what the true demand is so in many
00:16:03
ways this is the this is the tip of the
00:16:04
iceberg this is the top of the first
00:16:06
inning I think it's going to be a long
00:16:08
game ahead for understanding the role of
00:16:11
dynamic pricing in this setting and so
00:16:13
many others well I guess we'll see you
00:16:14
back for your next to your next base hit
00:16:16
so sex L Peter thank for thank you so
00:16:17
much for being to Magnus
00:16:36
you
