Methods and systems for identifying early terminating loan customers

• US 7,392,221 B2
• Filed: 04/06/2001
• Issued: 06/24/2008
• Est. Priority Date: 04/06/2001
• Status: Active Grant

• Current Assignee: GE Capital US Holdings, Inc. (General Electric Company)
• Original Assignee: General Electric Capital Corporation (General Electric Company)
• Inventors: Nabe, Oumar, Valtchanov, Rossen G., Bell, Ian R., Fogarty, David J., Karim, Maria Abdul
• Primary Examiner(s): Felten; Daniel S.
• Application Number:
  US09/827,721
• Publication Number:
  US 20030097329A1
• Time in Patent Office: 2,636 Days
• Field of Search: 705/38, 705/41, 705/42, 235/379, 235/380
• US Class Current: 705/38
• CPC Class Codes: G06Q 20/105 involving programming of a portable memory device, e.g. IC cards, "electronic purses"
• CPC Class Codes: G06Q 20/108 Remote banking, e.g. home banking
• CPC Class Codes: G06Q 30/02 Marketing; Price estimation or determination; Fundraising
• CPC Class Codes: G06Q 40/02 Banking, e.g. interest calculation or account maintenance credit or loans G06Q40/03
• CPC Class Codes: G06Q 40/03 Credit; Loans; Processing thereof

Abstract

Methods and systems for identifying and retaining customers who are likely to terminate an existing loan contract at a date earlier than the loan termination date identified in the loan contract are disclosed. The method includes the steps of using an early termination model to identify customers likely to terminate an existing loan contract at a date earlier than the loan termination date identified in the loan contract, notifying those customers of new loan opportunities at competitive loan interest rates and attempting to cross sell new loans to the customers before the customer early terminates their existing loan.

Claims

1. A method for identifying and retaining customers who are likely to terminate an existing loan contract at a date earlier than the loan termination date identified in the loan contract, wherein the existing loan contract is provided by a lender that provides financing for purchasing a product by a customer from a dealer, said method comprising the steps of:

storing customer data in a database including a payment history for each existing loan contract by the corresponding customer;

using an early termination model to identify for the lender customers likely to terminate an existing loan contract at a date earlier than the loan termination date identified in the loan contract, the early termination model uses the customer data stored within the database;

providing a list of early termination customers identified by the lender to the dealer, each customer identified by the lender on the list of early termination customers is a customer that satisfies the early termination model and is a customer the lender would like to retain as a customer;

notifying the customers identified on the early termination list by the dealer of new products and new loan opportunities at competitive loan interest rates, wherein the new loan opportunities are provided by the lender to finance a future purchase of new products from the dealer by the customers identified on the early termination list; and

cross selling the new loans to the customers identified in the early termination list before the identified customers terminate the existing loan contracts associated therewith.

2. A method according to claim 1 wherein said step of using an early termination model to identify for the lender customers likely to terminate an existing loan contract further comprises the step of utilizing six months of performance and demographic data to predict likely loan terminations at least three months before the customer is likely to prepay the existing loan.

3. A method according to claim 1 wherein said step of using an early termination model to identify for the lender customers likely to terminate an existing loan contract further comprises the steps of:

predicting a probability of early termination at every stage of the lifetime of the existing loan;

combining the probability of early termination with financial calculations to determine an expected profit at each stage of the existing loan;

comparing the expected profit from the existing loan with an expected profit from a new loan; and

determining which customers to target for a new loan based upon the expected profits from the existing and the new loans.

4. A method according to claim 3 wherein said step of predicting a probability of early termination further comprises the steps of predicting probabilities using probability properties P(A₁∪

A₂∪

A₃)=1−

P((A₁∪

A₂∪

A₃^c))=1−

P(A₁^c∩

A₂^c∩

A₃^c)=1−

P(A₁^c)P(A₂^c|A₁^c)P(A₃^c|A₁^c∩

A₂^c), where the events A₁^c, A₂^c|A₁^c, A₃^c|A₁^c∩

A₂^c represent the probabilities that the customer has not early terminated.

5. A method according to claim 4 wherein said step of combining the probability of early termination with financial calculations further comprises the step of calculating an expected income, given a set of probabilities according to

$\begin{array}{c}E\end{array}$

( Income ) = ⁢

( 1 - P 1 ) ⁢

I 1 + ( 1 - P 1 ) ⁢

( 1 - P 2 ) ⁢

I 2 + ⁢

. . . ⁢

+ ( 1 - P 1 ) ⁢

. . . = ⁢

∑

i = 1 T ⁢

⁢

[ ∏

k = 1 i ⁢

⁢

( 1 - P k ) ] * I i . ⁢

( 1 - P T ) ⁢

I T =

6. A method according to claim 5 wherein said step of comparing the expected profit from the existing loan with an expected profit from a new loan comprises the step of calculating a monthly internal rate of return for each customer according to:

$\underset{t=1}{\overset{T}{\sum <br><br>}}<br><br><br><br>\frac{C_t}{{\left(1+I_{\mathrm{RR}}\right)}^t}=0.$

7. A method according to claim 1 wherein said step of using an early termination model to identify for the lender customers likely to terminate further comprises the step of compiling customer data from multiple sources including at least one of world wide web, legacy customer data, facsimile, telephone and cellular data.

8. A method according to claim 1 wherein said step of cross selling the new loans to the customers further comprises the step of cross selling at least one of insurance product, loan products different from existing loan product and refinancing of existing loan product.

9. A method according to claim 1 further comprising the step of building logistic regression model to test accuracy and a strength of association between variables which describe customers and their loans within the early termination model for an i^th experimental unit according to:

${\prod <br><br>}_i<br><br>=F<br><br>\left({\mathrm{\beta <br><br>}}_0+\underset{j=1}{\overset{J}{\sum <br><br>}}<br><br><br><br>{\mathrm{\beta <br><br>}}_j<br><br>z_{\mathrm{ji}}\right)=\frac{\exp <br><br>\left({\mathrm{\beta <br><br>}}_0+\underset{j=1}{\overset{J}{\sum <br><br>}}<br><br><br><br>{\mathrm{\beta <br><br>}}_j<br><br>z_{\mathrm{ji}}\right)}{1+\exp <br><br>\left({\mathrm{\beta <br><br>}}_0+\underset{j=1}{\overset{J}{\sum <br><br>}}<br><br><br><br>{\mathrm{\beta <br><br>}}_j<br><br>z_{\mathrm{ji}}\right)}$

where

$F<br><br>\left({\mathrm{\beta <br><br>}}_0+\underset{i=1}{\overset{J}{\sum <br><br>}}<br><br><br><br>{\mathrm{\beta <br><br>}}_j<br><br>z_{\mathrm{ji}}\right)$

stands for the cumulative density function for a logistic distribution.

10. A method according to claim 9 further comprising the step of evaluating associations between variables within a matrix based on concordance according to

$C=\sum _{}^{}$

i = 1 I ⁢

⁢

∑

j = 1 J ⁢

⁢

X ij ⁡

[ ∑

k >

i ⁢

⁢

∑

l >

j ⁢

⁢

X kl + ∑

k <

i ⁢

⁢

∑

l <

j ⁢

⁢

X kl ]

where Xij stands for the number of observations for i^th row and j^th column, or disconcordance, when the subject rankings are higher on the row variable and lower on the column variable according to

$D=\underset{i=1}{\overset{I}{\sum <br><br>}}<br><br><br><br>\underset{j=1}{\overset{J}{\sum <br><br>}}<br><br><br><br>X_{\mathrm{ij}}<br><br>\left[\underset{k><br><br>i}{\sum <br><br>}<br><br><br><br>\underset{l<<br><br>j}{\sum <br><br>}<br><br><br><br>X_{\mathrm{kl}}+\underset{k<<br><br>i}{\sum <br><br>}<br><br><br><br>\underset{l<<br><br>j}{\sum <br><br>}<br><br><br><br>X_{\mathrm{kl}}\right]$

or when the subject rankings are tied on the row variable and the column variable according to

$T=\underset{i=1}{\overset{I}{\sum <br><br>}}<br><br><br><br>X_{\mathrm{ii}}<br><br>\left[\underset{k=j\ne <br><br>i}{\sum <br><br>}<br><br><br><br>X_{\mathrm{kk}}\right].$

11. A method according to claim 10 further comprising the step of ensuring that parameter estimates β

_k in the logistic regression model have comparable magnitudes for the different independent variables x_k, which have different units by standardizing all continuous variables according to;

$\tilde{X}=\frac{X-E<br><br>\left(X\right)}{\mathrm{\sigma <br><br>}<br><br>\left(X\right)}$

where E(X) is the mean of X and σ

(X) is the standard deviation of X.

12. A system configured for identifying and retaining customers who are likely to terminate an existing loan contract at a date earlier than the loan termination date identified in the loan contract, wherein the existing loan contract is provided by a lender that provides financing for purchasing a product by a customer from a dealer, said system comprising:

a server;

a network; and

at least one computer connected to said server via said network, said server configured to;

store customer data in a database including a payment history for each existing loan contract by the corresponding customer;

utilize an early termination model to identify for the lender customers likely to terminate an existing loan contract at a date earlier than the loan termination date identified in the loan contract, the early termination model uses the customer data stored within the database;

provide a list of early termination customers identified by the lender to the dealer, each customer identified by the lender on the list of early termination customers is a customer that satisfies the early termination model and is a customer the lender would like to retain as a customer;

notify the customers identified on the early termination list by the dealer of new products and new loan opportunities at competitive loan interest rates, wherein the new loan opportunities are provided by the lender to finance a future purchase of new products from the dealer by the customers identified on the early termination list; and

prompt the dealer to cross sell the new loans to the customers identified in the early termination list before the identified customers terminate the existing loan contracts associated therewith.

13. A system according to claim 12 wherein said server configured to utilize loan performance and demographic data to predict likely loan terminations at least three months before the customer is likely to prepay the existing loan.

14. A system according to claim 12 wherein said sewer configured to:

predict a probability of early termination at every stage of the lifetime of the existing loan;

combine a probability of early termination with financial calculations to determine an expected profit at each stage of the existing loan;

compare expected profit from the existing loan with an expected profit from a new loan; and

determine which customers to target for a new loan based upon the expected profits from the existing and the new loans.

15. A system according to claim 12 wherein said sewer is configured to predict a probability of early termination using probability properties P(A₁∪

A₂∪

A₃)=1−

P((A₁∪

A₂∪

A₃^c))=1−

P(A₁^c∩

A₂^c∩

A₃^c)=1−

P(A₁^c)P(A₂^c|A₁^c)P(A₃^c|A₁^c∩

A₂^c), where the events A₁^c, A₂^c|A₁^c, A₃^c|A₁^c∩

A₂^c represent the probabilities that the customer has not early terminated.

16. A system according to claim 15 wherein said server is configured to calculate an expected income, given a set of probabilities according to:

$\begin{array}{c}E<br><br>\left(\mathrm{Income}\right)=\left(1-P_1\right)<br><br>I_1+\left(1-P_1\right)<br><br>\left(1-P_2\right)<br><br>I_2+<br><br>...<br><br>+\left(1-P_1\right)<br><br>...\left(1-P_T\right)<br><br>I_T=\\ =\underset{i=1}{\overset{T}{\sum <br><br>}}<br><br><br><br>\left[\underset{k=1}{\overset{i}{\prod <br><br>}}<br><br><br><br>\left(1-P_k\right)\right]*I_i.\end{array}<br><br>$

17. A system according to claim 16 wherein said server is configured to calculate a monthly internal rate of return for each customer according to:

$\underset{t=1}{\overset{T}{\sum <br><br>}}<br><br><br><br>\frac{C_t}{{\left(1+I_{\mathrm{RR}}\right)}^t}=0.$

18. A system according to claim 12 wherein said server configured to compile customer data from multiple sources including at least one of world wide web, legacy customer data, facsimile, telephone and cellular data.

19. A system according to claim 12 wherein said server configured to build a regression model to test accuracy and a strength of association between variables which describe customers and their loans for an i^th experimental unit according to:

${\prod <br><br>}_i<br><br>=F<br><br>\left({\mathrm{\beta <br><br>}}_0+\underset{j=1}{\overset{J}{\sum <br><br>}}<br><br><br><br>{\mathrm{\beta <br><br>}}_j<br><br>z_{\mathrm{ji}}\right)=\frac{\exp <br><br>\left({\mathrm{\beta <br><br>}}_0+\underset{j=1}{\overset{J}{\sum <br><br>}}<br><br><br><br>{\mathrm{\beta <br><br>}}_j<br><br>z_{\mathrm{ji}}\right)}{1+\exp <br><br>\left({\mathrm{\beta <br><br>}}_0+\underset{j=1}{\overset{J}{\sum <br><br>}}<br><br><br><br>{\mathrm{\beta <br><br>}}_j<br><br>z_{\mathrm{ji}}\right)}$

where

$F<br><br>\left({\mathrm{\beta <br><br>}}_0+\underset{j=1}{\overset{J}{\sum <br><br>}}<br><br><br><br>{\mathrm{\beta <br><br>}}_j<br><br>z_{\mathrm{ji}}\right)$

where stands for the cumulative density function for a logistic distribution.

20. A system according to claim 19 wherein said server configured to evaluate associations between variables within a matrix based on concordance according to

$C=\sum _{}^{}$

i = 1 I ⁢

⁢

∑

j = 1 J ⁢

⁢

X ij ⁡

[ ∑

k >

i ⁢

⁢

∑

l >

j ⁢

⁢

X kl + ∑

k <

i ⁢

⁢

∑

l <

j ⁢

⁢

X kl ]

where Xij stands for the number of observations for i^th row and j^th column, or disconcordance, when the subject rankings are higher on the row variable and lower on the column variable according to

$D=\underset{i=1}{\overset{I}{\sum <br><br>}}<br><br><br><br>\underset{j=1}{\overset{J}{\sum <br><br>}}<br><br><br><br>X_{\mathrm{ij}}<br><br>\left[\underset{k><br><br>i}{\sum <br><br>}<br><br><br><br>\underset{l<<br><br>j}{\sum <br><br>}<br><br><br><br>X_{\mathrm{kl}}+\underset{k<<br><br>i}{\sum <br><br>}<br><br><br><br>\underset{l><br><br>j}{\sum <br><br>}<br><br><br><br>X_{\mathrm{kl}}\right]$

or when the subject rankings are tied on the row variable and the column variable according to

$T=\underset{i=1}{\overset{I}{\sum <br><br>}}<br><br>X_{\mathrm{ii}}<br><br>\left[\underset{k=j\ne <br><br>i}{\sum <br><br>}<br><br><br><br>X_{\mathrm{kk}}\right].$

21. A system according to claim 20 wherein said server configured to ensure that parameter estimates β

_k in the logistic regression model have comparable magnitudes for the different independent variables x_k, which have different units by standardizing all continuous variables according to

$\tilde{X}=\frac{X-E<br><br>\left(X\right)}{\mathrm{\sigma <br><br>}<br><br>\left(X\right)}$

where E(X) is the mean of X and σ

(X) is the standard deviation of X.

22. A computer for identifying and retaining customers who are likely to terminate an existing loan contract at a date earlier than the loan termination date identified in the loan contract, wherein the existing loan contract is provided by a lender that provides financing for purchasing a product by a customer from a dealer, said computer programmed to:

store customer data in a database including a payment history for each existing loan contract by the corresponding customer;

prompt the lender to start an analysis of the customer data stored in the database including accessing an early termination model to identify for the lender customers likely to terminate an existing loan contract at a date earlier than the loan termination date identified in the loan contract;

provide a list of early termination customers identified by the lender to the dealer, each customer identified by the lender on the list of early termination customers is a customer that satisfies the early termination model and is a customer the lender would like to retain as a customer; and

prompt the dealer to notify the customers identified on the early termination list of new products and new loan opportunities, wherein the new loan opportunities are provided by the lender to finance a future purchase of new products from the dealer by the customer identified on the early termination list.

23. A computer according to claim 22 programmed to utilize loan performance and demographic data to predict which loan customers are likely to prepay an existing loan at least three months before the predicted loan prepayment.

24. A computer according to claim 22 programmed to:

prompt a user with a prediction of a probability of early termination at every stage of the lifetime of the existing loan;

combine a probability of early termination with financial calculations to determine an expected profit at each stage of the existing loan;

compare expected profit from the existing loan with an expected profit from a new loan; and

prompt a user with a determination of customers to target for new loans based upon calculations of expected profits from the existing and the new loans.

25. A computer according to claim 22 programmed to predict a probability of early termination using probability properties P(A₁∪

A₂∪

A₃)=1−

P((A₁∪

A₂∪

A₃^c))=1−

P(A₁^c∩

A₂^c∩

A₃^c)=1−

P(A₁^c)P(A₂^c|A₁^c)P(A₃^c|A₁^c∩

A₂^c), where the events A₁^c, A₂^c|A₁^c, A₃^c|A₁^c∩

A₂^c represent the probabilities that the customer has not prepaid the loan.

26. A computer according to claim 25 programmed to calculate an expected income, given a set of probabilities according to

$\begin{array}{c}E\end{array}$

( Income ) = ( 1 - P 1 ) ⁢

I 1 + ( 1 - P 1 ) ⁢

( 1 - P 2 ) ⁢

I 2 + ⁢

. . . ⁢

+ ( 1 - P 1 ) ⁢

. . . = ∑

i = 1 T ⁢

⁢

[ ∏

k = 1 i ⁢

⁢

( 1 - P k ) ] * I i . ⁢

( 1 - P T ) ⁢

I T =

27. A computer according to claim 26 programmed to calculate a monthly internal rate of return for each customer according to:

$\underset{t=1}{\overset{T}{\sum <br><br>}}<br><br><br><br>\frac{C_t}{{\left(1+I_{\mathrm{RR}}\right)}^t}=0.$

28. A computer according to claim 22 programmed to test accuracy and strength of association between customers and their loans by building a regression model for an i^th experimental unit according to:

${\prod <br><br>}_i<br><br>=F<br><br>\left({\mathrm{\beta <br><br>}}_0+\underset{j=1}{\overset{J}{\sum <br><br>}}<br><br><br><br>{\mathrm{\beta <br><br>}}_j<br><br>z_{\mathrm{ji}}\right)=\frac{\exp <br><br>\left({\mathrm{\beta <br><br>}}_0+\underset{j=1}{\overset{J}{\sum <br><br>}}<br><br><br><br>{\mathrm{\beta <br><br>}}_j<br><br>z_{\mathrm{ji}}\right)}{1+\exp <br><br>\left({\mathrm{\beta <br><br>}}_0+\underset{j=1}{\overset{J}{\sum <br><br>}}<br><br><br><br>{\mathrm{\beta <br><br>}}_j<br><br>z_{\mathrm{ji}}\right)}$

where

$F<br><br>\left({\mathrm{\beta <br><br>}}_0+\underset{j=1}{\overset{J}{\sum <br><br>}}<br><br><br><br>{\mathrm{\beta <br><br>}}_j<br><br>z_{\mathrm{ji}}\right)$

stands for the cumulative density function for a logistic distribution.

29. A computer according to claim 28 programmed to evaluate associations between variables within a matrix based on concordance according to

$C=\sum _{}^{}$

i = 1 I ⁢

⁢

∑

j = 1 J ⁢

⁢

X ij ⁡

[ ∑

k >

i ⁢

⁢

∑

l >

j ⁢

⁢

X kl + ∑

k <

i ⁢

⁢

∑

l <

j ⁢

⁢

X kl ]

where Xzj stands for the number of observations for i^th row and j^th column, or disconcordance, when the subject rankings are higher on the row variable and lower on the column variable according to

$D=\underset{i=1}{\overset{I}{\sum <br><br>}}<br><br><br><br>\underset{j=1}{\overset{J}{\sum <br><br>}}<br><br><br><br>X_{\mathrm{ij}}<br><br>\left[\underset{k><br><br>i}{\sum <br><br>}<br><br><br><br>\underset{l<<br><br>j}{\sum <br><br>}<br><br><br><br>X_{\mathrm{kl}}+\underset{k<<br><br>i}{\sum <br><br>}<br><br><br><br>\underset{l><br><br>j}{\sum <br><br>}<br><br><br><br>X_{\mathrm{kl}}\right]$

or when the subject rankings are tied on the row variable and the column variable according to

$T=\underset{i=1}{\overset{I}{\sum <br><br>}}<br><br><br><br>X_{\mathrm{ii}}<br><br>\left[\underset{k=j\ne <br><br>i}{\sum <br><br>}<br><br>X_{\mathrm{kk}}\right].$

30. A computer according to claim 29 programmed to ensure that parameter estimates β

_k in the logistic regression model have comparable magnitudes for the different independent variables x_k, which have different units by standardizing all continuous variables according to

$\tilde{X}=\frac{X-E<br><br>\left(X\right)}{\mathrm{\sigma <br><br>}<br><br>\left(X\right)}$

where E(X) is the mean of X and σ

(X) is the standard deviation of X.

31. A computer readable medium for identifying and retaining customers who are likely to terminate an existing loan contract at a date earlier than the loan termination date identified in the loan contract, wherein the existing loan contract is provided by a lender that provides financing for purchasing a product by a customer from a dealer, said computer readable medium comprising:

a record of customer data including a payment history for each existing loan contract by the corresponding customer;

a plurality of rules including an early termination model for identifying for the lender customers likely to terminate an existing loan contract at a date earlier than the loan termination date identified in the loan contract, the early termination model uses the record of customer data;

a record of a list of early termination customers identified by the lender to the dealer, each customer identified by the lender on the list of early termination customers is a customer that satisfies the early termination model and is a customer the lender would like to retain as a customer; and

a record of customers notified by the dealer of new products and new loan opportunities, the customers notified including the customers on the early termination list, wherein the new loan opportunities are provided by the lender to finance a future purchase of new products from the dealer by the customers identified on the early termination list.

32. A computer readable medium according to claim 31 wherein said plurality of rules identify customers likely to early terminate at least three months before the customer is likely to terminate the loan contract.

33. A computer readable medium according to claim 31 wherein said plurality of rules comprises:

rules to predict a probability of early termination at every stage of an existing loan;

rules to combine the probability of early termination with financial calculations to determine an expected profit at every stage of the loan; and

rules to compare expected profit from the existing loan to an expected profit from a new loan.

34. A computer readable medium according to claim 33 further comprising rules to identify which customers to target for a new loan based upon expected profits form the existing and the new loans.

35. A computer readable medium according to claim 31 wherein said plurality of rules comprises rules to cluster groups of customers according to customer characteristics.

36. A computer readable medium according to claim 31 wherein said plurality of rules comprises rules to measure a strength of association between variables which describe customers and their loans.

37. A method for identifying customers likely to terminate an existing loan contract at a date earlier than the loan termination date identified in the loan contract, wherein the existing loan contract is provided by a lender that provides financing for purchasing a product by a customer from a dealer, said method comprising the steps of:

selecting, from an electronic interface, a record of customer information, wherein the customer information includes a payment history for each existing loan contract by the corresponding customer;

selecting by the lender, from the electronic interface, an early termination model for an analysis of the selected customer information;

requesting by the lender, from the electronic interface, a list of customers identified by the model as being likely to early terminate corresponding existing loan contracts, each customer identified by the lender on the list of early termination customers is a customer that satisfies the early termination model and is a customer the lender would like to retain as a customer; and

requesting by the dealer, from the electronic interface, a notification provided to the customers identified on the early termination list of new products and new loan opportunities, wherein the new loan opportunities are provided by the lender to finance a future purchase of new products from the dealer by the customers identified on the early termination list.

38. A method according to claim 37 further comprising the step of requesting by the dealer, from the electronic interface, which of the identified customers have been notified of new loan opportunities.

39. A method according to claim 37 wherein said step of selecting a record of customer information further comprises the step of selecting a source of customer information from at least one of world wide web, legacy customer data, facsimile, telephone and cellular data.

40. Apparatus for identifying and retaining customers who are likely to terminate an existing loan contract at a date earlier than the loan termination date identified in the loan contract, wherein the existing loan contract is provided by a lender that provides financing for purchasing a product by a customer from a dealer, said apparatus comprising:

means for storing customer information including a payment history for each existing loan contract by the corresponding customer;

means for identifying for the lender customers likely to terminate an existing loan contract at a date earlier than the loan termination date identified in the loan contract; and

means for providing to the dealer a list of early termination customers identified by the lender;

means for prompting the dealer to notify the customers identified on the early termination list of new products and new loan opportunities, wherein the new loan opportunities are provided by the lender to finance a future purchase of new products from the dealer by the customers identified on the early termination list.

41. Apparatus according to claim 40 further comprising means for notifying customers likely to early terminate of new loan opportunities.

42. Apparatus according to claim 40 wherein said means for identifying for the lender customers likely to terminate an existing loan contract comprises means for using loan performance and demographic data to identify customers likely to early terminate at least three months before the customer is likely to terminate the existing loan.

43. Apparatus according to claim 40 wherein said means for identifying for the lender customers likely to terminate an existing loan contract comprises:

means for predicting a probability of early termination at every stage of the lifetime of the existing loan;

means for combining the probability of early termination with financial calculations to determine an expected profit at each stage of the existing loan;

means for comparing expected profit from the existing loan with an expected profit from a new loan; and

means for determining which customers to target for a new loan based upon the expected profits from the existing and the new loans.

44. Apparatus according to claim 40 wherein said means for identifying for the lender customers likely to early terminate comprises means for using a cluster analysis to segment customers and identify sales opportunities for both active and inactive customers.

45. Apparatus according to claim 40 wherein said means for identifying for the lender customers likely to early terminate comprises means for measuring a strength of association between variables which describe customers and their loans.

46. Apparatus according to claim 40 wherein said means for identifying for the lender customers likely to early terminate comprises means for clustering groups of customers according to characteristics of the customers.