00:00:01
we're here with Raj sivakumar who is
00:00:05
head of the travel technology and
00:00:06
strategy unit at wns a global business
00:00:10
process management company and Peter
00:00:14
fader who is a warden marketing
00:00:15
professor and most recently co-director
00:00:18
of the Wharton customer analytics
00:00:20
initiative and we're also joined by Mike
00:00:23
Nemeth who is head of the insurance
00:00:25
practice in north america for wns so
00:00:29
welcome everyone good morning good
00:00:31
morning so people are talking a lot
00:00:34
about big data and customer analytics
00:00:36
what is customer analytics and why
00:00:39
should companies pay attention to that
00:00:41
so on one in customer analytics has been
00:00:44
around forever from the time that
00:00:46
marketing as we know it today was born
00:00:49
let's think about the 1950s or so when
00:00:52
we started realizing that customers are
00:00:54
different from each other and that
00:00:56
there's different ways that we can meet
00:00:57
their wants and needs and anticipate
00:00:59
what is that they might want next and
00:01:01
get smarter about how we'll deliver it
00:01:04
so we started collecting a lot of data
00:01:06
started with demographics sprinkled in a
00:01:09
little bit of behavior start asking
00:01:11
questions about attitudes started
00:01:14
getting physiological measurements as
00:01:16
well then let's talk a mix in a little
00:01:18
bit of social too so a lot of it is both
00:01:23
being smart about the kinds of data that
00:01:25
we should be collecting in order to make
00:01:27
better decisions but then the analytics
00:01:30
part is getting beyond the data or more
00:01:34
specifically below the data it's telling
00:01:37
stories about the true underlying
00:01:39
unobservable processes that are driving
00:01:42
that data and driving business success
00:01:44
so if you think about analytics one of
00:01:47
the ways that we like to break it down
00:01:48
is into three broad buckets we have
00:01:51
descriptive analytics we have predictive
00:01:53
analytics and prescriptive analytics the
00:01:55
name is a reasonably self-explanatory
00:01:57
but it's interesting to see where the
00:01:59
the boundaries and the synergies are
00:02:01
between them so with the script of
00:02:03
analytics that's just all about the data
00:02:04
so let's collect data let's let's come
00:02:07
up with with suitable summaries of it
00:02:08
let's let's do some data visualization
00:02:10
let's do some data science to really
00:02:13
take the raw data
00:02:15
and best frame what's going on then when
00:02:18
we get to predictive analytics the word
00:02:20
predictive is a little bit misleading
00:02:22
it's not only about prediction but it is
00:02:24
this idea of drawing insights that
00:02:26
aren't directly observable in the data
00:02:29
that's where we want to pull out
00:02:30
people's true underlying propensity
00:02:32
which is going to help us make
00:02:34
predictions and it's going to help us
00:02:35
make better decisions but predictive
00:02:37
analytics I mean that's really quite
00:02:39
literally the heart of the analytics is
00:02:41
is the models that we build stories that
00:02:44
we tell to really understand what's
00:02:46
going on and then we layer on top the
00:02:48
prescriptive part so now that we know
00:02:50
what's really going on and now we can
00:02:52
project what's going to happen next what
00:02:54
do we do about it so how are we going to
00:02:56
optimize if we have a pile of money to
00:02:58
spend how are we going to allocate it
00:03:00
across different kinds of activities or
00:03:02
different customer segments or different
00:03:04
geographic areas so it's all about this
00:03:06
this notion of descriptive predictive
00:03:09
and prescriptive and that of course
00:03:11
leads to the decision making which is
00:03:14
what my colleagues here can talk with
00:03:15
much more expertise about so what are
00:03:17
some of the data sources companies are
00:03:19
using in customer analytics so as I
00:03:22
mentioned earlier there's a lot of
00:03:23
generic data sources some of which are
00:03:26
becoming super hot some of which are
00:03:28
tried-and-true so it all starts with
00:03:30
demographics not to suggest that that's
00:03:33
necessarily the best but it certainly is
00:03:35
the oldest in the most common there's a
00:03:37
lot of companies that even when they
00:03:38
find better data sources they care so
00:03:40
much about just simple observable
00:03:42
characteristics of the customer so it's
00:03:44
going to be things like age and gender &
00:03:47
geography moving a little bit into media
00:03:50
habits what kind of car you own tell me
00:03:53
more about the characteristics of the
00:03:55
zip code that you live in so
00:03:56
demographics will kind of spread itself
00:03:58
out and will sometimes get into things
00:04:00
like I said media habits there wouldn't
00:04:02
be a demographic it would really be more
00:04:04
of a behavior but it's still something
00:04:06
that we often use to label people and
00:04:09
then as we move from kind of who people
00:04:12
are we move to what is it the thinking
00:04:14
about so that's what we're going to move
00:04:16
into attitudes so things like your wants
00:04:18
and needs your frustrations one of the
00:04:21
real common attitudinal metrics that we
00:04:23
focus on today would be net promoter
00:04:26
score so you know would you reckon
00:04:28
and this particular service to someone
00:04:31
else that's just one of a myriad
00:04:33
different attitudinal metrics the other
00:04:36
end there will be different kinds of
00:04:38
behavioral metrics so we might say
00:04:40
doesn't matter what people look like
00:04:42
doesn't matter what they say it's all
00:04:44
about what they do and so that's going
00:04:46
to be the transactions that people make
00:04:49
it's going to be their interactions with
00:04:51
a website it's going to be their
00:04:53
interactions with each other it's going
00:04:54
to be their responses to inbound and
00:04:56
outbound marketing activities and then
00:04:59
we can that gives us a segue to the next
00:05:02
one which would be social so we care a
00:05:04
lot about who someone is connected with
00:05:06
so how many people do you have in your
00:05:09
social graph how many of those links are
00:05:11
inbound people looking at you versus
00:05:14
outbound you're looking at other people
00:05:16
what's the how central are you to the
00:05:20
overall social network so as many
00:05:21
different kinds of social activities and
00:05:24
then a real big one that that's really
00:05:25
taking off today would be different
00:05:27
kinds of physiological measures so if we
00:05:29
think about the whole a wearables
00:05:33
revolution so let's measure heart rate
00:05:36
let's let's track people's eyes let's
00:05:40
let's let's look at that they're at
00:05:42
their at their movements not just where
00:05:43
they're going but how fast they're
00:05:44
moving and so on so so is a range of
00:05:47
different kinds of metrics and the real
00:05:48
beauty of analytics isn't just
00:05:51
collecting a lot of data but it's figure
00:05:53
out ways to do it in a really
00:05:54
synergistic manner that we can draw
00:05:56
insights from these different kinds of
00:05:58
metrics collectively that we couldn't
00:06:00
draw from any one of these types by
00:06:03
itself so what are some of the common
00:06:06
mistakes and companies make when they
00:06:08
collect and use big data as well as when
00:06:11
they deploy analytics tools so what
00:06:14
should they be doing well I think the
00:06:18
first mistake that people make is in the
00:06:21
insurance industry in particular i
00:06:23
should say is that they assume that the
00:06:25
way they have the data organized and and
00:06:28
the data they're storing is going to be
00:06:31
useful in an analytics project and and
00:06:35
it isn't always one of the first
00:06:36
barriers is how can i reorganize
00:06:41
information from the various places that
00:06:44
I have connections to internal and
00:06:46
external how can I organize that data
00:06:48
and how can i transform the information
00:06:52
to make it usable in an analytics
00:06:56
project so it comes as a surprise
00:06:58
sometimes two people they tend to hire a
00:07:02
bunch of expert analytics people they
00:07:08
buy tools they put them all in a room
00:07:11
and they say we have a lot of data which
00:07:14
insurance companies have massive amounts
00:07:16
of data and they say tell us some
00:07:19
insights and it's just really not that
00:07:22
simple the very first step is what data
00:07:24
are we going to use where are we going
00:07:27
to create a new data store that's used
00:07:30
specifically for analytics purposes how
00:07:33
do we manage that data replicate that
00:07:36
data over time is is really the first
00:07:39
challenge that people tend to tend to
00:07:41
face under the travel industry of just
00:07:45
tried what Mike said with the with the
00:07:48
data collection becoming so much cheaper
00:07:50
storing data becoming so much cheaper
00:07:52
unfortunately the emphasis on data
00:07:55
collection okay has overshadowed the
00:07:58
emphasis on analytics and so a lot of
00:08:01
lot of companies lot of lot of people
00:08:03
collect data to what purpose and the key
00:08:06
is to be able to ask the right question
00:08:07
to get to the right answer okay you know
00:08:09
the asking the right question is so much
00:08:12
more important okay and because we asked
00:08:14
the wrong question with the technology
00:08:16
that we have we can really quickly get
00:08:18
to the wrong answer so the emphasis on
00:08:20
analytics the emphasis on interpreting
00:08:22
the data the emphasis not realizing that
00:08:24
it's all about trade-offs is so much
00:08:26
more important in the current
00:08:27
environment I surely agree and and I
00:08:29
think about the old days again I'm a
00:08:31
real historian of marketing and business
00:08:33
there's so much that we can learn when
00:08:35
we didn't have all of this data when
00:08:37
there really was more focused on
00:08:38
decision-making than on data collection
00:08:41
and data management and companies were
00:08:44
pretty good at taking the limited amount
00:08:45
of data and squeezing as much value out
00:08:47
of as possible unfortunately today a lot
00:08:50
of companies are saying well we have all
00:08:52
of this data that we didn't have four
00:08:54
years ago so therefore whatever we knew
00:08:56
back then is irrelevant so think it's
00:08:58
important to understand to think before
00:09:00
collecting data to think about what kind
00:09:02
of data you need in order to address the
00:09:05
specific questions and hypotheses that
00:09:07
you have in mind rather than this idea
00:09:10
of if we build it that is a data
00:09:11
warehouse amazing things are going to
00:09:13
happen so we're all on the same page
00:09:15
about that well some companies as
00:09:19
particularly once in the banking and
00:09:21
insurance industries sit on a lot of
00:09:23
data as you guys mentioned but they
00:09:25
don't really mine it to great effect can
00:09:28
you give more details and information
00:09:29
about the barriers that stop companies
00:09:32
from mining the data to great effect
00:09:35
sure I think it begins with the fact
00:09:39
that the data that's been traditionally
00:09:41
collected in the insurance industry is
00:09:43
not about the customer we have
00:09:47
elementary information like Peter
00:09:49
mentioned earlier gender age location
00:09:52
things of that nature but really most of
00:09:55
the data that is being sat upon by
00:09:58
insurance companies is information about
00:10:01
the risk not about the customer so it's
00:10:04
about the house it's about the car it's
00:10:07
about business as opposed to about the
00:10:10
customer and we can Intuit a certain
00:10:13
amount of customer information from
00:10:17
information about the risk but most of
00:10:19
it is not about the customer so
00:10:21
collecting information specific to the
00:10:24
customer it's a relatively new thing in
00:10:26
the insurance industry and so these
00:10:30
doors are suddenly wide open and like
00:10:32
Peter mentioned just a moment ago all of
00:10:34
a sudden we have this influx of
00:10:37
information that insurance companies
00:10:38
have not had a lot of experience with
00:10:41
and so they don't really know how to
00:10:44
interpret the meaning of some of that
00:10:46
information how to combine it with
00:10:49
information they do have experience with
00:10:52
to come up with good analytic results
00:10:55
and and and I think one of the keys to
00:10:58
making that work is to add domain
00:11:01
expertise to the analytics project teams
00:11:04
and this is sometimes overlooked
00:11:06
unfortunately we hire analysts we buy
00:11:10
tools we have data we think those are
00:11:13
the three components to to produce these
00:11:16
fantastic results and they forget about
00:11:19
the domain expertise that needs to go
00:11:22
into the mix and Raj hit it right on the
00:11:24
nail when he said and and I I think we
00:11:28
have more questions about this so i
00:11:31
won't go too deeply into it yeah at the
00:11:33
moment but analytics is all about asking
00:11:36
the right questions and the people who
00:11:38
know what the right questions are are
00:11:40
the domain experts I do want to take
00:11:43
everything that might just said take out
00:11:45
the word insurance and financial
00:11:46
services and plug in pretty much any
00:11:48
other industry and this name applies in
00:11:50
fact in many ways insurance it might be
00:11:52
a step ahead of many other sectors
00:11:54
because traditionally they have looked
00:11:56
at say risks differently for different
00:11:58
kinds of customers as opposed to a lot
00:12:00
of other sectors that have looked at the
00:12:02
customer in some kind of singular way
00:12:04
there is the customer but but indeed the
00:12:07
idea that our data collection has been
00:12:09
much more focused on the products that
00:12:12
we develop and the activities that we do
00:12:14
to develop and serve those products as
00:12:15
opposed to those previously faceless
00:12:18
nameless customers out there that we're
00:12:20
creating the demand for them that is a
00:12:22
change and I like to believe that a lot
00:12:24
of the activities that I'm doing and
00:12:26
that happened at say in academia in
00:12:28
general are trying to get companies to
00:12:30
kind of wake up and realize that it's
00:12:33
not just a matter of collecting more
00:12:34
data about your products it's about
00:12:36
changing the kind of data the kinds of
00:12:38
questions that you're asking in a very
00:12:39
transformational way and just to add to
00:12:42
what Peter said and Mike said and
00:12:43
unfortunately like Peter exactly
00:12:45
referred to the the issues that Mike
00:12:48
talked to board he cannot take ownership
00:12:49
just in the insurance industry its
00:12:51
industry agnostic and perhaps what kind
00:12:54
of rares that head in the travel
00:12:55
industry and particularly with Airlines
00:12:57
is the issue of trade-offs okay let me
00:12:59
give you a very simple example right so
00:13:01
the marketing department would like to
00:13:03
ensure that they're the customer the
00:13:05
highest year gets the preference in
00:13:08
terms of seat assignment and travel
00:13:09
where is the revenue management
00:13:11
department would like to make sure that
00:13:13
every single passenger pays the highest
00:13:14
price so this trade-off between what do
00:13:17
you charge a customer vs
00:13:18
you alligator cuz allocate a high-value
00:13:20
customer who may not be paying a high
00:13:22
value sorry high value on that flight
00:13:24
becomes a classic trade-off so you know
00:13:26
the companies that understand the
00:13:28
trade-offs better again leverages the
00:13:30
data to understand the trade-offs better
00:13:31
is going to be well served so what's
00:13:34
next for customer analytics so right now
00:13:37
it's been so much about the data and I
00:13:39
think we've we've made it very clear and
00:13:41
I hope that that people resonate with
00:13:43
ASD is that it's not just a matter of
00:13:45
collecting more data so let's go back to
00:13:46
the basic rubric of the descriptive the
00:13:49
predictive and the prescriptive there's
00:13:52
been so much attention these days on the
00:13:55
descriptive part which is let's collect
00:13:56
lots of data let's make lots of pretty
00:13:58
pictures let's do a lot of what we call
00:14:01
data science the problem is when we talk
00:14:03
about data science there's been too much
00:14:05
emphasis on the data and not enough
00:14:07
emphasis on the science and so I think
00:14:10
that the next generation as we start
00:14:12
seeing that there's limits first of all
00:14:13
not only to how much data we can collect
00:14:15
but the quality of data we collect is
00:14:17
going to start saying let's not collect
00:14:19
any more let's think more carefully
00:14:21
about that data let's understand the
00:14:24
processes that are driving in the first
00:14:25
place and let's get smarter about the
00:14:27
ways that we can layer on top different
00:14:30
kinds of prescriptive or optimal
00:14:33
elements so I think we're going to see I
00:14:35
don't want to say a shift i'm not saying
00:14:37
we're moving away from data by any means
00:14:39
but a broadening of our horizons a
00:14:41
little bit more of the science to
00:14:43
balance out the focus that we've had on
00:14:45
data so far
00:15:01
you
