Description
Using Business Intelligence in the Insurance Industry
Using Business Intelligence in the Insurance Industry (Session VAL-3) 1
Proceedings of the Canadian Institute of Actuaries, Vol. XXXVII, No. 2, June 2006
Session VAL-3: Using Business Intelligence in the Insurance Industry
Séance VAL-3 : Utiliser l’intelligence d’affaires dans le secteur des assurances
Moderator/Modérateur: F. Wesley Reynolds
Speakers/Conférenciers: André Giroux
Philippe Torres
(?? = Inaudible/Indecipherable; U-M = Unidenti?ed Male / U-F = Unidenti?ed Female; ph = phonetic)
Moderator F. Wesley Reynolds: Alright. Let’s start again. Welcome to this session entitled Using Business Intelligence in the
Insurance Industry. I am pleased to introduce our presenters, Philippe Torres and André Giroux who are here to discuss the use
of Business Intelligence in insurance and the business organizational issues that come with it. Philippe is President of SQLiaison
and has been working in the Business Intelligence ?eld for 12 years. Philippe has a Bachelor of Computer Science from Con-
cordia University. André works with Philippe and is here to assist with any questions that you may have. SQLiaison has of?ces
in Montreal and Toronto, 60 employees, and clients in the insurance, banking and pharmaceutical industries. These diverse
clients are located across Canada and some are located in the U.S.; Philippe.
Mr. Philippe Torres: First off, thank you very much for attending this session. We are going to talk about Business Intelligence
in the Insurance Industry. Now if I look at the description it says more than ever insurance decision makers need to have easy
and timely access to information. This session will focus on the issues associated with data warehousing technology and provide
examples of effective applications of this technology.
Well, my talk, I guess isn’t too far from that point. We will be talking about BI, Business Intelligence, BI for short, in the insur-
ance industry. The insurance industry is a big industry. There are people on the pension side, the group bene?ts side. I am not
an actuary so I am quite honoured to be amongst you today. I’ll try to live up to the expectation, but the insurance industry
is a rather large industry and applications of Business Intelligence are very wide as well. It would be absolutely impossible to
try to pinpoint speci?c things in your particular practices because BI can be applied as well to Life Insurance, Investments,
Group Bene?ts, Property and Casualty, Reinsurance, Broker Management – there are a lot of applications of BI in the insurance
industry. My talk will be more generic in nature and will focus on issues, not only technology issues but organizational issues as
well in and around deploying Business Intelligence for this industry. The goal really of this session is to give you a few reference
points. I think if you leave the session with more questions than answers we will have achieved our objective. The idea is for
you to come out of this session with things to think about when you think about Business Intelligence (BI) in your particular
companies, corporations.
To give you reference points regarding what is BI and how it applies to the insurance industry; who are the actors involved in
a successful BI program… So, there are different people with different types of responsibilities throughout the enterprise. Who
are these actors, what do they do, what should they be doing? Also, to leave you with some reference points as to how to tri-
angulate your own company’s position within BI; where are you? Every day we do what we do, every day we crunch numbers
or develop reports or whatever we do and I think it’s important for you; you have, as actuaries, a very in?uential role in any BI
program for any type of practices within any branches of the Insurance Industry. So I think it’s important for you to be able to
step out of your day-to-day work and triangulate your own company’s position with respect to BI because more and more BI is
Délibérations de l’Institut canadien des actuaires, Vol. XXXVII, n
o
2, juin 2006
Utiliser l’intelligence d’affaires dans le secteur des assurances (Séance VAL-3) 2
becoming not only an accepted practice but also a strategic differentiator in the marketplace. It’s very competitive and having
Intelligence, being able to develop Business Intelligence about where we stand I think is rather important.
Now, the agenda will be looking at de?nitions of Business Intelligence, of BI, so that we’re all I guess reading from the same
page. Who are the actors; I’ve already talked about that; examples of BI in the Insurance Industry, just going through a couple
of cases to tell you what people have done with this technology; then I want to talk about an Insurance BI roadmap. So what
are the steps, what is the vision that an Insurance Company should perhaps adopt when thinking about a Business Intelligence
program, and then some introspection, recognizing where you stand. This is more of a question session rather than, again, as I
said before, to give you some reference point to see where you stand, and then we’ll follow up with a Q&A. If you have ques-
tions, please feel free to ask them even during the session. André and I can provide you with – André is a business development
manager for SQLiaison. He has a lot of experience in dealing with Insurance Companies and developing BI programs. I think as
a combination, André and I will be able hopefully to answer your questions. Logistics, I have some business cards in the back
if you want to talk some more about that later on at a later date, I would be happy to do that with you.
So, de?nitions – what I did is simply, like what we do today, I have four teenagers between the ages of 14 and 16 at home and
whenever they have a school project or school research, what they do is they go on the internet. If it’s on the internet it must be
true so what I did is I went on the internet and I typed in Business Intelligence and I went on Webopedia site and Webopedia
said that – I won’t read the whole thing for you but – the term BI represents tools and systems and they play a key role in stra-
tegic planning processes, allow you to store, access and analyze corporate data to aid in decision-making and it goes on to talk
about different examples of customer pro?ling, customer support, market research, statistical analysis and on and on. That’s
one de?nition. Then I went to another source on the internet, again if the internet says it then it must be true. Wikipedia says
that Business Intelligence typically refers to a set of business processes. I said, that’s interesting. Immediately we’re digressing
from technology per se and we’re talking about business processes for collecting and analyzing business information. It does
include a technology component of course. Then it goes on to say that competitive organizations accumulate BI, they accumu-
late Business Intelligence in order to gain a sustainable competitive advantage. Another interesting point here and it certainly
is viewed as a valuable core competence within the company; that’s a de?nition. Now, I went on our own website and found
a more even – I mean, we claim to be Business Intelligence experts; that’s what we do, but it’s a very convoluted de?nition to
me. We talk here about the primary goal of BI is to provide accurate, complete and integrated information, another key term,
“integrated information”, so thinking about information holistically rather than in a silo fashion or looking at only one perspec-
tive of information, so integrated information. So lets you gain visibility, so look at things that weren’t exposed before, perhaps
to pinpoint inef?ciencies. Then we talk about the identi?cation of new opportunities, perhaps it’s new products, perhaps it’s
new price points to deliver maximum value throughout your value chain.
We’ve looked at three de?nitions. It seems to me they’re fairly complicated and convoluted, even our own website. Now Busi-
ness Intelligence really can be de?ned within three perspectives. There is certainly a business perspective to it, so what’s this BI
thing for the business? That’s one aspect. Certainly a technology and infrastructure perspective and traditionally that’s how BI has
been thought of, as a technology and infrastructure kind of thing rather than a beast with three heads, with three perspectives.
A very important aspect is the people and process because the accumulation of Business Intelligence is to serve the purposes of
people ultimately and when people have to do anything within a large enterprise, then we need to be talking about processes. A
very important and often not very well thought about aspect of BI is the people and process perspective. So we’ll be exploring
those three perspectives in the following slides.
From a business perspective really when we talk about BI, you know in any enterprise one of the key things that we do as
managers or as top management of any company, we are looking at the development of business strategies. So we’re in a given
market, maybe it’s P&C maybe it’s Life, Investment, whatever market we’re in, so we’re looking at developing business strategies
to obviously gain competitive advantage, to gain a position, to gain a market, whatever the case may be. We certainly want to
look at markets, products, distribution channels, broker channels, direct channels, and some ?nancial objectives, some of the
customers you want to go after speci?cally. All of that is really the development of the business strategies. When we develop the
business strategies we put them into place with tactical things, action plans that we adopt, and obviously we need information.
Information is the life-blood of any Insurance Company.
Using Business Intelligence in the Insurance Industry (Session VAL-3) 3
Proceedings of the Canadian Institute of Actuaries, Vol. XXXVII, No. 2, June 2006
No matter what industry you’re in, information is your life-blood. Again, in order to support that, we need an architecture
component, a Business Intelligence component.
We need to be talking about process, how to distribute the information, how to develop the ability to answer ad hoc queries that
the business may have. Once we master that, what happens then is that we’re in an environment where we can start to manage
insight and our business knowledge. We have the ability now to be looking at trends, to see how we’re digressing from compli-
ance standards. We have the ability to experiment with the data, perhaps mine the data, so automated knowledge discovery.
Ultimately that information can help us in redeveloping or readjusting or readapting our business strategies. It really means that
now with the history of what happened, the trending, we’re able to create new products and develop new business strategies. It
is really a cycle of development of business strategies, the support mechanism and then the development of insight, knowledge
that allows us to circle back into developing new business strategies. From a business perspective BI really should be an enabler
that allows us to do this, very simply put. We want to use information that allows us to do this very cycle.
From a technology infrastructure perspective, I’m not going to try to be too technical here so, like any good IT (Information
Technology) chart, I have squares and bubbles and triangles. It’s all basic shapes that we can all recognize. From a technology
perspective how we de?ne BI really is through what’s called a Corporate Information Factory. A Corporate Information Factory
is simply a set of tools that allows us to take data from the source systems. You have source systems whether you’re in Life,
Investment, Pension, P&C. You have source systems that will manage the day-to-day operations of the business. An example
in the P&C world might be a policy administration system or a claims management system. Those systems are a source of
data. This is where your data lives. This is the data that you want to go after to be able to derive new information from that
data. Traditionally what happens in a CIF, in a Corporate Information Factory, that architecture blueprint, without being too
technical, has been developed over the last 20 years, so this is a known quantity, this is how we do this stuff. You will recognize
some components. You certainly have heard of the term “Data Warehouse”. How many of you have heard of the term “Data
Warehouse”? The company has a Data Warehouse so that’s where we store the corporate data at a transaction level. We are able
to look at all the transactions over time and develop trending or whatever. There are a few other components which are more or
less important in this image, but the next image is what we call, or the next stage of the Corporate Information Factory would
be, the data delivery step. It’s one thing to store the data, right, to store the raw data; it’s yet another to deliver it in a format that
is suitable for the ultimate user.
Now, as actuaries you have a particular requirement for the data and it’s not very complicated. Every time I have talked to an
actuary they tell me “I want everything”. I mean that’s ?ne. It’s okay. You want everything because you need to be able to have
access to all the data for your pricing models or whatever the case may be. In a standard Business Intelligence architecture that
data delivery happens in what are called “Data Marts”. So Data Marts are collections of databases that serve a particular purpose.
One example of a Data Mart is the famous actuarial database. Well, the actuarial database contains everything and this is where
you will go out and probably code a bunch of SAS programs to get access to the data. That’s one example. Another example
might be a Data Mart for Key Performance Indicators. Management wants to have a very nice dashboard to see how we’re doing,
our number of submissions, what’s our loss ratio, what’s our hit ratio and, you know, nice little dials and levers to see how the
company is doing. That data might come from another type of Data Mart, maybe it’s a KPI Data Mart. Or, as a third example
you might have some marketing people who want to combine data from StatsCan or some socio-demographic data, whatever
the case may be, that might be the third Data Mart. So it’s important from a technology perspective that the environment has the
ability to create these Data Marts fairly easily that may live or die after a certain amount of time. It doesn’t really matter because
all the data is in the data warehouse so we can recreate those Data Marts at will.
Then the ?nal piece, if you will, or next to ?nal piece of the puzzle, is reporting and analytics. Those are the set of tools that
allow you to get access to the data, you as human beings to access that data. In an ideal world access to that data is point and
click. You go click, click, click, I want to see this, I want to see that, I want to combine it with this, and off you go. The system
generates a procedure for you so that automatically it knows where the data is and retrieves the data and then you can do what-
ever you want with it, whether it’s put it in an Excel spreadsheet, create a report, whatever the case may be. Now it brings up
a point – I’ll be a little controversial about this. – I’ve been working in this industry for the last 12 years and for about 10 years
speci?cally with insurance companies and it is the case that a lot of actuarial folks think about a procedure to get access to the
Délibérations de l’Institut canadien des actuaires, Vol. XXXVII, n
o
2, juin 2006
Utiliser l’intelligence d’affaires dans le secteur des assurances (Séance VAL-3) 4
data, procedural thinking. I’m going to open a data set; I’m going to scan it from top to bottom and I’m going to summarize; I’m
going to do this and I’m going to do that. One should realize that in today’s environments, today’s environments are very declara-
tive in nature. We tell the system what we want and then the system generates the procedure for us and retrieves the data. It’s
a little of a leap of faith, I guess, in trusting that the system can ?nd the right procedure for us to create that information. I just
wanted to mention in passing that technologies are there to do that today so we need not be thinking about it in a procedural
manner; we can start thinking in a declarative manner where the system generates the procedure for us. What allows the system
to do this is what is called metadata. Metadata is data about data. It says my policies are here and my claims are over here and
my broker is over there, my products are over here; another very important component. From a technology perspective this is
how we can de?ne Business Intelligence.
Enough technicalities -- from a people and process perspective, in an ideal world this is what would happen. Business Users talk
to Information Managers, who interact with Information Technology people who are responsible for managing this architecture
that we just called the Corporate Information Factory. This is in an ideal world. Not all organizations have this level of maturity
in terms of their BI programs but this is where you should tend to strive for, to make BI work as a process. Now, it leads me
to talk about who are the actors in Business Intelligence. There are three types of actors. We saw Business Users, we saw the
management piece and we saw the IT (Information Technology) piece. If you are a Business User – and you may very well be a
Business User yourself – you certainly want to use information to support your business process or business strategy. You might
be someone from Marketing; you might be from Underwriting; you might be from Claims; you might be from Actuarial; you
might be a Business User. You are an information consumer. You consume information and ideally you should not care how this
information is stored, how it’s accessed, how it’s managed. You just want information to work with. You are a Business User.
One of the challenges that Business Users typically see is that gleaning information from data is dif?cult. So things like how
are my renewal rates doing, where can I steer the broker channel. The analysis of detail is also very dif?cult and requires many
cycles of effort. There are also inconsistencies of information across the enterprise, so you get to a meeting and then you have
your own set of reports and Joe Smith over here has his own set of reports and Jane Smith over there has her own set of reports
and, lo and behold, the numbers do not match. So there are inconsistencies of information across the enterprise.
The other type of actor or the other actor in this play is members of the Management Information Systems team. Well, a typi-
cal MIS (Management Information Systems) team member is someone who is responsible to provide information based on the
company data to the Business User. They respond to information demands from the Business Users. These are not necessarily
technical experts. They don’t necessarily know technology, but they certainly know the business very well.
In the case of actuarial it might not be surprising that you are both a Business User and an MIS user because you are typically
Power Users or information producers. I am making a difference here between an information consumer and an information
producer. You may be both, granted, but there is a signi?cant difference between the two. One is responsible for producing
information, interpretation about information. On the other hand, an information consumer just consumes it. Now, some of the
typical challenges. The systems that are in place do not allow you a quick turnaround in information demands so it’s dif?cult to
produce information. Mapping the business requirements may be dif?cult because the data is hard to understand and you are
spending 80% of your time assembling data rather than analyzing data. Let’s take an Easytrieve over here an Excel over here, a
SAS extract over here, it’s assembly. A lot of time is spent on assembly of the data. We might not have any IT people in the room
but if you are a member of the IT team you have a strong technical expertise, you understand the company’s data structures,
the architecture of support systems; those sort of things.
I’d be surprised if anyone in the room is an actual IT team member, but nevertheless an important actor in this whole BI picture.
Some of the challenges are more related to technology and very often what happens is that the technology or the infrastructure
really does not allow the IT team to be very responsive to the MIS team and in turn to the Business Users.
Do you recognize yourself in this Shakespearean play? Are you a Business User? Are you an MIS user? A good question is, is there
such a group in your organizations? Maybe not -- the information producers may be scattered all over. You might have some in
Actuarial; you might have some in Underwriting; you might have some in Claims. Who has not heard of Natalie Johnson over
here who knows how to get things done in terms of producing reports but she’s not in the MIS group and there is no such thing
Using Business Intelligence in the Insurance Industry (Session VAL-3) 5
Proceedings of the Canadian Institute of Actuaries, Vol. XXXVII, No. 2, June 2006
in the enterprise? It might very well be a distributed function; nevertheless an important function and we have to recognize
that it’s there. Are you a Power User? I suspect that most people who have gone through the ranks of the actuarial discipline in
all of your respective companies have at one point or another been labeled as a Power User. As an information producer you
produce information, there’s no two ways about that. Are you an IT member or do you sometimes feel that you act like one, so
you are spending a lot of time learning about Oracle, databases and DB2 databases and relational technology and stuff that has
nothing to do with the actuarial science and actuarial discipline, perhaps.
Moving on to some of the examples of BI in the insurance industry; Business Intelligence Jeopardy – I wanted to have the Jeop-
ardy team going at the same time here – the question is which industry segment has the most interest in deploying Business
Intelligence. It’s a good question. Well, we have different practices: Life, Investment, Property and Casualty, Group Bene?ts,
Reinsurance Brokers, and there’s a whole lot more. Without answering the question directly, let me say this: Business Intelligence
is a discipline. Historically it came from retail.
You probably have heard the Wal-Mart story where there’s a correlation between selling beer and diapers. Anybody heard that
story before? No, nobody has heard it. Well, I’ll tell it; it’s a classic story. Apparently when diapers are on sale retailers sell more
beer and they found out using a data warehouse and using Business Intelligence apparently the explanation for this is that when
diapers are on sale mommy sends daddy to the corner store to buy diapers and by the way he picks up a six pack of beer at the
same time. I’m not making this up; this is very well documented in the history of Business Intelligence.
The point I’m making is in the retail industry – very, very, competitive – there’s a lot of rotations of movement of products and all
that sort of thing, so it’s an industry with a very, very quick dynamic of new products, of very, very high churning of customers.
Customers come in and out, in and out, in and out. There is absolutely no brand loyalty; absolutely none. Business Intelligence
helps the retail industry because it allows them again to develop business strategies and the stuff that we talked about at the
beginning of our presentation, but there’s a clear correlation between a high dynamic market and the use of Business Intelligence;
a very, very huge correlation. So the more dynamics you have in your market the more important or the better use you can make
of Business Intelligence. If your market is a very, very slow market and it takes years to develop and people come in but they
never leave, such as life insurance, term products and all that, the market dynamics are much slower.
The Business Intelligence, it has been our experience and the industry’s experience that it is probably the P&C market that can
bene?t the most from Business Intelligence because of the market dynamics of P&C. You know, I have car insurance today, I
can cancel it and get another one next year after 12 months; a very, very high dynamic. So the survey says really that P&C is
probably the market that can bene?t the most from Business Intelligence. I’m not saying that the other markets can’t bene?t
from it but the high dynamics make it very suitable for that type of business environment.
Now, business cases for why do insurance companies invest in BI; business cases vary and they can vary widely. It all has to do
with what is the strategy that the company is adopting. Do we want to go after new clients? Do we want to increase the pro?t-
ability of underwriting? What is the incentive? So they may have to do a reduction of operational risks, speed of market response.
Some companies want to deploy BI to foster better relationships with their brokers, so broker relationship management, or even
the direct writers with client relationship management, or exploiting opportunities for expense reduction. So, a question that
I have for you, and I’m not expecting an answer, is can you think of more? There are dozens and dozens of business cases that
can be developed for BI in an insurance company. An assignment when you leave the room is to think of a business case for BI
or a few business cases of BI in your respective organizations.
A few examples, broker management – broker management is a subject area that supports the analytical requirements for the
enterprise to be able to analyze and report on brokers. In other words, looking at how the broker is doing is an area unto itself.
A lot of insurance companies have deployed BI for better broker management. Their rationale is if I’m going to be dealing with
brokers I want an information-based relationship with my broker. I want to be able to go to my broker and say the piece of
business or the agent, the piece of business that you’ve written here is not so good because you used the experience. In other
words, drive the relationship using information. That’s an example. Another example is simply reg ?lings. People have been
using a BI infrastructure for regulatory ?lings to the IBC for instance in the case of P&C. Bene?ts include streamlined electronic
?lings and reduction of errors in ?lings because they’re using a BI infrastructure.
Délibérations de l’Institut canadien des actuaires, Vol. XXXVII, n
o
2, juin 2006
Utiliser l’intelligence d’affaires dans le secteur des assurances (Séance VAL-3) 6
Another example, risk attributes variance analysis, the ability to track over time the set of risks or the set of risk attributes
and how they affect the experience. You know, what if we’re rating on this, now we’re not rating on this and we are rating on
this other attribute? The bene?ts include the ability to track the evolution of a risk pro?le over time. Another example, a ?ne
example, is on-level analysis. This is actually an actuarial requirement for developing what are called a book of tariffs. So how
can we use historical information to drive our pricing models?
Here is an example, a company that shall remain unnamed that had an amalgam of operational support systems, a lot of support
systems, and very, very dif?cult to do the infamous month-end process. Every time at the end of the month those reports come
out and it’s very dif?cult. Results come out X number of days after the month-end and certainly not an environment that allowed
them to do any kind of analytics to dig into the data. What they did is they implemented a BI infrastructure for consolidating
and reporting on premiums and sales. Now the turnaround time is much quicker. They went from I would say 4-5 days after
month end to less than 24 hours after month end; 2 or 3 days faster turnaround time on month end information. They are much
closer to the famed single version of the truth; that is, the ability to get consistent reporting across the enterprise.
Another example of a company that shall also remain unnamed is a company that has adopted BI to reduce the operational
risk. This company, a P&C company – I’ll just give you some statistics – wanted to introduce a new driver class in Ontario and
it was estimated that it would take perhaps 75 person days to implement the new product and the support systems, the policy
management, the policy admin system. It would have taken about 500 days for the decision support system, so a high risk.
What they did is they implemented a BI infrastructure that allows them to simplify the reporting infrastructure; certainly a lower
risk from an operational point of view. Actuarial now have their own sandbox in which they can play with the data at their own
will. Again, the ability to implement multiple Data Marts, multiple sandboxes for different purposes.
I want to shift gears a little bit and talk about a roadmap for BI in insurance. What I propose here is a roadmap that is really
a set of milestones that an insurance company should be looking at when thinking about a BI program for the corporation.
It’s derived from experience. It’s derived from our own internal R&D and what’s happening in the market. It’s fairly simple to
understand. The roadmap really proposes four key milestones. The ?rst milestone is integrating and warehousing the data. The
second milestone – and we’ll go through these – the second milestone is to analyze and produce consumer information. The
third one is to model and simulate the operational transactions and ?nally it’s the management of business knowledge. Let’s
look at each of these steps individually.
The ?rst step is really the integration and the warehousing of data. In any BI program, if you think about the future, this is the
?rst thing that needs to happen. You need to have an environment whereby your enterprise data is integrated and warehoused
in a central location. If that’s not there – this is a question for introspection – then the future for BI is shaky at best because the
foundation will not be there. So you need that foundation of the ability to integrate your operational data into a central loca-
tion.
Step number two, once you’ve done that, once you have the ability to store data in a central location, is really get into a proc-
ess, and that involves again tools and technology but also involves people and processes to analyze, produce and consume
information. I would say that a lot of the companies that have implemented Business Intelligence are struggling with this step
of the roadmap because they have an infrastructure in place but they don’t necessarily have the people and processes in place
to really bene?t from the infrastructure. The good news is that step one and step two, really those two ?rst critical steps, are
supported with existing technologies, what we talked about at the beginning of the presentation. We talked about the Corpo-
rate Information Factory where you have your enterprise at a warehouse, where you have your data delivery, where you have
your reporting infrastructure and where you have your metadata management. Those two steps are well supported with this
Corporate Information Factory model. This I would say is going to happen in the next couple of years for most companies that
have implemented the BI program. It’s really the ability to model and simulate transactions. Really what we’re talking about here
is the ability not only to question the history because that’s one thing, question the history and look at trends and compliance
and things of that nature.
The other question is really the ability to ask the “what if” questions. You know, what if what happened in the past didn’t really
happen like that and it happened in another way. Or, what if we would price our products like this instead of like that, so the
Using Business Intelligence in the Insurance Industry (Session VAL-3) 7
Proceedings of the Canadian Institute of Actuaries, Vol. XXXVII, No. 2, June 2006
ability to ask “what if” is really the next step in the Business Intelligence roadmap. An example of this, I thought about on-level
analytics or the ability to generate an on-level premium. That’s a very good example of a “what if” scenario. So what if my rating
assumptions are varied or what if I have a new set of rating assumptions, what’s my experience? That is, it has nothing to do
with the past, it has everything to do with the future and it probably has to do with pricing, product development and such
things. That’s an example of modeling your transactions, the ability to ask the “what if” questions.
The ?nal step of the roadmap is really I guess the top of the pyramid, is really the management and the re?nement of business
knowledge. When we think about our strategic step, the ?rst step is the development of our strategic direction. A lot of it is
based on assumptions that we make, or we pose a certain set of hypotheses but they’re not necessarily always veri?ed. In this
step of the roadmap you’re in a place where you can start asking questions like this. You can, for instance, think about a rule
that says when the price of gas increases the loss ratio decreases, right, so your experience is better. Well, that clearly relies on
external data. The price of gas is not something that you have in your own data. The price of gas is something that you would go
out and get from some external data provider. It is certainly based on correlation operators, you know, an increase, a decrease, a
correlation with perhaps internal data that you may have. The loss ratio or your experience is your own, so that’s internal data.
So you can imagine, if you will, an environment where you have a set of rules, you have a rulebook that you can simply test
against your data. That’s called directed data mining. You want to mine your data to verify certain hypotheses, certain assump-
tions that you’re making. It sounds far-fetched but really it’s not that far away. It is really at the crossroads of existing technology
such as data warehousing, business rules, management, data provision, external data provision, BI, data mining, CPM (Corporate
Performance Management). All these things are coalescing really to create an environment where now we can start managing
our own hypotheses, managing our own rules by which we make strategic investments or strategic decisions.
The last piece of this talk, I want to talk about a bit of introspection, recognizing where you stand. As I said before, actuaries
have a strong in?uence over BI products for the enterprise because you are the main information producers of the enterprise.
Actuaries understand data and its derivation. You understand the business rules and how to manipulate the data. You understand
the structure of the data. Now, BI, however, is more than the actuarial database. That’s one thing I want to leave you with that BI
is simply more than the actuarial database. It’s more than procedural programming with sequential data sets. I talked about that
before. I think it’s important to think about how to change our paradigm from a procedural paradigm to a declarative paradigm.
Certainly, BI is more than impossibly large reports containing many metrics. Believe it or not, a lot of other consumers in the
enterprise are really not interested in the stuff that actuaries are interested in, in terms of their numbers or detailed numbers,
all they want is their premiums and their losses and that’s about it.
One observation, I went to a website of a highly reputable actuarial school in Canada, I won’t say where, but I was, well, I
shouldn’t say that I was surprised, by looking at the basic curriculum of actuarial science there is only one course related to the
use of computers. This one course is a course on how to use Microsoft Word and how to use Microsoft Excel, that kind of thing.
That’s a mandatory course. Then there is an elective course which is a Java programming course. This is not a very exhaustive
study on my part but as I was preparing for this talk I said let me go and look at the curriculum and I had lunch with a client of
ours in Montreal last week and he concurred. The actuarial discipline today, there is a lot of I guess data-related work with the
actuarial discipline and I found it very surprising that in these university curriculums there’s not a whole lot given to structured
programming and data modeling and information management, those sort of things. Yet, if you look at most actuarial depart-
ments today what actuaries do, a lot of it has to do with data. I was very surprised to ?nd how little focus there is on the things
I just talked about, data, structures and programming in the actuarial schools. Again, you know, we can’t change the world but
it might be interesting in the future if more attention was given to using more computer science related topics in an actuarial
program. So that’s something that I’d like you to think about.
From the business perspective, do you rely on your data to make strategic decisions? Would you be able to answer any ques-
tions that come your way regarding any line of business? Can you drive dynamic broker relationships based on information? Do
you view your decision support systems as an operational risk? Perhaps the answer is yes; perhaps the answer is no, you want
to think about that. From a technology infrastructure perspective, do you have a single integrated source of decision support
data? Do you have an auditable data integration process? Do you have metadata? Those are all questions to think about from a
technology infrastructure perspective.
Délibérations de l’Institut canadien des actuaires, Vol. XXXVII, n
o
2, juin 2006
Utiliser l’intelligence d’affaires dans le secteur des assurances (Séance VAL-3) 8
Finally, on the people and process perspectives, do you consider that any particular group within your company owns the data?
Who owns the data? Is there a centralized multi-disciplinary group responsible for information production? Do you have an
MIS component? Do you have an executive-level steering committee driving BI in your corporation? Finally, is BI considered
an IT project or a business project? These are all good questions that I want to leave you with and then I’d like to open it up
for questions.
Questions, comments, issues? That’s it, then. Thank you very much.
(applause)
doc_465126308.pdf
Using Business Intelligence in the Insurance Industry
Using Business Intelligence in the Insurance Industry (Session VAL-3) 1
Proceedings of the Canadian Institute of Actuaries, Vol. XXXVII, No. 2, June 2006
Session VAL-3: Using Business Intelligence in the Insurance Industry
Séance VAL-3 : Utiliser l’intelligence d’affaires dans le secteur des assurances
Moderator/Modérateur: F. Wesley Reynolds
Speakers/Conférenciers: André Giroux
Philippe Torres
(?? = Inaudible/Indecipherable; U-M = Unidenti?ed Male / U-F = Unidenti?ed Female; ph = phonetic)
Moderator F. Wesley Reynolds: Alright. Let’s start again. Welcome to this session entitled Using Business Intelligence in the
Insurance Industry. I am pleased to introduce our presenters, Philippe Torres and André Giroux who are here to discuss the use
of Business Intelligence in insurance and the business organizational issues that come with it. Philippe is President of SQLiaison
and has been working in the Business Intelligence ?eld for 12 years. Philippe has a Bachelor of Computer Science from Con-
cordia University. André works with Philippe and is here to assist with any questions that you may have. SQLiaison has of?ces
in Montreal and Toronto, 60 employees, and clients in the insurance, banking and pharmaceutical industries. These diverse
clients are located across Canada and some are located in the U.S.; Philippe.
Mr. Philippe Torres: First off, thank you very much for attending this session. We are going to talk about Business Intelligence
in the Insurance Industry. Now if I look at the description it says more than ever insurance decision makers need to have easy
and timely access to information. This session will focus on the issues associated with data warehousing technology and provide
examples of effective applications of this technology.
Well, my talk, I guess isn’t too far from that point. We will be talking about BI, Business Intelligence, BI for short, in the insur-
ance industry. The insurance industry is a big industry. There are people on the pension side, the group bene?ts side. I am not
an actuary so I am quite honoured to be amongst you today. I’ll try to live up to the expectation, but the insurance industry
is a rather large industry and applications of Business Intelligence are very wide as well. It would be absolutely impossible to
try to pinpoint speci?c things in your particular practices because BI can be applied as well to Life Insurance, Investments,
Group Bene?ts, Property and Casualty, Reinsurance, Broker Management – there are a lot of applications of BI in the insurance
industry. My talk will be more generic in nature and will focus on issues, not only technology issues but organizational issues as
well in and around deploying Business Intelligence for this industry. The goal really of this session is to give you a few reference
points. I think if you leave the session with more questions than answers we will have achieved our objective. The idea is for
you to come out of this session with things to think about when you think about Business Intelligence (BI) in your particular
companies, corporations.
To give you reference points regarding what is BI and how it applies to the insurance industry; who are the actors involved in
a successful BI program… So, there are different people with different types of responsibilities throughout the enterprise. Who
are these actors, what do they do, what should they be doing? Also, to leave you with some reference points as to how to tri-
angulate your own company’s position within BI; where are you? Every day we do what we do, every day we crunch numbers
or develop reports or whatever we do and I think it’s important for you; you have, as actuaries, a very in?uential role in any BI
program for any type of practices within any branches of the Insurance Industry. So I think it’s important for you to be able to
step out of your day-to-day work and triangulate your own company’s position with respect to BI because more and more BI is
Délibérations de l’Institut canadien des actuaires, Vol. XXXVII, n
o
2, juin 2006
Utiliser l’intelligence d’affaires dans le secteur des assurances (Séance VAL-3) 2
becoming not only an accepted practice but also a strategic differentiator in the marketplace. It’s very competitive and having
Intelligence, being able to develop Business Intelligence about where we stand I think is rather important.
Now, the agenda will be looking at de?nitions of Business Intelligence, of BI, so that we’re all I guess reading from the same
page. Who are the actors; I’ve already talked about that; examples of BI in the Insurance Industry, just going through a couple
of cases to tell you what people have done with this technology; then I want to talk about an Insurance BI roadmap. So what
are the steps, what is the vision that an Insurance Company should perhaps adopt when thinking about a Business Intelligence
program, and then some introspection, recognizing where you stand. This is more of a question session rather than, again, as I
said before, to give you some reference point to see where you stand, and then we’ll follow up with a Q&A. If you have ques-
tions, please feel free to ask them even during the session. André and I can provide you with – André is a business development
manager for SQLiaison. He has a lot of experience in dealing with Insurance Companies and developing BI programs. I think as
a combination, André and I will be able hopefully to answer your questions. Logistics, I have some business cards in the back
if you want to talk some more about that later on at a later date, I would be happy to do that with you.
So, de?nitions – what I did is simply, like what we do today, I have four teenagers between the ages of 14 and 16 at home and
whenever they have a school project or school research, what they do is they go on the internet. If it’s on the internet it must be
true so what I did is I went on the internet and I typed in Business Intelligence and I went on Webopedia site and Webopedia
said that – I won’t read the whole thing for you but – the term BI represents tools and systems and they play a key role in stra-
tegic planning processes, allow you to store, access and analyze corporate data to aid in decision-making and it goes on to talk
about different examples of customer pro?ling, customer support, market research, statistical analysis and on and on. That’s
one de?nition. Then I went to another source on the internet, again if the internet says it then it must be true. Wikipedia says
that Business Intelligence typically refers to a set of business processes. I said, that’s interesting. Immediately we’re digressing
from technology per se and we’re talking about business processes for collecting and analyzing business information. It does
include a technology component of course. Then it goes on to say that competitive organizations accumulate BI, they accumu-
late Business Intelligence in order to gain a sustainable competitive advantage. Another interesting point here and it certainly
is viewed as a valuable core competence within the company; that’s a de?nition. Now, I went on our own website and found
a more even – I mean, we claim to be Business Intelligence experts; that’s what we do, but it’s a very convoluted de?nition to
me. We talk here about the primary goal of BI is to provide accurate, complete and integrated information, another key term,
“integrated information”, so thinking about information holistically rather than in a silo fashion or looking at only one perspec-
tive of information, so integrated information. So lets you gain visibility, so look at things that weren’t exposed before, perhaps
to pinpoint inef?ciencies. Then we talk about the identi?cation of new opportunities, perhaps it’s new products, perhaps it’s
new price points to deliver maximum value throughout your value chain.
We’ve looked at three de?nitions. It seems to me they’re fairly complicated and convoluted, even our own website. Now Busi-
ness Intelligence really can be de?ned within three perspectives. There is certainly a business perspective to it, so what’s this BI
thing for the business? That’s one aspect. Certainly a technology and infrastructure perspective and traditionally that’s how BI has
been thought of, as a technology and infrastructure kind of thing rather than a beast with three heads, with three perspectives.
A very important aspect is the people and process because the accumulation of Business Intelligence is to serve the purposes of
people ultimately and when people have to do anything within a large enterprise, then we need to be talking about processes. A
very important and often not very well thought about aspect of BI is the people and process perspective. So we’ll be exploring
those three perspectives in the following slides.
From a business perspective really when we talk about BI, you know in any enterprise one of the key things that we do as
managers or as top management of any company, we are looking at the development of business strategies. So we’re in a given
market, maybe it’s P&C maybe it’s Life, Investment, whatever market we’re in, so we’re looking at developing business strategies
to obviously gain competitive advantage, to gain a position, to gain a market, whatever the case may be. We certainly want to
look at markets, products, distribution channels, broker channels, direct channels, and some ?nancial objectives, some of the
customers you want to go after speci?cally. All of that is really the development of the business strategies. When we develop the
business strategies we put them into place with tactical things, action plans that we adopt, and obviously we need information.
Information is the life-blood of any Insurance Company.
Using Business Intelligence in the Insurance Industry (Session VAL-3) 3
Proceedings of the Canadian Institute of Actuaries, Vol. XXXVII, No. 2, June 2006
No matter what industry you’re in, information is your life-blood. Again, in order to support that, we need an architecture
component, a Business Intelligence component.
We need to be talking about process, how to distribute the information, how to develop the ability to answer ad hoc queries that
the business may have. Once we master that, what happens then is that we’re in an environment where we can start to manage
insight and our business knowledge. We have the ability now to be looking at trends, to see how we’re digressing from compli-
ance standards. We have the ability to experiment with the data, perhaps mine the data, so automated knowledge discovery.
Ultimately that information can help us in redeveloping or readjusting or readapting our business strategies. It really means that
now with the history of what happened, the trending, we’re able to create new products and develop new business strategies. It
is really a cycle of development of business strategies, the support mechanism and then the development of insight, knowledge
that allows us to circle back into developing new business strategies. From a business perspective BI really should be an enabler
that allows us to do this, very simply put. We want to use information that allows us to do this very cycle.
From a technology infrastructure perspective, I’m not going to try to be too technical here so, like any good IT (Information
Technology) chart, I have squares and bubbles and triangles. It’s all basic shapes that we can all recognize. From a technology
perspective how we de?ne BI really is through what’s called a Corporate Information Factory. A Corporate Information Factory
is simply a set of tools that allows us to take data from the source systems. You have source systems whether you’re in Life,
Investment, Pension, P&C. You have source systems that will manage the day-to-day operations of the business. An example
in the P&C world might be a policy administration system or a claims management system. Those systems are a source of
data. This is where your data lives. This is the data that you want to go after to be able to derive new information from that
data. Traditionally what happens in a CIF, in a Corporate Information Factory, that architecture blueprint, without being too
technical, has been developed over the last 20 years, so this is a known quantity, this is how we do this stuff. You will recognize
some components. You certainly have heard of the term “Data Warehouse”. How many of you have heard of the term “Data
Warehouse”? The company has a Data Warehouse so that’s where we store the corporate data at a transaction level. We are able
to look at all the transactions over time and develop trending or whatever. There are a few other components which are more or
less important in this image, but the next image is what we call, or the next stage of the Corporate Information Factory would
be, the data delivery step. It’s one thing to store the data, right, to store the raw data; it’s yet another to deliver it in a format that
is suitable for the ultimate user.
Now, as actuaries you have a particular requirement for the data and it’s not very complicated. Every time I have talked to an
actuary they tell me “I want everything”. I mean that’s ?ne. It’s okay. You want everything because you need to be able to have
access to all the data for your pricing models or whatever the case may be. In a standard Business Intelligence architecture that
data delivery happens in what are called “Data Marts”. So Data Marts are collections of databases that serve a particular purpose.
One example of a Data Mart is the famous actuarial database. Well, the actuarial database contains everything and this is where
you will go out and probably code a bunch of SAS programs to get access to the data. That’s one example. Another example
might be a Data Mart for Key Performance Indicators. Management wants to have a very nice dashboard to see how we’re doing,
our number of submissions, what’s our loss ratio, what’s our hit ratio and, you know, nice little dials and levers to see how the
company is doing. That data might come from another type of Data Mart, maybe it’s a KPI Data Mart. Or, as a third example
you might have some marketing people who want to combine data from StatsCan or some socio-demographic data, whatever
the case may be, that might be the third Data Mart. So it’s important from a technology perspective that the environment has the
ability to create these Data Marts fairly easily that may live or die after a certain amount of time. It doesn’t really matter because
all the data is in the data warehouse so we can recreate those Data Marts at will.
Then the ?nal piece, if you will, or next to ?nal piece of the puzzle, is reporting and analytics. Those are the set of tools that
allow you to get access to the data, you as human beings to access that data. In an ideal world access to that data is point and
click. You go click, click, click, I want to see this, I want to see that, I want to combine it with this, and off you go. The system
generates a procedure for you so that automatically it knows where the data is and retrieves the data and then you can do what-
ever you want with it, whether it’s put it in an Excel spreadsheet, create a report, whatever the case may be. Now it brings up
a point – I’ll be a little controversial about this. – I’ve been working in this industry for the last 12 years and for about 10 years
speci?cally with insurance companies and it is the case that a lot of actuarial folks think about a procedure to get access to the
Délibérations de l’Institut canadien des actuaires, Vol. XXXVII, n
o
2, juin 2006
Utiliser l’intelligence d’affaires dans le secteur des assurances (Séance VAL-3) 4
data, procedural thinking. I’m going to open a data set; I’m going to scan it from top to bottom and I’m going to summarize; I’m
going to do this and I’m going to do that. One should realize that in today’s environments, today’s environments are very declara-
tive in nature. We tell the system what we want and then the system generates the procedure for us and retrieves the data. It’s
a little of a leap of faith, I guess, in trusting that the system can ?nd the right procedure for us to create that information. I just
wanted to mention in passing that technologies are there to do that today so we need not be thinking about it in a procedural
manner; we can start thinking in a declarative manner where the system generates the procedure for us. What allows the system
to do this is what is called metadata. Metadata is data about data. It says my policies are here and my claims are over here and
my broker is over there, my products are over here; another very important component. From a technology perspective this is
how we can de?ne Business Intelligence.
Enough technicalities -- from a people and process perspective, in an ideal world this is what would happen. Business Users talk
to Information Managers, who interact with Information Technology people who are responsible for managing this architecture
that we just called the Corporate Information Factory. This is in an ideal world. Not all organizations have this level of maturity
in terms of their BI programs but this is where you should tend to strive for, to make BI work as a process. Now, it leads me
to talk about who are the actors in Business Intelligence. There are three types of actors. We saw Business Users, we saw the
management piece and we saw the IT (Information Technology) piece. If you are a Business User – and you may very well be a
Business User yourself – you certainly want to use information to support your business process or business strategy. You might
be someone from Marketing; you might be from Underwriting; you might be from Claims; you might be from Actuarial; you
might be a Business User. You are an information consumer. You consume information and ideally you should not care how this
information is stored, how it’s accessed, how it’s managed. You just want information to work with. You are a Business User.
One of the challenges that Business Users typically see is that gleaning information from data is dif?cult. So things like how
are my renewal rates doing, where can I steer the broker channel. The analysis of detail is also very dif?cult and requires many
cycles of effort. There are also inconsistencies of information across the enterprise, so you get to a meeting and then you have
your own set of reports and Joe Smith over here has his own set of reports and Jane Smith over there has her own set of reports
and, lo and behold, the numbers do not match. So there are inconsistencies of information across the enterprise.
The other type of actor or the other actor in this play is members of the Management Information Systems team. Well, a typi-
cal MIS (Management Information Systems) team member is someone who is responsible to provide information based on the
company data to the Business User. They respond to information demands from the Business Users. These are not necessarily
technical experts. They don’t necessarily know technology, but they certainly know the business very well.
In the case of actuarial it might not be surprising that you are both a Business User and an MIS user because you are typically
Power Users or information producers. I am making a difference here between an information consumer and an information
producer. You may be both, granted, but there is a signi?cant difference between the two. One is responsible for producing
information, interpretation about information. On the other hand, an information consumer just consumes it. Now, some of the
typical challenges. The systems that are in place do not allow you a quick turnaround in information demands so it’s dif?cult to
produce information. Mapping the business requirements may be dif?cult because the data is hard to understand and you are
spending 80% of your time assembling data rather than analyzing data. Let’s take an Easytrieve over here an Excel over here, a
SAS extract over here, it’s assembly. A lot of time is spent on assembly of the data. We might not have any IT people in the room
but if you are a member of the IT team you have a strong technical expertise, you understand the company’s data structures,
the architecture of support systems; those sort of things.
I’d be surprised if anyone in the room is an actual IT team member, but nevertheless an important actor in this whole BI picture.
Some of the challenges are more related to technology and very often what happens is that the technology or the infrastructure
really does not allow the IT team to be very responsive to the MIS team and in turn to the Business Users.
Do you recognize yourself in this Shakespearean play? Are you a Business User? Are you an MIS user? A good question is, is there
such a group in your organizations? Maybe not -- the information producers may be scattered all over. You might have some in
Actuarial; you might have some in Underwriting; you might have some in Claims. Who has not heard of Natalie Johnson over
here who knows how to get things done in terms of producing reports but she’s not in the MIS group and there is no such thing
Using Business Intelligence in the Insurance Industry (Session VAL-3) 5
Proceedings of the Canadian Institute of Actuaries, Vol. XXXVII, No. 2, June 2006
in the enterprise? It might very well be a distributed function; nevertheless an important function and we have to recognize
that it’s there. Are you a Power User? I suspect that most people who have gone through the ranks of the actuarial discipline in
all of your respective companies have at one point or another been labeled as a Power User. As an information producer you
produce information, there’s no two ways about that. Are you an IT member or do you sometimes feel that you act like one, so
you are spending a lot of time learning about Oracle, databases and DB2 databases and relational technology and stuff that has
nothing to do with the actuarial science and actuarial discipline, perhaps.
Moving on to some of the examples of BI in the insurance industry; Business Intelligence Jeopardy – I wanted to have the Jeop-
ardy team going at the same time here – the question is which industry segment has the most interest in deploying Business
Intelligence. It’s a good question. Well, we have different practices: Life, Investment, Property and Casualty, Group Bene?ts,
Reinsurance Brokers, and there’s a whole lot more. Without answering the question directly, let me say this: Business Intelligence
is a discipline. Historically it came from retail.
You probably have heard the Wal-Mart story where there’s a correlation between selling beer and diapers. Anybody heard that
story before? No, nobody has heard it. Well, I’ll tell it; it’s a classic story. Apparently when diapers are on sale retailers sell more
beer and they found out using a data warehouse and using Business Intelligence apparently the explanation for this is that when
diapers are on sale mommy sends daddy to the corner store to buy diapers and by the way he picks up a six pack of beer at the
same time. I’m not making this up; this is very well documented in the history of Business Intelligence.
The point I’m making is in the retail industry – very, very, competitive – there’s a lot of rotations of movement of products and all
that sort of thing, so it’s an industry with a very, very quick dynamic of new products, of very, very high churning of customers.
Customers come in and out, in and out, in and out. There is absolutely no brand loyalty; absolutely none. Business Intelligence
helps the retail industry because it allows them again to develop business strategies and the stuff that we talked about at the
beginning of our presentation, but there’s a clear correlation between a high dynamic market and the use of Business Intelligence;
a very, very huge correlation. So the more dynamics you have in your market the more important or the better use you can make
of Business Intelligence. If your market is a very, very slow market and it takes years to develop and people come in but they
never leave, such as life insurance, term products and all that, the market dynamics are much slower.
The Business Intelligence, it has been our experience and the industry’s experience that it is probably the P&C market that can
bene?t the most from Business Intelligence because of the market dynamics of P&C. You know, I have car insurance today, I
can cancel it and get another one next year after 12 months; a very, very high dynamic. So the survey says really that P&C is
probably the market that can bene?t the most from Business Intelligence. I’m not saying that the other markets can’t bene?t
from it but the high dynamics make it very suitable for that type of business environment.
Now, business cases for why do insurance companies invest in BI; business cases vary and they can vary widely. It all has to do
with what is the strategy that the company is adopting. Do we want to go after new clients? Do we want to increase the pro?t-
ability of underwriting? What is the incentive? So they may have to do a reduction of operational risks, speed of market response.
Some companies want to deploy BI to foster better relationships with their brokers, so broker relationship management, or even
the direct writers with client relationship management, or exploiting opportunities for expense reduction. So, a question that
I have for you, and I’m not expecting an answer, is can you think of more? There are dozens and dozens of business cases that
can be developed for BI in an insurance company. An assignment when you leave the room is to think of a business case for BI
or a few business cases of BI in your respective organizations.
A few examples, broker management – broker management is a subject area that supports the analytical requirements for the
enterprise to be able to analyze and report on brokers. In other words, looking at how the broker is doing is an area unto itself.
A lot of insurance companies have deployed BI for better broker management. Their rationale is if I’m going to be dealing with
brokers I want an information-based relationship with my broker. I want to be able to go to my broker and say the piece of
business or the agent, the piece of business that you’ve written here is not so good because you used the experience. In other
words, drive the relationship using information. That’s an example. Another example is simply reg ?lings. People have been
using a BI infrastructure for regulatory ?lings to the IBC for instance in the case of P&C. Bene?ts include streamlined electronic
?lings and reduction of errors in ?lings because they’re using a BI infrastructure.
Délibérations de l’Institut canadien des actuaires, Vol. XXXVII, n
o
2, juin 2006
Utiliser l’intelligence d’affaires dans le secteur des assurances (Séance VAL-3) 6
Another example, risk attributes variance analysis, the ability to track over time the set of risks or the set of risk attributes
and how they affect the experience. You know, what if we’re rating on this, now we’re not rating on this and we are rating on
this other attribute? The bene?ts include the ability to track the evolution of a risk pro?le over time. Another example, a ?ne
example, is on-level analysis. This is actually an actuarial requirement for developing what are called a book of tariffs. So how
can we use historical information to drive our pricing models?
Here is an example, a company that shall remain unnamed that had an amalgam of operational support systems, a lot of support
systems, and very, very dif?cult to do the infamous month-end process. Every time at the end of the month those reports come
out and it’s very dif?cult. Results come out X number of days after the month-end and certainly not an environment that allowed
them to do any kind of analytics to dig into the data. What they did is they implemented a BI infrastructure for consolidating
and reporting on premiums and sales. Now the turnaround time is much quicker. They went from I would say 4-5 days after
month end to less than 24 hours after month end; 2 or 3 days faster turnaround time on month end information. They are much
closer to the famed single version of the truth; that is, the ability to get consistent reporting across the enterprise.
Another example of a company that shall also remain unnamed is a company that has adopted BI to reduce the operational
risk. This company, a P&C company – I’ll just give you some statistics – wanted to introduce a new driver class in Ontario and
it was estimated that it would take perhaps 75 person days to implement the new product and the support systems, the policy
management, the policy admin system. It would have taken about 500 days for the decision support system, so a high risk.
What they did is they implemented a BI infrastructure that allows them to simplify the reporting infrastructure; certainly a lower
risk from an operational point of view. Actuarial now have their own sandbox in which they can play with the data at their own
will. Again, the ability to implement multiple Data Marts, multiple sandboxes for different purposes.
I want to shift gears a little bit and talk about a roadmap for BI in insurance. What I propose here is a roadmap that is really
a set of milestones that an insurance company should be looking at when thinking about a BI program for the corporation.
It’s derived from experience. It’s derived from our own internal R&D and what’s happening in the market. It’s fairly simple to
understand. The roadmap really proposes four key milestones. The ?rst milestone is integrating and warehousing the data. The
second milestone – and we’ll go through these – the second milestone is to analyze and produce consumer information. The
third one is to model and simulate the operational transactions and ?nally it’s the management of business knowledge. Let’s
look at each of these steps individually.
The ?rst step is really the integration and the warehousing of data. In any BI program, if you think about the future, this is the
?rst thing that needs to happen. You need to have an environment whereby your enterprise data is integrated and warehoused
in a central location. If that’s not there – this is a question for introspection – then the future for BI is shaky at best because the
foundation will not be there. So you need that foundation of the ability to integrate your operational data into a central loca-
tion.
Step number two, once you’ve done that, once you have the ability to store data in a central location, is really get into a proc-
ess, and that involves again tools and technology but also involves people and processes to analyze, produce and consume
information. I would say that a lot of the companies that have implemented Business Intelligence are struggling with this step
of the roadmap because they have an infrastructure in place but they don’t necessarily have the people and processes in place
to really bene?t from the infrastructure. The good news is that step one and step two, really those two ?rst critical steps, are
supported with existing technologies, what we talked about at the beginning of the presentation. We talked about the Corpo-
rate Information Factory where you have your enterprise at a warehouse, where you have your data delivery, where you have
your reporting infrastructure and where you have your metadata management. Those two steps are well supported with this
Corporate Information Factory model. This I would say is going to happen in the next couple of years for most companies that
have implemented the BI program. It’s really the ability to model and simulate transactions. Really what we’re talking about here
is the ability not only to question the history because that’s one thing, question the history and look at trends and compliance
and things of that nature.
The other question is really the ability to ask the “what if” questions. You know, what if what happened in the past didn’t really
happen like that and it happened in another way. Or, what if we would price our products like this instead of like that, so the
Using Business Intelligence in the Insurance Industry (Session VAL-3) 7
Proceedings of the Canadian Institute of Actuaries, Vol. XXXVII, No. 2, June 2006
ability to ask “what if” is really the next step in the Business Intelligence roadmap. An example of this, I thought about on-level
analytics or the ability to generate an on-level premium. That’s a very good example of a “what if” scenario. So what if my rating
assumptions are varied or what if I have a new set of rating assumptions, what’s my experience? That is, it has nothing to do
with the past, it has everything to do with the future and it probably has to do with pricing, product development and such
things. That’s an example of modeling your transactions, the ability to ask the “what if” questions.
The ?nal step of the roadmap is really I guess the top of the pyramid, is really the management and the re?nement of business
knowledge. When we think about our strategic step, the ?rst step is the development of our strategic direction. A lot of it is
based on assumptions that we make, or we pose a certain set of hypotheses but they’re not necessarily always veri?ed. In this
step of the roadmap you’re in a place where you can start asking questions like this. You can, for instance, think about a rule
that says when the price of gas increases the loss ratio decreases, right, so your experience is better. Well, that clearly relies on
external data. The price of gas is not something that you have in your own data. The price of gas is something that you would go
out and get from some external data provider. It is certainly based on correlation operators, you know, an increase, a decrease, a
correlation with perhaps internal data that you may have. The loss ratio or your experience is your own, so that’s internal data.
So you can imagine, if you will, an environment where you have a set of rules, you have a rulebook that you can simply test
against your data. That’s called directed data mining. You want to mine your data to verify certain hypotheses, certain assump-
tions that you’re making. It sounds far-fetched but really it’s not that far away. It is really at the crossroads of existing technology
such as data warehousing, business rules, management, data provision, external data provision, BI, data mining, CPM (Corporate
Performance Management). All these things are coalescing really to create an environment where now we can start managing
our own hypotheses, managing our own rules by which we make strategic investments or strategic decisions.
The last piece of this talk, I want to talk about a bit of introspection, recognizing where you stand. As I said before, actuaries
have a strong in?uence over BI products for the enterprise because you are the main information producers of the enterprise.
Actuaries understand data and its derivation. You understand the business rules and how to manipulate the data. You understand
the structure of the data. Now, BI, however, is more than the actuarial database. That’s one thing I want to leave you with that BI
is simply more than the actuarial database. It’s more than procedural programming with sequential data sets. I talked about that
before. I think it’s important to think about how to change our paradigm from a procedural paradigm to a declarative paradigm.
Certainly, BI is more than impossibly large reports containing many metrics. Believe it or not, a lot of other consumers in the
enterprise are really not interested in the stuff that actuaries are interested in, in terms of their numbers or detailed numbers,
all they want is their premiums and their losses and that’s about it.
One observation, I went to a website of a highly reputable actuarial school in Canada, I won’t say where, but I was, well, I
shouldn’t say that I was surprised, by looking at the basic curriculum of actuarial science there is only one course related to the
use of computers. This one course is a course on how to use Microsoft Word and how to use Microsoft Excel, that kind of thing.
That’s a mandatory course. Then there is an elective course which is a Java programming course. This is not a very exhaustive
study on my part but as I was preparing for this talk I said let me go and look at the curriculum and I had lunch with a client of
ours in Montreal last week and he concurred. The actuarial discipline today, there is a lot of I guess data-related work with the
actuarial discipline and I found it very surprising that in these university curriculums there’s not a whole lot given to structured
programming and data modeling and information management, those sort of things. Yet, if you look at most actuarial depart-
ments today what actuaries do, a lot of it has to do with data. I was very surprised to ?nd how little focus there is on the things
I just talked about, data, structures and programming in the actuarial schools. Again, you know, we can’t change the world but
it might be interesting in the future if more attention was given to using more computer science related topics in an actuarial
program. So that’s something that I’d like you to think about.
From the business perspective, do you rely on your data to make strategic decisions? Would you be able to answer any ques-
tions that come your way regarding any line of business? Can you drive dynamic broker relationships based on information? Do
you view your decision support systems as an operational risk? Perhaps the answer is yes; perhaps the answer is no, you want
to think about that. From a technology infrastructure perspective, do you have a single integrated source of decision support
data? Do you have an auditable data integration process? Do you have metadata? Those are all questions to think about from a
technology infrastructure perspective.
Délibérations de l’Institut canadien des actuaires, Vol. XXXVII, n
o
2, juin 2006
Utiliser l’intelligence d’affaires dans le secteur des assurances (Séance VAL-3) 8
Finally, on the people and process perspectives, do you consider that any particular group within your company owns the data?
Who owns the data? Is there a centralized multi-disciplinary group responsible for information production? Do you have an
MIS component? Do you have an executive-level steering committee driving BI in your corporation? Finally, is BI considered
an IT project or a business project? These are all good questions that I want to leave you with and then I’d like to open it up
for questions.
Questions, comments, issues? That’s it, then. Thank you very much.
(applause)
doc_465126308.pdf