welcome to decision analyst insider
series webinar and pricing analytics are you leaving money on the table my name
is Cristian I am the marketing director Activision analyst and the moderator
today before I introduce our presenters that
have a few notes for everyone there is a copy of today’s presentation in the
handouts section available for everyone to download along with some relevant
white papers and case studies also please feel free to ask questions by
typing in the chat box we will attempt to answer as many questions as we can at
the end of the presentation if we don’t answer your question during the webinar
someone will respond to your question within a day or two today’s presenters
are John colias senior vice president of advanced analytics John has
approximately 30 years of experience applying advanced metrics and models
using data of many kinds and Elizabeth Horn senior vice president of advanced
analytics who is co presenting she specializes in product and price
optimization and marketing segmentation and with that I will hand the
presentation over to John okay thank you today we’re going to talk about pricing
analytics and we will cover typical business objectives when that pricing
analytic supports some methodology that is used in the pricing pricing analytics
have gone together with pros and cons of the different methods and we’ll give
some examples as well as we go through the presentation today so first of all
pricing analytics is supporting many different business objectives and we’ve
listed five of the most common here to begin with a most common probably
business objective is to optimize or maximize revenue profit or market share
and we basically try to measure the price response and use that information
in order to determine what is an optimal or revenue or profit maximizing price
point in some cases we might for a new product for example want to establish
market share as a first entrance in the market and might also want to optimize
or maximize market share a second common objective
for pricing analytics is to to optimize product line pricing based upon
customers perceived value and this is the idea of using pricing analytics to
determine which features or you know in which products in a product line or more
upper tier products where customers are willing to pay more and which ones are a
little bit lower tier products customers or women are not was willing to pay as
much and what is the relative price between these more upper or higher
priced products in the product line and lower priced products and where is that
optimal price difference so that’s the product line pricing based upon
customers perceived value and a third objective is to customize price based
upon who are actually a customer segment who is the customer and this is kind of
this is really a price discrimination strategy that has to be approached with
caution but in some cases with service or experience based products that it can
also be a pricing objective that firm could use in the actual marketplace a
fourth pricing objective is to price based upon purchase channel so this is a
very old method it’s for example if there is a organic or natural food
product sold and for example Walmart it might is probably priced differently
than it would be in a specialty natural foods grocery store and a fifth
objective revolves around new products and typically with a new product many
times there’s not a benchmark for what price to charge and it’s very important
to enter the market with exactly the best price point for the new product and
there’s special research methods that are used to address this type of issue
in pricing analytics so before we go further let’s talk about what we’re
truly trying to measure with pricing line analytics and that is the price
elasticity and this is an example here of a demand
curve on the left side where the price elasticity and we’ll talk about this in
a moment moment is minus one point zero which implies that a price drop of ten
percent from 200 to 180 causes a 10 percent increase from 100 to 110 in
units of demand a higher an absolute value price elasticity in the right
graph of minus 1.5 it shows that a same price drop of 10 percent from 200 to 180
causes a 15 percent increase or 100 to 115 and unit demand so measuring this
kind of price elasticity is really measuring the response that customers
have to price and this is the foundation of all of the pricing analytics work
that we do to support the various objectives that we talked about so now
we’re going to talk a little bit about pricing analytics different methods that
are used in pricing analytics and the first method would like to discuss is
econometrics demand modeling this is a an older technique still heavily used
today of course but it requires historical data on unit sales and prices
and essentially it’s estimating a line that best fits data on the historical
unit sales and prices in order to measure that price elasticity that we
were talking about and one moment ago it requires that there is enough historical
data long enough time series of it it requires that there are good data for
this type of analysis to to work and it also requires that the data is not just
available for just price but it’s also available for other factors that impact
demand so that the price effect can be isolated separately and uniquely from
those other factors like market size macroeconomic just conditions and
competitor prices that also impact demand so one of the advantages of this
approach is that it uses sophisticated econometrics modelling methods that are
really well grounded in eken theory and econometrics theory and in
statistics and so this well grounding this well-developed discipline results
in unbiased estimates of price elasticity and it’s based upon actual in
market data so there’s a lot of face validity and to this kind of analysis
the disadvantages however are there and particularly the price variation in the
actual market may not be that great it may not be sufficient to to reliably
measure the price effects and in addition we there could be price points
in the analytics for pricing that we need to investigate that are not a knot
that are not present in the historical data so we can’t deliver price
elasticity measurement for price points outside the range of the historical data
and then as we see today with rapidly changing markets for example ecommerce
coming in and disrupting the brick-and-mortar retail business we see
that older data perhaps older than five years old and maybe even even younger
data than that might not be really relevant to today’s market marketplace
where consumer behaviors are changing so rapidly a second type of analytical
approach or methodology for measuring price elasticity is to use choice
modeling and we’re going to refer to this as stated preference choice
modeling for all of you who are in the audience who have done conjoint or
choice-based conjoint or choice modeling with survey data this is what we’re
talking about and basically the respondent is given a task in a survey
where a product is present and not just the product that we’re investigating and
measuring the price elasticity for but a full competitive set around that product
and the respondents states and that’s where the word stated preference comes
from states how many if any of each product would be purchased it could
include a lot of variety in the menu task itself with bundles choices
available and so forth this type of task is very common in
market research and is really a bread-and-butter of a lot of what we do
in marketing research today it has some great advantages it includes competitive
and context effects because the entire competitive set is shown in a choice
screen or a all the major competitive factors are shown in the choice screen
so they’re taken into account when think when these survey respondent makes
decisions in States what they would buy if it can deliver brand specific price
and feature effects because we actually control what were what prices were
varying by brand so that we can actually measure price elasticity a much more
granular detail than we can with other methods and we can evaluate not just a
product but an entire product line and we noted I mentioned before packages and
bundles as well it can deliver using your more sophisticated modeling
techniques such as hierarchical Bayes choice modeling or a latent class choice
modeling segment level or even individual customer level price effects
and if we can include new products in these types of State of choice modeling
choice taps and surveys the products that are not even in the market today
so you can see it has that advantage over econometric modeling where the new
products are just simply not in the data it does have some disadvantages the
these this approach where the respondent focuses on a particular task one without
all the distractions that happen in the real world and in an actual purchase
decision one could it’s typical for the respondent to to overstate the impact of
price but you know with survey choice modeling if we design it really well it
can mitigate some of this effect is overstatement so for example if we
remind the respondent to about what their purchase constraints are then they
will like in the real world make choices if there are there are trade-offs and
they’re trading off benefits and prices in order to make their choices much like
they would in the real world so we’re simulating realism by that so by
injecting as much realism as we can in a stated preference choice task we can get
more accurate price elasticity however as a general rule there is a tendency to
overstate the price effect not as much as with other techniques but certainly
to some extent in stated preference choice modeling there is a solution to
this overstatement and sometimes it could be understatement as well that
comes from state of preference choice modeling and that is to use what’s
called joint stated reveals of preference choice modeling we refer to
that as JSR P for joint stated revealed preference and this is exactly the same
task in terms of the survey as we saw with the state of preference choice
modeling but we have real-world purchased data that’s also included at
the modeling step so that the the using the appropriate techniques to adjust and
calibrate the model the actual price elasticity is automatically calibrated
into a more realistic level because we have actually in markets purchased data
included in the analysis and the development of the price elasticity and
the way it’s measured so the primary advantage of this technique is in
reducing the bias and measuring price elasticity and of course getting the
right measure of price elasticity is critical and when we’re doing pricing
analytics it avoids a false positive this tip it’s more common that stated
choice profile experiments would cause an overstatement of the price effect
then it’s less likely that we’re going to have a false positive where we’re
saying we should reduce price when actually we should in cry increase
products for example it does require specialized software and experienced
modeler to implement however these tools are available and this technique can be
used and has been used before we move on it’s worth mentioning a couple of other
survey techniques that are commonly are frequently
used in pricing analytics and one is the Gabor Granger method which is a case
where we select say for example five price points and we randomly select the
first price point or select the middle price point and just simply at show a
concept or a product to a respondent and ask them directly if they would buy it
at this price and if they say if typically a purchase rating scale was
used such as definitely would buy probably would buy and so forth and if
the respondent says top – box definitely or probably would buy then the
respondent is shown the higher price point to see if they would buy it at
even a higher price point and if they are not taught to box and responding to
show a lower price point and until the lowest price point is reached among the
test price points in a survey experiment so this is the key advantage of this
approach is that it’s easy and inexpensive to program the survey and to
do the analysis the disadvantage though is that it’s a direct questioning
approach it’s subject to more responsible eyes than an indirect
approach such as a stated preference choice modeling experiment where the
respondent is indirectly revealing how much they desire a product by having
trade-offs between benefits and prices among other products there’s no
trade-off there’s no competitive context in its
task and the respondent can kind of gain the system if they notice that if they
say don’t definitely or probably purchase a product that they will get a
lower price point they might wait for an even lower price point so it tends to
overstate the price effect even more so than stated preference choice modeling
another technique that’s frequently used and is still used and everybody many
phones do this is the answer is what’s called the Van westendorf pricing
methodology and the in this methodology I won’t go into the group details
because many of you will know and if the working market
research about this there’s a four question series about whether the
respondent would buy at what price were this product that is shown to the
respondent be too cheap cheap expensive or too expensive and the the key and
then what’s done in this approach is typically a too expensive or too in a
too cheap price intersection point of Sikkim units and lines that are drawn is
determined at the optimum price point with the logic that that there’s no this
is a maximizes the number of people out of the market for which price is not an
issue when they’re buying the product but it’s the key advantage is that it’s
inexpensive to program again and to do the analysis the disadvantage is it’s
really not based on any kind of economic theory again it’s a direct question and
approach and subject to some response bias versus indirect questioning
approaches it really doesn’t replicate anything close to an actual purchase
decision process and and the closer we can get to an actual purchase decision
process replicating that in a survey environment the more accurate the
results would be so as a little bit of a case study in these first three methods
we mentioned become a metric modeling stated preference choice modeling and
joint stated revealed preference choice modeling we took some data and for a
product category that is a consumer good category and this was a fairly loyal
category where a lot of respondents are loyal to the product and brand that they
buy and we found with econometrics modelling of the day of some of the data
for this product category of price elasticity of minus two point one the
price in the market varied somewhat but we see the standard deviation of that
variability of price of zero point four which is lower than what will see with
stated preference and joint stated preference modeling and why is it lower
because in the juice in these bottom two techniques we’re actually injecting some
variability of price we’re controlling the price variability
because we’re using an experimental design to vary price in choice
experiments that are done in a survey environment and with state of preference
choice modeling we found the price elasticity of minus 1.8 when we combine
the stated preference the choice stated responses from the survey with some in
market choice data we found a price elasticity which is lower lower in
absolute value of minus 0.8 and this is a common result that we find with joint
stated revealed preference modeling where we find that the the calibration
of the price elasticity is towards a smaller number and absolute value so
this is an example of the kind of differences we see of the different
techniques in an actual application so now what we’ll do is talk about some
examples of pricing analytics before we do that this is a short tutorial on
price elasticity because we’ve been talking about that a lot if the price
elasticity according to economic theory is greater than one an absolute value
then that implies that if wind price increases as we see in the bottom left
here that revenue would decrease if it’s unit elastic and the price elasticity is
exactly 1 then a price increase would make would cause no change in revenue
and if it’s inelastic with that is to say the price elasticity is less than 1
and absolute value that a price increase would cause a revenue increase and you
can see how this relates to maximizing product or product line revenue and and
we can how we can use price elasticity itself towards that objective
so this first type of objective of maximizing that we talked about earlier
of that of maximizing revenue profit or market share can be accomplished by
stated preference choice modeling or joint stated preference choice modeling
but we can measure the price elasticity for different products and product lines
such as in this graph product a B and C where a has the highest price
elasticity and absolute value if you lastic minus one point four B is a minus
nine point one and C has the lowest it’s a price elasticity and absolute value
it’s an inelastic range if we were to optimize revenue get provided using the
same price point for all of these pre three products in a product line we
would find for example and this is an we would find for example that the price
for the entire product line the optimal price would be about three dollars and
thirty cents in this example however knowing that the price elasticity is
very and how much they vary if we were to allow pricing to vary and set
independently for each of the three products we would find as according to
the theory that that would raise price for the product that’s inelastic and
lower price for the products that are more elastic in order to maximize
revenue and this is an example of using price elasticity that can be measured in
a price in pricing analytics by pricing analytics to support the objective of
maximizing revenue for a product line by varying the prices of the individual
products in that product the second of pricing objective that we talked about
earlier was to optimize product line pricing based upon the customers
perceived value which we abbreviate as CPV customer perceived value and again
this is the kind of experiment where we are the kind of analysis where we’re
trying to measure for different products which ones are the customers valuing
more than others the ones that they value more typically have a lower price
elasticity customer perceived value CPV is inversely related to price elasticity
the absolute value price elasticity and those that they value less tend to have
a higher price elasticity so by measuring price elasticity
we’re actually measuring customers perceived value and we can leverage the
CPD or customers see value by charging a higher price for those products that are
have a higher perceived value and to do this the seller has to be able to
differentiate the products based upon features and the benefits of the
products themselves and communicate those very well the goal is to charge
more for what the customer values more so communicating that isn’t very
important and measuring what it is that they’re valuing is very important the
communication can be done in packaging it can be done in advertising it can
done be done marketing for programs but how do we know what to communicate you
know what is it that the customers are valuing more and we can use choice
modeling to do this in this example with some power drills we see we’ve taken the
drills and broken out some features in brand and some features of the drill
that has different levels like it’s courted or has a long life battery it
has dual speed or not and it’s wider or not and those kinds of things are each
features in the product which in a choice modeling experiment we can vary
individual features measure the CPV for each individual features including brand
and then determine the customer perceived value for the product by
summing those in customer perceived values across the individual features to
get a total value for each product and then because we can measure the total
value for each product but in this manner we can actually measure that how
the price elasticity varies but also why it varies so that we can communicate the
proper things in our marketing programs to to demonstrate to consumers those
features that they actually do value more and to chart in to enable us to
support charging a higher price for a particular product so that’s an example
of using a choice modeling to support the CPV objectives a third objective we
mentioned was customizing products based upon customer segments and
pricing analytics canoes can support this is with choice modeling we can use
a technique such as a hierarchical Bayes choice modeling as we mentioned earlier
to measure individual price elasticity for a representative sample of
respondents and by doing that we can actually measure a distribution of price
elasticity and this data here again was a product category we talked about
earlier we did a case study and it was a very loyal product category and we see
the large bar on the left with a very low price elasticity represented by the
green square here representing that a large part of the distribution is
falling into this low and absolute value price elasticity and so by knowing who
these people are in this bar we can actually charge more for those people
for this product now we have to be very careful about that
typically we need to have a reason to charge more because there may be some
legal considerations or even regulatory considerations and and so you know
reasons might be that it costs more to deliver to a certain customer group or
customer type or the the competition is greater in certain customer segments and
therefore the it’s a bit justifies differential pricing by type of customer
once we know this we can use the the pricing analytics can support this by
looking at the demographics and the digital footprint and lifestyle
segmentation for the people that are have a lower and absolute value price
elasticity in order to find them in order to market to them and then and
actually implement the dip customer differential pricing so a a fourth of
business objectives we mentioned earlier was to customize price based on purchase
channel and for example a manufacturer might be a manufacturer of a generic
brand that is sold in different channels and by measuring price elasticity by
channel we can support varying price by channel so we would do this
by drawing a representative sample of consumers into a choice modeling
experiment where we have different selves for the different shoppers at
different channels for example Walmart Walgreens Target and CBS and this is in
this example here which is hypothetical we have a particular demand with a price
elasticity that’s in the inelastic range for CBS and in the elastic range for
Walgreens Walmart which might suggest that we could raise the price or suggest
the manufacturer could suggest to a retailer that they’re doing business
that they’re set offering their products with to raise the price in a certain
retail environment if they know that that price elasticity is relatively low
and they can provide the evidence from the survey experiment that the price
elasticity in absolute value is relatively low so that’s an example how
we can use pricing analytics to support price variation by purchase channel so
now I’m going to let Beth handle the next topic all right thank you John so
let’s look at an example for another business objective which is optimizing
pricing for a new product so in this example we would first design a survey
that contains a pre concept exposure choice task now this choice task does
not include the new product it only includes current products next we would
expose survey respondents to the new product concept and we would obtain
their reactions using purchase motivation questions such as purchase
intent value uniqueness believability could be appeal and as a last phase we
would introduce a post concept exposure choice task so this time the task
includes the new product so here’s an example of a pre concept twist ask for a
canned beverage category and as you can see the task displays what we’re calling
the current market current products respondents can click on an image to
enlarge they can about the product you can even have a
have a description displayed and the test for the respondent is just how many
have you purchased of the one shown in the past 30 days typically five or so
tasks would be presented to the respondent for this pre concept choice
task and then they would go through the product concept of a new product concept
and answer the various purchase motivation type questions we would then
show them a pose concept choice task and again now the new product is showing up
we ask a similar question as with the pre thinking of the drinks you’ll buy in
the next 30 days so how many of each of these would you buy pricing varies the
attributes for the new product can vary across typically five screens so we
would pool together or combine the selections made in the pre and post
concept exposure choice tab and model them to measure price elasticity for the
new product so there will be some overstatement of the new product
purchase intent and we use the purchase motivation questions to calibrate or
adjust down the enthusiasm that consumers typically have for new
products within a survey setting we use standard volumetric forecasting methods
to estimate the percent trial for that new product and once we apply that as a
calibration to the model the price elasticity results can be used to find
an optimal price for that new product ok great thanks Beth so we have essentially
talked today about different types of pricing objectives that can be supported
by pricing analytics some of the methodologies and the pros and cons and
given a little bit more detail and some examples about how these methodologies
can support the different objectives and a little bit about the
implementation strategy and what’s important in implementing particularly
pricing objective and how the analytics can support that is raw so now we’re
we’re ready and open for questions okay so John it looks like we have a few
questions and the first one is what is an example of an industry that could
implement different price points for different customer segments yeah that’s
a good question because it’s not as common strategy some that might come to
mind would be maybe a personal trainer it’s more industries that are service or
experience based where the individual is receiving some benefit in a service or
experience in experiencing that benefit and they can perceive some value
differences differently than other consumers so that might be an example a
personal trainer would be an example okay housecleaner maybe or esprit
members in the life coach and the exec it also could be true okay our next
question is what is a really the feasibility or
ability of someone to get that in market data and other I think they’re they’re
probably talking about it sounds like a great technique but how can we get that
data for the revealed preference yeah that’s another great question I think
that this technique has not been used as much as it should have been used in
recent years because it’s difficult to get the in market data and especially
the pricing with that in market data and it requires a lot of effort but one of
the things that’s happened in recent times is that with e-commerce coming in
we have a lot of those prices that are actually online and even if they’re not
the same prices as might be in in stores for example and brick-and-mortar in a
retail setting in econometrics and in choice modeling we have something called
an instrumental variable and so as long as the prices that we get on
scrape that we might scrape from often online for example from the web are
correlated with the prices that are in store for example then we then those
prices satisfy the requirement of being an instrumental variable and we
theoretically get an unbiased measurement of price elasticity but
that’s a good question I think we’re getting to a market place today we’re
going to be we’re swimming and data we had lots of data available it’s going to
be easier to get that in market data to do this type of technique okay and while
you’re talking on our first question we did have a an attendee remind us that
uber is trying this differential price segment and I can’t believe we forgot
about that and so far I’m not entirely sure how successful it is but that is an
example within there using real-time data feedback to adjust those prices
thank you thank you for that so the final question that we have is how large
or do you think is the bias in price elasticity estimates and choice modeling
yeah you know it really it depends so we have a well-designed choice experience
and you know that then others that might not be so well designed I think the
closer you get to realism in a choice experiment where consumers are thinking
about particularly their constraints their budget constraint the the more
realistic the price elasticity however it is generally true that in a survey
setting the respondent is more focused on the task at hand the purchase
decision than they might be in the actual purchase in the marketplace so as
a general rule the way choice models survey based choice models are done in
market research in general I’d say as a general rule there’s probably an
overstatement some couple of studies have been done in
the past and presented for example the advanced research techniques form in the
past where there were in consumer packaged goods categories a sort of a
two to one ratio of the amount of overstatement so that the price response
was doubled and survey based choice modelling when compared with scanner
data example so I think it’s true that there is a bias how much bias kind of
depends the other thing that we might say is that if we have a price category
product category where prices are very people are very sensitive to price we
may be interested primarily in the relative price elasticity so it might
not be as essential that we get the actual level of price elasticity but the
relative price elasticity is what’s important okay terrific well that’s all
the questions we have currently so we will turn the presentation back over to
Christy for wrap up thank you John and Beth and thank you everyone for
attending today’s insider series webinar if you have any questions please feel
free to email John and Beth our next insider series webinar will be Wednesday
September 6 decision analyst Jerry Thomas and clay debt law will be
presenting ignition liftoff accelerating new product development we hope you
enjoyed today’s session and are looking forward to seeing you for next webinar
have a wonderful rest of the day thank you you