• An insightful guide to the use ofstatistics for solving key problems in modern-day business and

industry

Highlighting the relevance of statistical methods in everyday applications, The Role of Statistics in
Business and Industry bridges the gap between the tools of statistics and their use in today's business
world. This one-of-a-kind resource encourages the proactive use of statistics in three well-organized
and succinct parts:

Setting the Stage provides an introduction to statistics, with a general overview of its uses in business
and industry

Manufactured Product Applications explains how statistical techniques assist in designing, building,
improving, and ensuring the reliability of a wide variety of manufactured products such as appliances,
plastic materials, aircraft engines, and locomotives

Other Applications describe the role of statistics in pharmaceuticals, finance, and business services, as
well as more specialized areas including the food, semiconductor, and communications industries

This book is truly unique in that it first describes case studies and key business problems, and then
shows how statistics is used to address them, while most literature on the topic does the reverse. This
approach provides a comprehensive understanding of common issues and the most effective methods
for their treatment. Each chapter concludes with general questions that allow the reader to test their
understanding of the presented statistical concepts as well as technical questions that raise more
complex issues. An extensive FTP site provides additional material, including solutions to some of the
applications.

With its accessible style and real-world examples, The Role ofStatistics in Business and Industry is a
valuable supplement for courses on applied statistics and statistical consulting at the upper-
undergraduate and graduate levels. It is also an ideal resource for early-career statisticians and
practitioners who would like to learn the value of applying statistics to their everyday work.

The Role of Statistics in Business and Industry provides insights, not available elsewhere, into what
business and industrial statistics is all about. It prepares students and aspiring statisticians for a career
in business and industry, broadens the perspective of those already practicing, and shows non-
statisticians the value of using statistics. It supplements other texts by describing key problems that
can be addressed by statistics--based on the internationally renowned authors' extensive experience--
and then demonstrates the tools to address them successfully.

• importance of Statistics in Different Fields

Statistics plays a vital role in every fields of human activity. Statistics has important role in
determining the existing position of per capita income, unemployment, population growth rate,
housing, schooling medical facilities etc…in a country. Now statistics holds a central position in almost
every field like Industry, Commerce, Trade, Physics, Chemistry, Economics, Mathematics, Biology,
Botany, Psychology, Astronomy etc…, so application of statistics is very wide. Now we discuss some
important fields in which statistics is commonly applied.

(1) Business:
Statistics play an important role in business. A successful businessman must be very quick and
accurate in decision making. He knows that what his customers wants, he should therefore, know what
to produce and sell and in what quantities. Statistics helps businessman to plan production according
to the taste of the costumers, the quality of the products can also be checked more efficiently by using
statistical methods. So all the activities of the businessman based on statistical information. He can
make correct decision about the location of business, marketing of the products, financial resources
etc…

(2) In Economics:
Statistics play an important role in economics. Economics largely depends upon statistics.
National income accounts are multipurpose indicators for the economists and administrators.
Statistical methods are used for preparation of these accounts. In economics research statistical
methods are used for collecting and analysis the data and testing hypothesis. The relationship
between supply and demands is studies by statistical methods, the imports and exports, the inflation
rate, the per capita income are the problems which require good knowledge of statistics.

(3) In Mathematics:
Statistical plays a central role in almost all natural and social sciences. The methods of natural
sciences are most reliable but conclusions draw from them are only probable, because they are based
on incomplete evidence. Statistical helps in describing these measurements more precisely. Statistics
is branch of applied mathematics. The large number of statistical methods like probability averages,
dispersions, estimation etc… is used in mathematics and different techniques of pure mathematics like
integration, differentiation and algebra are used in statistics.

(4) In Banking:
Statistics play an important role in banking. The banks make use of statistics for a number of
purposes. The banks work on the principle that all the people who deposit their money with the banks
do not withdraw it at the same time. The bank earns profits out of these deposits by lending to others
on interest. The bankers use statistical approaches based on probability to estimate the numbers of
depositors and their claims for a certain day.

(5) In State Management (Administration):

Statistics is essential for a country. Different policies of the government are based on statistics.
Statistical data are now widely used in taking all administrative decisions. Suppose if the government
wants to revise the pay scales of employees in view of an increase in the living cost, statistical
methods will be used to determine the rise in the cost of living. Preparation of federal and provincial
government budgets mainly depends upon statistics because it helps in estimating the expected
expenditures and revenue from different sources. So statistics are the eyes of administration of the
state.

(6) In Accounting and Auditing:

Accounting is impossible without exactness. But for decision making purpose, so much precision
is not essential the decision may be taken on the basis of approximation, know as statistics. The
correction of the values of current asserts is made on the basis of the purchasing power of money or
the current value of it.
In auditing sampling techniques are commonly used. An auditor determines the sample size of
the book to be audited on the basis of error.

(7) In Natural and Social Sciences:

Statistics plays a vital role in almost all the natural and social sciences. Statistical methods are
commonly used for analyzing the experiments results, testing their significance in Biology, Physics,
Chemistry, Mathematics, Meteorology, Research chambers of commerce, Sociology, Business, Public
Administration, Communication and Information Technology etc…

(8) In Astronomy:
 Astronomy is one of the oldest branch of statistical study, it deals with the measurement of
distance, sizes, masses and densities of heavenly bodies by means of observations. During these
measurements errors are unavoidable so most probable measurements are founded by using
statistical methods.
Example: This distance of moon from the earth is measured. Since old days the astronomers have
been statistical methods like method of least squares for finding the movements of stars.
 • Statistical methods for business intelligence

Innovation Matters

Business data exist in many different forms, including time series and longitudinal data, and may
contain various types of variables, such as time to event, occurrences, categorical and continuous
measurements. Statistical data analysis includes data management, modeling, segmentation and
application integration, and seeks to uncover useful information; forecast business processes at
various horizons; and detect hidden structural changes.

Recent examples include:

Revenue forecasting
Demand forecasting
Network traffic forecasting
Expense tracking
Customer profitability analysis
Targeting analytics
Capacity planning
System monitoring
Warranty data analysis

The challenge of business analytics is that one has to estimate the parameters of the model without
comprehensive information. For instance, IBM Research may be asked to develop a control center (or
dashboard) that will track a portfolio of large IT services customers and identify those that might
decide to terminate their contract, or renegotiate it to achieve a lower price, which would lead to a
significant loss of revenue to the company. In this case, as shown in Figure 1, the inputs to the model
are numerous variables that describe the following five risk factors: 1) financial health of client
companies; 2) previous relationships with the service provider; 3) price and competitiveness of the
offered service; 4) significant events in the client company that could have a potential impact on the
decision to cancel the service (e.g. change of CEO, merger, restructuring, etc.); and 5) previous history
of contract terminations or renegotiations. The output variable is the likelihood that a customer will
terminate its contract (or a part of it).

Conventional methods are limited to estimating the likelihood of termination, without providing
insights into which factor is most influential in the decision. Yet, knowing the impact of different factors
to the client's decision can help the service provider influence the outcome. For example, if the
decision to terminate is based on limited cash availability, the service provider might arrange different
financing. On the other hand, if the decision is based on the low satisfaction with the service, the
service provider can influence the outcome by improving service and mobilizing its sales and
marketing teams to save the relationship. Typically only the variables that influence the risk factors
are known, not the risk factors themselves, which means they are considered hidden states.
In traditional methods, such as logistic regression, the values of these hidden states are computed as
a byproduct of the model. However, in many applications, at least some of the relationships among
these factors (i.e. hidden variables) are known. For example, in the aforementioned problem of the
dashboard design, it is often possible to provide additional information in the form of "Company A has
been more satisfied than Company B" or "Company C has better financial health than Company D."
Because the traditional parameter estimation techniques, such as (Multidimensional Analysis Platform
(MAP) or iteratively re-weighted least squares, do not account for these relationships, the estimation of
hidden variables obtained with the standard parameter estimation procedures is not optimal. An
estimation procedure that captures such relationships in the data is needed.

At IBM Research, an algorithm was developed to compute the regression parameters, taking into
account the issue of estimating hidden variables with constraints. A recursive adaptation procedure
has also been developed to identify the most significant nonlinear relationships in the data and to
adapt the model by introducing corresponding higher order terms. The entire approach is embedded in
a Web-based platform and used to track the portfolio of large IT services clients in the IBM Global
Services division, and identify those who are likely to terminate or reduce the scope of their
engagement.
Ibexi Solut ions Page 1

Solving Insurance Business Problems

Using Statistical Methods
Anup Cheriyan

Ibexi Solut ions Page 2

Table of Contents
Executive Summary...........................................................................................3
About the Author ...............................................................................................3
Introduction .......................................................................................................4
Common statistical methods ..............................................................................4
1. Descriptive statistics ............................................................................................ 5
2. Inferential statistics.............................................................................................. 6
Data Sources ......................................................................................................7
Use Of Statistics For Specific Business Problems ..............................................8
1. Insurance Agent Attrition .................................................................................... 9
2. Productivity Of Insurance Agents ...................................................................... 10
3. Lapse Ratio & NTU Of Policies ........................................................................ 11
4. New Product & Market Development ................................................................ 13
5. Cross Selling ..................................................................................................... 15
Conclusion.......................................................................................................17
Annexure 1- Important Statistical Definitions ..................................................18
Annexure 2- References...................................................................................22

Ibexi Solut ions Page 3

Executive Summary
This paper deals with the use of statistical methods for solving insurance business problems.
It explains why statistical methods are well suited for insurance, and identifies some business
problems that can be solved using these approaches.
It introduces common statistical methods used in solving business problems. Statistical
methods like descriptive statistics, inferential statistics, hypothesis testing, chi square analysis
of frequencies, strength of association and scatter plots are described.
Finally, some specific business problems are discussed. These include insurance agent
attrition, productivity of insurance agents and high lapse ratio. For each problem, the impact
and cost to the business are explained. Then some approaches to solve the problem, the data
required and the statistical method that can be employed are described.
About the Author
Anup Cheriyan is Associate consultant with Ibexi Solutions. He
has experience in implementation of Business Intelligence
solutions for insurance companies in India.

Ibexi Solut ions Page 4

Introduction
Decision making processes must be based on data, not on personal opinion or on belief. In the
real world decisions need to be made even though there are uncertainties. Business statistics
is a science assisting managers to make business decisions under uncertainties based on some
numerical and measurable scales. It is used in many disciplines, such as financial analysis,
production, operations and marketing research.
Business statistics is used to make inferences about
certain characteristics of a population based on
information contained in a random sample from the
entire population. This enables managers to:
! Solve problems in a diversity of contexts.
! Add substance to decisions.
! Minimize guesswork.
The insurance business is rich in data, and in data complexity. Retail insurance business (e.g.
life and motor) typically has a large number of clients and policies. Mature companies in
many countries have millions of policies. In emerging countries such as India, even new
companies less than 5 years old have a million clients; older large companies e.g. LIC have
over 130 million policies. Insurance policies have a large amount of data, and they are
complex in structure, with variations such as benefits, face amounts, schemes, pricing,
claims, multiple client relationships, medical history and family history and underwriting.
This combination of volume and complexity is unusual; this makes it difficult to manually
understand the data, and its trends. Thus, the insurance business is ideally suited for the
application of statistical methods.
Currently, use of statistics is largely limited to actuaries for determination of the insurance
premium rates. Statistics can have wide applications in other departments of an insurance
company. For instance, the agency department can use statistical methods for combating
high agent attrition rates and hiring productive agents. Also, the marketing department can
use statistics to identify target customers for cross selling a new insurance policy.
Common statistical methods
Before we venture into the applications, let us review common
statistical methods like descriptive statistics, inferential
statistics, hypothesis testing, chi-square analysis of frequencies,
scatter plots, simple linear regression, multiple regression and
non-linear regression, and their direct uses. Descriptive
statistics are used to summarize or describe the population
sample, whereas inferential statistics are used to test theories
about the population and for generalizing the behaviour of a
population sample.

Ibexi Solut ions Page 5

1. Descriptive statistics
Descriptive statistics are used to describe the shape of a sample, where the data tend to
cluster, and how the data are dispersed. This type of analysis is supported by most of the
OLAP tools available in the market, and can be used to get insights into the business.
Type of analysis Example
Central
Tendency, using
! Mode
! Median
! Mean
What is the number of policies per month sold?
Month Number of policies sold
January 295
February 599
March 630
April 509
May 465
June 600
July 309
August 630
September 645
October 500
November 307
December 780

Mode=630;
Median=554;
Average=522
This reveals that we sell 522
policies per month, on an
annual average; however, as
the median is 554, which is
greater than the average, it
indicates that the lowervolume months are pulling
down the performance more
sharply than can be made up
by the higher-selling
months.
Distribution of
Data, using
! Frequency
Distribution
! Histogram
What is the distribution of policies by product class, assured gender?
Illustration:
Product / Gender Company Female Male
Level Term Insurance 32,525 25,296 63,295
Pure Endowment 2,090 56,092 217,092
Universal Life 490 7,267 20,880
Term Insurance
27%
21%
52%
Endowment
1% 20%
79%
Universal life
2% 25%
73%
The distribution data
indicate a difference in the
purchase pattern for pure
term insurance among
genders, as opposed to
investment products like
endowment and universal
life.

Ibexi Solut ions Page 6

Variability,
using
! Variance
! Standard
Deviation
! Coefficient
of variation
What is the seasonal variance across months in a year? Is this trend
changing? Is seasonality growing with time?
In this example, the standard deviation for monthly sales for each year
can indicate whether the seasonality is increasing or decreasing.
2. Inferential statistics
Inferential statistics involves a decision-making process that allows us to objectively quantify
if treatment effects are significant or not. The success of this process requires that we make
certain assumptions about how well the sample represents the larger population. These
assumptions are based on two important concepts of statistical reasoning: probability and
sampling error. A statistical comparison of sample parameters is required to determine if the
difference is ìtrueî or ìfalseî: such a comparison is called hypothesis testing. The chi-square
statistic is a nonparametric statistical technique used to determine if a distribution of observed
frequencies differs from theoretical expected frequencies. Various strengths of association are
also discussed in this section.
Analysis Statistical method Example
Hypothesis testing
! Hypothesis testing
! Chi Square test
! Simulation
Is the increase in sales an outcome
of ad campaign?
Strength of association
! Correlation coefficient
! Linear regression
! Multiple regression
! Non-linear regression
Is there a correlation between ad
spending and sales volume?
Segmentation
! Cluster analysis
! Scatter plots
Which customer segment buys
maximum number of whole life
policies?

Ibexi Solut ions Page 7

Data Sources
Collection of relevant data is prerequisite for employing statistical methods for the same.
Data required can be categorized into primary and secondary data depending upon the data
source.
Primary data ñ fresh data collected by the
researcher himself
Secondary data ñ data already collected by
others or available internally within the
company, which will be "re-used"
Secondary data may be internal to the firm, such
as sales invoices and warranty cards, or may be
external to the firm such as published data or
commercially available data. The government
census is a valuable source of secondary data.
It is possible to categorise secondary data based on source and degree structure. This
typically indicates the ease of using the data. Internal data is more likely to be complete and
available easily; external data like those from research bureaus and government are often not
complete for the analysis. Also, structured data is more amenable to analysis than
unstructured data.

IMRB, Census
Interest rates, Tax changes,
Regulatory changes,
Product launched by
compet it ion
Typically from databases
and transactional systems
Launch of new products by
the company, new
partnership agreements
Secondary data has the advantage of saving time and reducing data gathering costs. However,
there are limitations of secondary data. These include:
D
A
T
A

S
O
U
R
C
E
External
Internal
DATA STRUCTURE
Structured Unstructured
Examples of secondary data in different categories

Ibexi Solut ions Page 8

! It is collected for a different purpose
! There might be problem of definitions
! Secondary data will have problem of comparability over time
! Lack of awareness of sources of error/bias in data is a limitation.
Therefore secondary data may not fit the problem perfectly. Secondary data alone is not
sufficient to analyse business problems. Secondary data must be supplemented by primary
data originated specifically for the study at hand. Primary data can be obtained by initiating a
market research.
Some common types of primary data are:
! Demographic and socio-economic characteristics
! Psychological and lifestyle characteristics
! Attitudes and opinions
! Awareness and knowledge - for example, brand awareness, insurance awareness
! Intentions - for example, purchase intentions. While useful, intentions are not a
reliable indication of actual future behaviour.
! Motivation - a person's motives are more stable than his/her behaviour, so motive is a
better predictor of future behaviour than is past behaviour.
Use Of Statistics For Specific Business Problems
In insurance, there are many business problems in different areas that can be tackled using
statistical approaches. These include:
Agency department:
! Agency force attrition
! Insurance agent productivity and agent success factors
Renewals department:
! High lapse in the initial years of the policy
Marketing & sales department:
! Identification of customer segment for cross selling
! Features to be added to a new product and understanding
customer needs
! Identifying gaps in product mix
! Identifying gaps in product mix
! Customer segmentation
Actuarial department:
! Enhancing product profitability
Underwriting department:

Ibexi Solut ions Page 9

! Fine tuning of Auto underwriting methods
Operations department:
! Reducing turnaround times of New business and policy owner servicing processes
! Fraud detection patterns
! Enhancing product profitability

To solve a real-world business problem, it is necessary to use a combination of standard

statistical methods, a combination of tools and data sources.
1. Insurance Agent Attrition
High attrition rate of insurance agents is one of the biggest challenges for an insurance
company. Conservative estimates put the attrition rates at 35-40 per cent.
For new insurance companies still struggling to break even, the rising attrition rate is yet
another challenge that they have to battle. For mature companies too, the attrition rate
especially in the face of rising competition is a growing threat.
Agents leave company because of various reasons like:
- Compet it ion offers more remuneration
- Company does not having good products to offer
- Attrition of sales managers
- Lack of proper training resulting in unsuccessful sales career
- Unsuccessful in adapting to sales
- Unsuccessful in adapting to insurance
- High work pressure and targets.
Solving this problem involves finding reasons resulting in high attrition rate, and identifying
actions, which can address these reasons. Business questions that would answer this include
- Are most of those who drop out non-performers?
- Can adequate training increase performance of agents? Will this help in reducing
attrition?
- Will introducing new products which appeal to the customers better, solve the
problem of high attrit ion?
- Has booming economy caused the rampant poaching of insurance agents? Will it
settle down once industry has their expanding networks in place?
To answer the business questions, some of the data that is required include:
- Historical data of existing agents and terminated agents: Agent profile, number of
policies sold, first premium collected, renewal premium collected
- Training period of agents in the company
- Remuneration and reward models for agency force within the company
- Agents monthly and annual targets within the company
- Exit interviews of agent who leave the company

Ibexi Solut ions Page 10

- Average training period of agents in the industry
- Remuneration and reward models for agency force of competitors
- Agents monthly and annual targets for competitors
Note that the last three are external data, and may be incomplete.
Recommended actions that can be taken to solve this problem include:
- Adjust the hiring pattern to match demographic and psychographic patterns
- Training effectiveness and duration can be changed if required
- Build a predict ive agent attrit ion model to determine agents likelihood of attrit ion
Illustration: Predictive agent attrition model
Predictive model is an equation used to predict a variable. Statistical methods like regression
analysis can be used to develop predictive models.
The solution involves the use of past data of companyís agents as well as data about
terminated agents. This data can used to develop a predictive model that would assign a score
to each and every agent. This score would indicate the agent's propensity to leave in a defined
timeline. Input data for modelling included key demographic and behavioural information.
Agents could now be scored for their likelihood of attrition. The attrition scores, used in
conjunction with the expected lifetime value of the agent can business managers with a
continuous trade-off framework will help decide on offers to negotiate with agents in a bid to
retain them.
2. Productivity Of Insurance Agents
New people are ultimately joining the insurance industry; these days an image makeover as
"life insurance advisors" is in fashion. But little has changed in the basic nature of the
business - insurance still needs to be sold to a reluctant populace. Sustenance requires
constant networking and acquiring new relationships for business.
In a business such as insurance one has to accept the fact that 20 per cent of the work force
will bring in 80 per cent of their business.
Agent productivity depends on factor like:
- Agentsí age, gender
- Educational qualification
- Work experience
- Product knowledge (training received)
- Selling skills
- Motivation

Ibexi Solut ions Page 11

Solving this problem will involve identifying factors that results in high productivity of
agents.
- Can adequate training increase performance of agents?
- Are older agents showing better performance than younger ones?
- Is higher educational qualification positively or negatively correlated with
performance?
- Are female agents more productive than male agents or wise versa?
- Does an agent with work experience perform better than a new agent? Is it wise to
poach insurance agents from competitors?
To answer the business questions, some of the data that is required include:
Internal data:
- Training period of agents in the company
- Remuneration and reward models for agency force within the company
- Work experience of agents
- Number of policies sold, first premium collected, renewal premium collected for
agents.
- Educational qualification of agents
- Agent details (age, gender, etc)
External data:
- Average training period of agents in the industry
- Remuneration and reward models for agency force of competitors
Illustration: Insurance agent performance and age
Did you ever wonder whether the insurance agents performance depends on his age? Strength
of such a relationship between performance and age can be quantified by using correlation
coefficient. The correlation coefficient summarizes the relationship between two variables
It might be that the agents who are aged have more experience, and they sell more policies.
This would be an example of a positive correlation, because high values of one variable (e.g.,
agents age) are associated with high values on the other variable (e.g., better performance on
sales). Or it might be the other way around: age of person is associated with poorer sales
performance. The latter is an example of a negative correlation, because high values on one
variable are associated with low values on another variable.
Similarly strength of association of other factors like agents training duration, educational
qualifications, etc on agent performance can be determined by using correlation coefficient.
3. Lapse Ratio & NTU Of Policies
Lapse ratio is calculated by dividing number of policies lapsed by total number of policies. It
is a ratio used to measure the effectiveness of an insurer's renewal strategy. A lower lapse

Ibexi Solut ions Page 12

ratio is better, particularly because insurance companies pay high commissions to brokers and
agents bringing new clients.
Higher lapse ratio and NTU can be due to various reasons like:
- Not providing timely and accurate (e.g. notice to wrong address) reminders to
customer for renewal, results in higher Lapse ratio.
- Churning. This is a major contributor to the high lapse and surrender rates. Churning
is the practice of encouraging customer to cancel an existing policy to take out a new
policy or another investment. What many policyholders do not realize is that this
results in double cost: penalties apply in most cases where a policy is cancelled and
then a whole new round of costs are incurred on the new investment. One of the major
reasons for churning is the commission structure of the life insurance industry.
Currently, in the case of most life assurance products, commissions are paid upfront,
with 35 to 40% of the commission paid in the first year of the policy. As a result
financial advisor will encourage customer to switch investments to earn a living.
- Pressure sales tactics by agents, resulting in purchase of policies for higher premiums
than can be continually paid by the client during renewal.
Business questions that may help address this problem are:
- Is there a correlation between lapse ratio and customers annual income, especially if
the premium is worked at as a percentage of their income?
- Does certain product characteristic of policy result in higher lapse and NTU?
- Is there a specific demographic profile of customers who NTU & lapse their policies?
- Do some channels contribute significantly more to high lapsed and NTU?
- Are there group of agents who contribute to higher lapse ratios?
- Does policies issued at end of a quarter have more probability of lapsing than policies
issued at other times?
To answer the business questions, some of the data that is required include:
Internal data:
- Remuneration and reward models for agency force within the company
- Agent targets and evaluation parameters
- Agent wise data of number of policies issued, Inforced, lapsed and NTU
- Agent details (age, gender, etc)
External data:
- Remuneration and reward models for agency force of competitors
-
Illustration: In 3
rd
Quarter of 2005 an insurance company decides to do Lapse ratio analysis
Histogram can be used to describe the lapse ratios of policies with respect to policy Inforced
year.

Ibexi Solut ions Page 13

It can be seen that companyís lapse
ratio is as high as 40%. Also, this is
consistent across policies issued in
different years. Changes in renewal
management and lapse reduction
schemes have made no impact on the
lapse ratio.
Now it will be useful if one know
which are the channels leading to high
lapses. Therefore a histogram showing
lapse ratio across channels for period from 2001 to Q3 2005 will serve our purpose.
The histogram shows
that 90% of the policies
have lapsed in ìsocial
and ruralî channel.
This indicates the need
to focus on this channel
for improving the lapse
ratio.
Now the company would like to know if owner income and annual premium have any impact
on lapses. This can be found out by calculating the coefficient of correlation.
It may also be instructive to correlate lapse ratio with the %(premium/ ownerís income). This
may lead to changes in financial underwriting. Further one can formulate a regression
equation. A regression equation allows us to express the relationship between two or more
variables algebraically. It indicates the nature of the relationship between two or more
variables. In particular, it indicates the extent to which you can predict some variables by
knowing others.
4. New Product & Market Development
Company should choose the expansion method that best fits company's product or service,
strengths and weaknesses, and the limitations of cash, credit and existing resources.
- New product development can be one of the methods of expansion.
- New market development can help you reach new customers
Year wise lapse ratio analysis
0%
10%
20%
30%
40%
2001
2002
2003
2004(Q1+Q2)
Inforce year
Lapse ratio
Channel wise lapse ratio analysis
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
Agency Bancassurance Direct marketing Corporate
agency
Social & rural
Channel
Lapse ratio

Ibexi Solut ions Page 14

While developing a new product business users would need to address the following
questions:
- Which product features are important to customer?
- What are the gaps in product portfolio that need to be filled?
- What are the gaps in companyís product mix? For example if the company has
very low sales of policies in insured age band between 25 to 30 years then it can
introduce new product to cater to this specific age band and fill in the gap in the
product mix.
Market segmentation analysis is a key tool for
understanding what drives consumersí purchase
behaviour and how their values and beliefs affect
their behaviours. By identifying distinctive groups
of consumersótheir characteristics, needs,
purchase patterns, attitudes and/or valuesó
segmentation research can help firms develop
offerings targeted to the most profitable
consumers.
While taking products to a new market following questions need to be addressed by the
management:
- What is the demographic profile of people who buy the product?
- Does the demographic profile vary across products?
- Has it varies across years?
To answer the business questions, some of the data that is required include:
Internal data:
- Using Existing customer base to vet a new product's potential market value.
- Customer profile for each policy (Demographic and socio-economic)
External Data:
- Psychological and lifestyle characteristics
- Attitudes and opinions
- Awareness and knowledge - for example, brand awareness
- Intentions - for example, purchase intentions. While useful, intentions are not a
reliable indication of actual future behaviour.
Illustration: Determine features of insurance policy important to a people in a particular
market segment

Ibexi Solut ions Page 15

Products promise a set of benefits like adjustable face amount, flexible premiums, option to
add riders, Ease of making claims, premium reminders, tacking policy details online, Tax
advantage, security, retirement planning, unit linked, etc.
These product features can be grouped into 3 factors:
Convenience (Factor X):
Variable 1: Ease of making claims
Variable 2: Premium reminders
Variable 3: Tacking policy details online
Flexibility (Factor Y):
Variable 1: Adjustable face amount
Variable 2: Flexible premiums
Variable 3: Option to add riders
Investment returns (Factor Z):
Variable 1: Tax advantage
Variable 2: Retirement planning
Variable 3: Unit linked

The factor loading can be defined as the correlations between the factors and their underlying
variables. A factor-loading matrix is a key output of the factor analysis.
Factor X Factor Y Factor Z
Variable 1 0.7 0.6 0.6
Variable 2 0.6 0.6 0.8
Variable 3 0.9 0.7 0.8
Columnís Sum of
Squares
1.66 1.21 1.64
Each cell in the matrix represents correlation between the variable and the factor associated
with that cell. The square of this correlation represents the proportion of the variation in the
variable explained by the factor. The sum of the squares of the factor loadings in each column
is called an eigenvalue. An eigenvalue represents the amount of variance in the original
variables that is associated with that factor. The communality is the amount of the variable
variance explained by common factors.
5. Cross Selling
It costs 5 times more to acquire a new customer than retaining an existing customer.
Encouraging existing customers to spend more can have a dramatic effect on companyís

Ibexi Solut ions Page 16

profit margins. For example, Telecom company entering Insurance business can sell
insurance to their existing customers; an Insurance company with composite license can sell
life insurance to there general insurance customers; an Insurance company can aim at repeat
business from existing satisfied customers by offering them riders and new policies based on
the customers needs. Cross selling if done successfully enhances loyalty and reduces the cost
of sales.
While encouraging existing customers to spend more
management should find answers to the following questions:
- Who are the target customers for cross selling a
particular product?
- Will investing heavily in call centres and technology
creates more satisfied customers for cross selling?
- Will training to customer-service representatives that
increases the quality and consistency of service, help
cross selling?
- For any given customer household, what product or
products should be marketed next?
To answer the business questions, some of the data that is required include:
Internal data:
- Details of policies sold to existing customers
- Customer profile (age, gender, annual income) that can be used to predict
financial needs.
- Training details of customer-service representatives that increases the quality and
consistency of service
- Investments in call centres and technology
External data:
- Life style data (wants, needs and desires)
Illustration: Identifying the next insurance product to sell to a particular client
The approach is to develop a targeting model for each possible product or service type. This
predictive model would use recent product or service acquisition as the dependent variable. It
would use other product usage and demographic and lifestyle data as predictors.
By using pre-existing product ownership, demographics and lifestyle variables predictors, we
can generate not only a powerful scoring model for cross selling, but also a richer
understanding of why a household's characteristics indicate that the household is a good
candidate for targeting.

Ibexi Solut ions Page 17

Each client is scored on each model separately, using a separate model for each product or
service. Then the model scores are ranked from high to low for each client. And the next
product to sell to a particular client is determined by the highest scoring product from among
the array of candidate products that the client has not yet acquired.
Conclusion
Today's competitive insurance market has made the critical link between good decisionmaking and
success more important than ever. Increasingly organizations are turning to
statistical analysis to guide decision-making processes.
Business data exist in many different forms, including time series and longitudinal data, and
may contain various types of variables, such as time to event, occurrences, categorical and
continuous measurements. The challenge in using statistics for business is that one has to
estimate the parameters without comprehensive information. There are many statistical
analysis tools on the market today that provides a complete, comprehensive and integrated
platform for data analysis. Using optimal statistical techniques can provide new information
that improves process turnaround time, reduces policy lapse ratios, and helps retain valued
and satisfied agents; hence driving development and revenues.

Ibexi Solut ions Page 18

Annexure 1- Important Statistical Definitions
Hypothesis Testing
Estimation of population parameters is also useful to answer questions about comparisons or
relationships: "Is one treatment more effective than the other?î or "Is there a relationship
between length of treatment and treatment outcome?" These types of questions usually
involve comparison of means, proportions, standard deviations, or some other statistic.
According to the concept of sampling error, sample parameters (e.g. means) of different
samples will be expected to differ based on the nature of sampling. A statistical comparison
of those parameters is required to determine if the difference is ìtrueî or ìfalseî. Such a
comparison is called hypothesis testing.
When evaluating observed differences between different samples, we must consider two
different outcomes: either the two samples are truly different, or the differences are due to
chance occurrence. The second of these two outcomes is called the null hypothesis, H0. The
first of these two outcomes is called the alternate hypothesis, H1.
Chi-Square Analysis of Frequencies
The chi-square statistic is a nonparametric statistical technique used to determine if a
distribution of observed frequencies differs from theoretical expected frequencies.
Goodness of Fit: tests if proportions are the same as theoretical expected proportions,
Independence Tests: tests for relationships among categorical data based on proportions
McNemur Test: used for correlated samples (e.g. comparing proportions of correct diagnosis
using two different modalities, such as MRI versus CT for detecting cartilage defects)
Coefficients of Association: tests if two categorical variables are related significantly
Any appropriately performed test determines the degree of confidence one can have in
accepting or rejecting a hypothesis. Typically, the hypothesis tested with Chi Square is
whether or not two different samples are different enough in some characteristic or aspect of
their behaviour that we can generalize from our samples that the populations from which our
samples are drawn are also different in the behaviour or characteristic. Chi Square is used
most frequently to test the statistical significance of results reported in bivariate tables
Correlation
The correlation coefficient summarizes the relationship between two variables. Correlation is
a measure of association that tests whether a relationship exists between two variables. It

Ibexi Solut ions Page 19

indicates both the strength of the association and its direction. A correlation coefficient is a
number between -1 and 1, which measures the degree to which two variables are linearly
related.
Let's take an example. Did you ever wonder whether the insurance agents performance
depends on his age? It might be that the agents who are aged have more experience, and they
sell more policies. This would be an example of a positive correlation, because high values of
one variable (e.g., agents age) are associated with high values on the other variable (e.g.,
better performance on sales). Or it might be the other way around: age of person is associated
with poorer sales performance. The latter is an example of a negative correlation, because
high values on one variable are associated with low values on another variable.
Scatter plots
It is useful to obtain a plot of the joint distribution of the values of the two variables, ìage of
the agentî(X) and ìsalesî(Y). These are called scatter plots If small values of X are
associated with small values for Y, and large values of X are associated with large values of
Y, then the data will stretch from the lower left hand corner of the plot to the upper right hand
corner of the plot. This indicates a positive relationship.
If small values of X are associated with large values for Y, and large values of X are
associated with small values of Y, then the data will stretch from the upper left hand corner of
the plot to the lower right hand corner of the plot. This indicates an inverse relationship. If
there is no discernible pattern to the distribution, then the two variables probably are not
related in a linear fashion.
Scatter plot
0
10
20
30
40
50
60
70
20 30 40 50 60
Agent's age
Number of policies sold
The above scatter plot depicts a negative correlation between agents age and sales
performance.

Ibexi Solut ions Page 20

Simple Linear Regression
Simple linear regression aims to find a linear relationship between a response variable and a
possible predictor variable by the method of least squares.
Multiple Regression
Multiple linear regression aims is to find a linear relationship between a response variable
and several possible predictor variables.
Non-linear Regression
Non-linear regression aims to describe the relationship between a response variable and one
or more explanatory variables in a non-linear fashion.
Regression Equation
A regression equation allows us to express the relationship between two (or more) variables
algebraically. It indicates the nature of the relationship between two (or more) variables. In
particular, it indicates the extent to which you can predict some variables by knowing others,
or the extent to which some are associated with others.
A linear regression equation is usually written as:
Y = a + bX + e
Where,
Y is the dependent variable
a is the intercept
b is the slope or regression coefficient
X is the independent variable (or covariate)
e is the error term
The equation will specify the average magnitude of the expected change in Y given a change
in X.
Regression line
The regression equation is often represented on a scatter plot by a regression line.
Regression lines can be used to visually depicting the relationship between the independent
(x) and dependent (y) variables in the graph. A straight line depicts a linear trend in the data
(i.e., the equation describing the line is of first order. For example, y = 2x + 7). A curved line
represents a trend described by a higher order equation (e.g., y = 3x
2
+ 7x - 9).

Ibexi Solut ions Page 21

How well this equation describes the data (the 'fit'), is expressed as a correlation coefficient
(R
2
). The closer R
2
 is to 1.00, the better the fit.
Regression line
y = -1.0403x + 71.45
R
2
= 0.6856
0
10
20
30
40
50
60
70
20 30 40 50 60
Agent's age
d Number of policies sol

Ibexi Solut ions Page 22

Annexure 2- References
www.georgetown.edu
www.entrepreneur.com
www.clearlybusiness.com
www.clomedia.com
www.thehindubusinessline.com
www.stats-consult.com
www.home.ubalt.edu
www.smartdrill.com
www.nowsell.com
www.financetech.com
www.persfin.co.za
www.stats.govt.nz
www.sas.com