U.S. patent application number 14/280925 was filed with the patent office on 2015-12-31 for intelligent collections models.
The applicant listed for this patent is Accenture Global Services Limited. Invention is credited to Indeer Preet Singh.
Application Number | 20150379631 14/280925 |
Document ID | / |
Family ID | 39493302 |
Filed Date | 2015-12-31 |
![](/patent/app/20150379631/US20150379631A1-20151231-D00000.png)
![](/patent/app/20150379631/US20150379631A1-20151231-D00001.png)
![](/patent/app/20150379631/US20150379631A1-20151231-D00002.png)
![](/patent/app/20150379631/US20150379631A1-20151231-D00003.png)
![](/patent/app/20150379631/US20150379631A1-20151231-D00004.png)
![](/patent/app/20150379631/US20150379631A1-20151231-D00005.png)
![](/patent/app/20150379631/US20150379631A1-20151231-D00006.png)
![](/patent/app/20150379631/US20150379631A1-20151231-D00007.png)
![](/patent/app/20150379631/US20150379631A1-20151231-D00008.png)
![](/patent/app/20150379631/US20150379631A1-20151231-D00009.png)
![](/patent/app/20150379631/US20150379631A1-20151231-D00010.png)
View All Diagrams
United States Patent
Application |
20150379631 |
Kind Code |
A1 |
Singh; Indeer Preet |
December 31, 2015 |
INTELLIGENT COLLECTIONS MODELS
Abstract
Apparatuses, computer media, and methods for analyzing credit
and tax form data and determining a collection treatment type to
collect revenue. A collections model is constructed to determine a
collections score that is based on raw credit data and tax form
data and is indicative of a debtor's propensity to pay an owed
amount. The collections model includes score bands, each score band
being associated with a range of credit scores. A collections score
is determined from a scoring expression that is associated with a
score band and that typically includes a subset of available raw
credit data and tax form data. A collections treatment type is
determined from a collections score. Each treatment type
corresponds to a treatment action that is directed to the debtor. A
collections model is constructed from historical tax data, in which
score bands and scoring expressions are constructed for the
collections model.
Inventors: |
Singh; Indeer Preet;
(Princeton Junction, NJ) |
|
Applicant: |
Name |
City |
State |
Country |
Type |
Accenture Global Services Limited |
Dublin |
|
IE |
|
|
Family ID: |
39493302 |
Appl. No.: |
14/280925 |
Filed: |
May 19, 2014 |
Related U.S. Patent Documents
|
|
|
|
|
|
Application
Number |
Filing Date |
Patent Number |
|
|
13655747 |
Oct 19, 2012 |
8732074 |
|
|
14280925 |
|
|
|
|
13419248 |
Mar 13, 2012 |
8306909 |
|
|
13655747 |
|
|
|
|
12900839 |
Oct 8, 2010 |
8135643 |
|
|
13419248 |
|
|
|
|
11566787 |
Dec 5, 2006 |
7827100 |
|
|
12900839 |
|
|
|
|
Current U.S.
Class: |
705/38 |
Current CPC
Class: |
G06Q 40/025 20130101;
G06Q 40/00 20130101; G06Q 40/02 20130101 |
International
Class: |
G06Q 40/02 20060101
G06Q040/02 |
Claims
1. (canceled)
2. A computer-implemented method comprising: generating training
data for training a predictive model to estimate a likelihood of a
debtor to pay a debt, the training data including, for each of a
plurality of debtors, (i) historic tax return data for the debtor,
(ii) commercial financial or credit data that is usable to
calculate a credit score for the debtor, and (iii) a label
indicating a likelihood of the debtor to pay a debt; training the
predictive model using the historic tax return data, the commercial
financial or credit data, and the labels included in the training
data; after training the predictive model, identifying a particular
debtor; obtaining (i) historic tax return data for the particular
debtor, and (ii) commercial financial or credit data that is usable
to calculate a credit score for the particular debtor; providing,
to the predictive model, (i) the historic tax return data for the
particular debtor, and (ii) the commercial financial or credit data
that is usable to calculate a credit score for the particular
debtor; and obtaining, from the predictive model, an indication of
a likelihood of the particular debtor to pay a debt.
3. The method of claim 2, wherein the historic tax return data
comprises one or more values were entered on a personal income tax
form by an associated debtor.
4. The method of claim 2, wherein the debt comprises a debt owed to
a government revenue agency.
5. The method of claim 2, wherein the historic tax return data
comprises data that is available only to a government revenue
agency or to designated agents of the government revenue
agency.
6. The method of claim 2, wherein the historic tax return data
comprises data that is not reflected on a credit report of an
associated debtor.
7. The method of claim 2, wherein the historic tax return data
comprises one or more values that reflect an amount of tax listed
as due on a tax return in relation to an amount of income listed on
the tax return.
8. The method of claim 2, wherein the historic tax return data
comprises one or more values that characterize a status of a
previous year's tax return.
9. The method of claim 2, wherein the historic tax return data
comprises one or more values that reflect an amount of tax listed
due on a tax return in relation to an amount of tax listed as owed
on the tax return.
10. The method of claim 2, wherein the historic tax return data
comprises one or move values that reflect a tax penalty amount
listed on a tax return.
11. The method of claim 2, wherein the historic tax return data
comprises one or more values that reflect an amount of time after a
tax deadline in which a tax return was filed.
12. The method of claim 2, wherein the historic tax return data
comprises one or more values that reflect an amount of tax that a
tax return lists as owed.
13. A computer-readable storage device encoded with a computer
program, the program comprising instructions that, if executed by
one or more computers, cause the one or more computers to perform
operations comprising: generating training data for training a
predictive model to estimate a likelihood of a debtor to pay a
debt, the training data including, for each of a plurality of
debtors, (i) historic tax return data for the debtor, (ii)
commercial financial or credit data that is usable to calculate a
credit score for the debtor, and (iii) a label indicating a
likelihood of the debtor to pay a debt; and training the predictive
model using the historic tax return data, the commercial financial
or credit data, and the labels included in the training data.
14. The device of claim 13, wherein the historic tax return data
comprises one or more values were entered on a personal income tax
form by an associated debtor.
15. The device of claim 13, wherein the debt comprises a debt owed
to a government revenue agency.
16. The device of claim 13, wherein the historic tax return data
comprises data that is available only to a government revenue
agency or to designated agents of the government revenue
agency.
17. The device of claim 13, wherein the historic tax return data
comprises one or more values that reflect an amount of tax listed
as due on a tax return in relation to an amount of income listed on
the tax return.
18. The device of claim 13, wherein the historic tax return data
comprises one or more values that reflect an amount of tax listed
due on a tax return in relation to an amount of tax listed as owed
on the tax return.
19. The device of claim 13, wherein the historic tax return data
comprises one or move values that reflect a tax penalty amount
listed on a tax return.
20. The device of claim 13, wherein the historic tax return data
comprises one or more values that reflect an amount of time after a
tax deadline in which a tax return was filed.
21. A system comprising: a processor configured to executed
computer program instructions; and a computer storage medium
encoded with computer program instructions that, when executed by
the processor, cause the system to perform operations comprising:
obtaining a predictive model that is trained to estimate a
likelihood of a debtor to pay a debt, wherein the predictive model
is trained using training data that includes, for each of a
plurality of debtors, (i) historic tax return data for the debtor,
(ii) commercial financial or credit data that is usable to
calculate a credit score for the debtor, and (iii) a label
indicating a likelihood of the debtor to pay a debt; identifying a
particular debtor; obtaining (i) historic tax return data for the
particular debtor, and (ii) commercial financial or credit data
that is usable to calculate a credit score for the particular
debtor; providing, to the predictive model, (i) the historic tax
return data for the particular debtor, and (ii) the commercial
financial or credit data that is usable to calculate a credit score
for the particular debtor; and obtaining, from the predictive
model, an indication of a likelihood of the particular debtor to
pay a debt.
Description
FIELD OF THE INVENTION
[0001] This application is a continuation application of U.S.
patent application Ser. No. 13/655,747 filed Oct. 19, 2012, naming
Inder Preet Singh as the inventor, which is a continuation
application of U.S. patent application Ser. No. 13/419,248 filed
Mar. 13, 2012, naming Inder Preet Singh as the inventor, which is a
divisional of U.S. patent application Ser. No. 12/900,839, filed
Oct. 8, 2010, naming Inder Preet Singh as the inventor, which is
itself a divisional of U.S. patent application Ser. No. 11/566,787,
filed Dec. 5, 2006, naming Inder Preet Singh as the inventor. These
applications are incorporated herein by reference in their entirety
and for all purposes.
BACKGROUND OF THE INVENTION
[0002] Revenue agencies typically have more accounts to be
collected than resources to collect and resolve the accounts.
Historically revenue agencies work all accounts through a single,
inflexible workflow with little consideration to the debtor's
willingness or ability to pay. Decisions to use outside collections
services occur at the end of the process at which time the accounts
are stale.
[0003] A revenue agency typically utilizes a FICO score, which is a
credit score developed by Fair Isaac & Co. Credit scoring and
is a method for determining the likelihood that credit users will
pay their bills. Fair, Isaac began its pioneering work with credit
scoring in the late 1950s and, since then, scoring has become
widely accepted by lenders as a reliable means of credit
evaluation. A credit score attempts to condense a borrower's credit
history into a single number. However, Fair, Isaac & Co. and
the credit bureaus do not reveal how the credit scores are
computed. The Federal Trade Commission has ruled this approach to
be acceptable. Credit scores are calculated by using scoring models
and mathematical tables that assign points for different pieces of
information which best predict future credit performance.
Developing these models involves studying how thousands, even
millions, of people have used credit. Score-model developers find
predictive factors in the data that have proven to indicate future
credit performance. Models can be developed from different sources
of data. Credit-bureau models are developed from information in
consumer credit-bureau reports.
[0004] Credit scores analyze a borrower's credit history
considering numerous factors such as:
[0005] Late payments
[0006] The amount of time credit has been established
[0007] The amount of credit used versus the amount of credit
available
[0008] Length of time at present residence
[0009] Employment history
[0010] Negative credit information such as bankruptcies,
charge-offs, collections, etc.
There are typically three FICO scores that are computed by data
provided by each of the three most prevalent credit bureaus:
Experian, TransUnion, and Equifax. Some lenders use one of these
three scores, while other lenders may use the middle score.
[0011] The use of a credit score to determine the propensity to pay
is inflexible in altering the collections model. A revenue agency,
for example, may wish to tailor its collection model to better fit
available data. Moreover, a revenue agency can customize its
collection practices to more effectively use collections resources
and to identify those accounts that will require private
collections services early in the process.
BRIEF SUMMARY OF THE INVENTION
[0012] Embodiments of invention provide apparatuses, computer
media, and methods for analyzing raw credit data and tax form data
to determine a collections score that is indicative of debtor's
(tax filer's) propensity to pay an owed amount to a revenue
agency.
[0013] With one aspect of the invention, a collections model is
formed from raw credit data, tax form data, and credit scores. The
collections model includes a plurality of score bands, in which a
score band is associated with range of credit scores.
[0014] With another aspect of the invention, a collections score is
determined from a scoring expression that is associated with each
score band. The scoring expression typically includes a subset of
available raw credit data and tax form data. A scoring expression
that is associated with a score band may utilize different
variables than another scoring expression that is associated with
another score band.
[0015] With another aspect of the invention, a collections
treatment type for a debtor is determined from a collections score.
The collections treatment type may be independent of the score band
of the debtor. Each collections treatment type corresponds to a
treatment action that is directed to the debtor. Moreover, the
collections treatment type for a given collections score range may
be modified if the revenue agency wishes to alter the collections
model.
[0016] With another aspect of the invention, a collections model is
constructed from historical tax data. A plurality of score bands is
constructed for the collections model, where a different scoring
expression is associated with each score band.
BRIEF DESCRIPTION OF THE DRAWINGS
[0017] The present invention is illustrated by way of example and
not limited in the accompanying figures in which like reference
numerals indicate similar elements and in which:
[0018] FIG. 1 shows an architecture of a computer system used in a
multi-lingual telephonic service in accordance with an embodiment
of the invention.
[0019] FIG. 2 shows a process for modeling revenue collections in
accordance with an embodiment of the invention.
[0020] FIG. 3 illustrates a process for assigning a debtor to a
score band in accordance with an embodiment of the invention.
[0021] FIG. 4 shows variables for scoring in a first score band in
accordance with an embodiment of the invention.
[0022] FIG. 5 shows variables for scoring in a second score band in
accordance with an embodiment of the invention.
[0023] FIG. 6 shows variables for scoring in a third score band in
accordance with an embodiment of the invention.
[0024] FIG. 7 shows variables for scoring in a fourth score band in
accordance with an embodiment of the invention.
[0025] FIG. 8 shows variables for scoring in a fifth score band in
accordance with an embodiment of the invention.
[0026] FIG. 9 shows variables for scoring in a sixth score band in
accordance with an embodiment of the invention.
[0027] FIG. 10 shows a process for determining a collections score
for a debtor in accordance with an embodiment of the invention.
[0028] FIG. 11 shows a process for determining a collections
treatment type from a collections score in accordance with an
embodiment of the invention.
[0029] FIG. 12 shows an apparatus that analyzes raw credit data and
tax form data to initiate a collections treatment action in
accordance with an embodiment of the invention.
DETAILED DESCRIPTION OF THE INVENTION
[0030] Elements of the present invention may be implemented with
computer systems, such as the system 100 shown in FIG. 1. Computer
100 may be incorporated in an apparatus (as shown in FIG. 12) that
analyzes input data and consequently initiates a collections
treatment action for collecting revenues. Computer 100 includes a
central processor 110, a system memory 112 and a system bus 114
that couples various system components including the system memory
112 to the central processor unit 110. System bus 114 may be any of
several types of bus structures including a memory bus or memory
controller, a peripheral bus, and a local bus using any of a
variety of bus architectures. The structure of system memory 112 is
well known to those skilled in the art and may include a basic
input/output system (BIOS) stored in a read only memory (ROM) and
one or more program modules such as operating systems, application
programs and program data stored in random access memory (RAM).
[0031] Computer 100 may also include a variety of interface units
and drives for reading and writing data. In particular, computer
100 includes a hard disk interface 116 and a removable memory
interface 120 respectively coupling a hard disk drive 118 and a
removable memory drive 122 to system bus 114. Examples of removable
memory drives include magnetic disk drives and optical disk drives.
The drives and their associated computer-readable media, such as a
floppy disk 124 provide nonvolatile storage of computer readable
instructions, data structures, program modules and other data for
computer 100. A single hard disk drive 118 and a single removable
memory drive 122 are shown for illustration purposes only and with
the understanding that computer 100 may include several of such
drives. Furthermore, computer 100 may include drives for
interfacing with other types of computer readable media.
[0032] A user can interact with computer 100 with a variety of
input devices. FIG. 1 shows a serial port interface 126 coupling a
keyboard 128 and a pointing device 130 to system bus 114. Pointing
device 128 may be implemented with a mouse, track ball, pen device,
or similar device. Of course one or more other input devices (not
shown) such as a joystick, game pad, satellite dish, scanner, touch
sensitive screen or the like may be connected to computer 100.
[0033] Computer 100 may include additional interfaces for
connecting devices to system bus 114. FIG. 1 shows a universal
serial bus (USB) interface 132 coupling a video or digital camera
134 to system bus 114. An IEEE 1394 interface 136 may be used to
couple additional devices to computer 100. Furthermore, interface
136 may configured to operate with particular manufacture
interfaces such as FireWire developed by Apple Computer and i.Link
developed by Sony. Input devices may also be coupled to system bus
114 through a parallel port, a game port, a PCI board or any other
interface used to couple and input device to a computer.
[0034] Computer 100 also includes a video adapter 140 coupling a
display device 142 to system bus 114. Display device 142 may
include a cathode ray tube (CRT), liquid crystal display (LCD),
field emission display (FED), plasma display or any other device
that produces an image that is viewable by the user. Additional
output devices, such as a printing device (not shown), may be
connected to computer 100.
[0035] Sound can be recorded and reproduced with a microphone 144
and a speaker 146. A sound card 148 may be used to couple
microphone 144 and speaker 146 to system bus 114. One skilled in
the art will appreciate that the device connections shown in FIG. 1
are for illustration purposes only and that several of the
peripheral devices could be coupled to system bus 114 via
alternative interfaces. For example, video camera 134 could be
connected to IEEE 1394 interface 136 and pointing device 130 could
be connected to USB interface 132.
[0036] Computer 100 can operate in a networked environment using
logical connections to one or more remote computers or other
devices, such as a server, a router, a network personal computer, a
peer device or other common network node, a wireless telephone or
wireless personal digital assistant. Computer 100 includes a
network interface 150 that couples system bus 114 to a local area
network (LAN) 152. Networking environments are commonplace in
offices, enterprise-wide computer networks and home computer
systems.
[0037] A wide area network (WAN) 154, such as the Internet, can
also be accessed by computer 100. FIG. 1 shows a modem unit 156
connected to serial port interface 126 and to WAN 154. Modem unit
156 may be located within or external to computer 100 and may be
any type of conventional modem such as a cable modem or a satellite
modem. LAN 152 may also be used to connect to WAN 154. FIG. 1 shows
a router 158 that may connect LAN 152 to WAN 154 in a conventional
manner.
[0038] It will be appreciated that the network connections shown
are exemplary and other ways of establishing a communications link
between the computers can be used. The existence of any of various
well-known protocols, such as TCP/IP, Frame Relay, Ethernet, FTP,
HTTP and the like, is presumed, and computer 100 can be operated in
a client-server configuration to permit a user to retrieve web
pages from a web-based server. Furthermore, any of various
conventional web browsers can be used to display and manipulate
data on web pages.
[0039] The operation of computer 100 can be controlled by a variety
of different program modules. Examples of program modules are
routines, programs, objects, components, data structures, etc.,
that perform particular tasks or implement particular abstract data
types. The present invention may also be practiced with other
computer system configurations, including hand-held devices,
multiprocessor systems, microprocessor-based or programmable
consumer electronics, network PCS, minicomputers, mainframe
computers, personal digital assistants and the like. Furthermore,
the invention may also be practiced in distributed computing
environments where tasks are performed by remote processing devices
that are linked through a communications network. In a distributed
computing environment, program modules may be located in both local
and remote memory storage devices.
[0040] An embodiment of the invention supports the development of
unique analytic models to score debtors(i.e., tax filers who owe
money) with outstanding accounts receivable that are owed to
government revenue agencies. The scores generated by the
collections model represent the propensity of a debtor to pay and
also provide insight into the level of effort that will be required
to collect the debt by the revenue agency. Collection models may
blend demographic and financial information maintained by the
revenue agency with commercial data that is reflective of a
debtor's ability to pay and credit history. While a revenue agency
is typically a governmental organization, revenue collections can
be performed by a private organization that has been contracted by
a government (Federal, state, or local). In such a case, required
tax and credit information is made available to the private agency
with proper security measures.
[0041] With the prior art, collections models for revenue agencies
typically use only internal revenue agency data. With an embodiment
of the invention, collections models involve the blending of the
internal revenue agency data and the use of commercial financial
and credit data. The final collections model may provide a
significant improvement in identifying receivables that debtors are
more likely to pay during the collections process. The final
collections model is typically more predictive compared to
FICO-only model as well as tax data-only model. Both tax form data
and credit data are often very predictive in explaining payment
behavior. Those who have good credit history are also good tax
payers. For example, the ratio of tax still owed and income
(corresponding to ratio_taxowed_ctincome as will be discussed) is a
predictive tax variable--those with higher ratio are less likely to
pay.
[0042] FIG. 2 shows process 200 for modeling revenue collections in
accordance with an embodiment of the invention. Process 200
demonstrates quantitative benefits of using a collections model for
prioritizing receivable cases. A collections model is built from
developed datasets. With an embodiment of the invention, process
200 provides a test-deploy collections model as a proof-of-concept
for developing a business case for a state government.
[0043] With module 201, client customer data is blended with credit
history data and other data as required to fulfill the specific
requirements of a collections model. In an embodiment of the
invention, Module 201 extracts historical individual tax data for
the State of Connecticut (CT) in the 2002 and 2003 tax years.
Payment behavior is primarily modeled on 2003 tax data to predict
payment in the 2003 year. Prior tax year's (2002) Paid/Not-Paid
flag is also used for additional predictive power.
[0044] These data are combined in a database record called the
Customer Analytic Record (CAR) by module 203. U.S. Pat. No.
7,047,251 and U.S. application Ser. No. 11/147,034, to Kenneth L.
Reed, et al., ('251 and '034, respectively) are incorporated herein
by reference. The '251 and '034 references disclose a system and
method for creating virtual flat customer records derived from
database customer data that may be used as standardized input for
analytical models. A Customer Analytic Record (`CAR") application
may be created as a database object to extract, transform, and
format the customer data needed for customer segmentation and
predictive modeling. The CAR may be a set of database "views" that
are defined using virtual stored queries and enabled using
capabilities of a data base management system and a structured
query language. The CAR is typically a virtual `flat" record of the
customer data needed for customer analytics. The customer data may
be extracted by running one or more SQL queries against the
database view(s). The CAR application may dynamically calculate
additional variables using predetermined transformations, including
custom transformations of an underlying behavior. If additional
variables are created, the CAR may be modified to include the
additional variables. The CAR is often a dynamic view of the
customer record that changes whenever any update is made to the
database. The definition of the CAR provides documentation of each
data element available for use in models and analytics.
[0045] Module 203 creates a CAR table that is used as model input
data set to drive the modeling effort. (With an embodiment, module
203 determined tax-filers who owed $50 or more on the cutoff date.
The tax filers who owed less than $50 were dropped to provide
sharper contrast.) Module 203 rolls up (accumulates) transactional
tax data for the identified tax filers (e.g., until the cutoff date
of Jul. 15, 2004) to one record per tax filer and creates derived
variables-like ratios. Inferred "Goods" (Payers) correspond to tax
filers who paid in a performance window of 9 months and "Bads"
(Non-Payers) correspond to tax filers who did not pay in the
performance window. Module 203 appends credit attributes to each
record. (In an embodiment, more than 850 credit attributes provided
by TransUnion were appended, in which TransUnion was able to match
98% of names for credit data.)
[0046] Module 205 provides address hygiene on the historical tax
data (e.g., for the years 2002 and 2003) so that latest and correct
address information is associated with the names of tax payers. In
an embodiment of the invention, a data provider e.g., Acxiom
Corporation, verifies address information with the names of the
identified tax payers. Enhanced address accuracy and completeness
via Acxiom's address hygiene process typically results in improved
targetability. Name and address information is then sent to a
credit bureau, e.g., TransUnion for credit information. Credit
information may include credit scores and raw credit information.
Because historical tax information is being analyzed, the credit
information typically corresponds to the same timeframe (e.g., for
the years 2002 and 2003 in this example).
[0047] Module 207 obtains the raw credit data, historical tax data,
and credit scores from module 205 to form a collections model using
an application developed on the CAR. (Raw data, sometimes called
source data or atomic data, is data that has not been processed for
meaningful use and that has been collected but not formatted or
analyzed. Raw data often is collected in a database, where the raw
can be analyzed and made useful for an application.) Modeling
activities begin after CAR is available. Preliminary data analysis
for basic checks and data validity may be performed. With an
embodiment, module 207 performs decision tree segmentation using a
statistical analysis package to analyze credit scores (e.g.,
SAS/STAT software) to find sufficiently differentiated segments
(score bands) and creates a separate segment model for each score
band (segment), thus increasing the overall predictive power.
[0048] The collections model may be dynamically retrained prior to
use in order to capture the latest information available. This
approach is different from the typical static credit model approach
where the models and the data variables are held constant. In this
case, the collections model and the data are allowed to change.
[0049] Module 207 creates a collections model using tax-return and
credit data that will identify and rank all future receivables on a
likelihood of payment during collections process. Collections
scores generated by the collections model will be used to rank
receivables--a higher score implies that creditor is more likely to
pay compared to creditor with a lower score. On the basis of
collections scores, differentiated collections treatments can be
designed and optimized over time for each risk score band of the
collections model.
[0050] With an embodiment of the invention, segment modeling is
performed using a KXEN data mining tool. The KXEN tool divides data
into estimation (75%) and validation (25%) sub-samples, where
validation results verify robustness/stability of the collections
model. The KXEN tool differentiates between behavior of "good" and
"bad" tax filers. The KXEN tool mines more than 1,000 tax and
credit variables and identified attributes that are predictive in
explaining payment behavior. The KXEN tool generates automated
final model equations (scoring expressions) that is used to score
tax filers who still owe tax-dues to find individuals who are most
likely to pay owed amounts. With an embodiment of the invention, a
scoring expression is a statistical regression equation determined
by the statistical tool. The regression equation typically includes
only the relevant variables from more than 1000 mined
variables.
[0051] Module 209 tests and verifies the collections model
developed by module 207. In an embodiment, module 209 extracts
receivables for the 2004 tax year and determines the collections
scores using the collections model. Treatment actions based on the
determined treatment type are directed test groups. The "Goods"
(those who pay) and the "Bads" (those who do not pay within a
predetermined time duration (performance window)) are measured.
[0052] One the collections model has been developed by module 207
and verified by module 209, module 211 prepares the collection
model for the targeted revenue agency. For example, the collection
model may be implemented as a computer-readable medium having
computer executable instructions and distributed to a revenue
agency over a secure communications channel (e.g., LAN 152 as shown
in FIG. 1) or as an apparatus that utilizes a computer platform,
e.g., computer 100.
[0053] FIG. 3 illustrates process 300 for configuring a plurality
of score bands in a collections model in accordance with an
embodiment of the invention. In an embodiment, process 300 is
performed by module 207 as shown in FIG. 2. A sampled population
350 of debtors (using historical tax data as previously discussed)
is analyzed to configure a plurality of score bands (segments) in
accordance with desired statistical characteristics. The tree based
algorithm finds the top variable which divides the debtors into
segments with similar percentage of "goods" and "bads." Sampled
population includes a combination of "Goods" (21966 debtors or 74%)
and "Bads" (7727 debtors or 26%). As will be further discussed, the
debtors are assigned to one of the score bands based on credit
score 351 (NA2OJTOT) that is built and produced by TransUnion (TU).
However, other embodiments may use other scores, e.g., another
credit score or a customized score that is determined from a
combination of tax form data and raw credit data.
[0054] Each debtor of the sampled population of debtors is assigned
to one of six score bands (segments) based on the associated credit
score 351. Debtors that satisfy criterion 301 (NA201TOT<491.5)
are assigned to score band 1, Debtors that satisfy criteria 303 and
305 (491.5<=NA201TOT<525.5) are assigned to score band 2, and
debtors that satisfy criteria 303 and 307
(525.5<=NA201TOT<581.5) are assigned to score band 3.
Similarly, debtors are assigned to score bands 4, 5, and 6 that
satisfy criteria 309, 311 and 313, respectively.
[0055] FIGS. 4-9 show configurations for segment models for each of
the score bands that are determined by process 300 as performed by
module 207 when constructing a collections model. As previously
discussed, a scoring expression is determined for each score band
(segment). Even though over a thousand credit and tax variables are
available, the scoring expressions shown in FIGS. 4-9 are limited
to twenty variables in order to reduce calculations for determining
a desired collections objective. In general, a scoring expression
(given that the j.sup.th score band is selected) may be expressed
as:
collections_score = i = 1 N W i , j .times. V i , j ( EQ . 1 )
##EQU00001##
where N is the numbers of variables used in a scoring expression,
W.sub.i.j is the weight for the i.sup.th variable of the j.sup.th
score band, and V.sub.i.j is the value of the i.sup.th variable of
the j.sup.th score band.
[0056] With an exemplary embodiment of the invention, module 207
selects 20 variables for each scoring expression. However, with
other embodiments module 207 may select a different number of
variables, where the variables vary with different scoring
expressions.
[0057] FIG. 4 shows scoring expression 400 for the first score band
as shown in FIG. 3 in accordance with an embodiment of the
invention. Scoring expression 400 utilizes twenty variables
selected from over one thousand raw credit data and tax form data.
For example, variable 401 (ratio_taxowed_ctincome) is considered as
having the greatest importance and is accordingly given the
greatest weight 405 (17.9%). Variable 403 (RE36) has the next
greatest importance and is given weight 407 (7.7%).
[0058] FIG. 5 shows scoring expression 500 for the second score
band as shown in FIG. 3 in accordance with an embodiment of the
invention. Scoring expression 500 utilizes twenty variables
selected from over one thousand raw credit data and tax form data.
For example, variable 401 (ratio_taxowed_ctincome) is considered as
having the greatest importance and is accordingly given the
greatest weight 503 (13.1%). Variable 501 (PS230) has the next
greatest importance and is given weight 505 (7.5%). In the
exemplary embodiment, scoring expressions 400 and 500 have one
common variable (variable 401) with the remaining variables being
different (e.g. variables 403 and 501).
[0059] FIG. 6 shows scoring expression 600 for the third score band
as shown in FIG. 3 in accordance with an embodiment of the
invention. Scoring expression 600 utilizes twenty variables
selected from over one thousand raw credit data and tax form data.
With an embodiment of the invention, the majority of the variables
of scoring expression 600 are different from the variables of the
other scoring expressions 400, 500, 700, 800, and 900,
[0060] FIG. 7 shows scoring expression 700 for the fourth score
band as shown in FIG. 3 in accordance with an embodiment of the
invention. Scoring expression 700 utilizes twenty variables
selected from over one thousand raw credit data and tax form data.
As shown in FIGS. 4-9, variable 401 (ratio_taxowed_ctincome) is
commonly used by scoring expressions 400-900. Moreover, sonic of
the variables of scoring expression 700 may be used by some of the
other scoring expressions. For example, variable 701
(home_ownership) is used by scoring expression 400 but not by the
other scoring expressions.
[0061] FIG. 8 shows scoring expression 800 for the fifth score band
as shown in FIG. 3 in accordance with an embodiment of the
invention. Scoring expression 800 utilizes twenty variables
selected from over one thousand raw credit data and tax form data.
The fifth score band contains debtors having a very low credit risk
with a small proportion of "Bads."
[0062] FIG. 9 shows scoring expression 900 for the sixth score band
as shown in FIG. 3 in accordance with an embodiment of the
invention. Scoring expression 900 utilizes twenty variables
selected from over one thousand raw credit data and tax form data.
The sixth score band contains debtors having the lowest credit risk
with almost no "Bads."
[0063] As previously discussed, a collections model is constructed
as shown in FIGS. 2-9. The collections model can then be used by a
revenue agency to determine and initiate collections treatment for
debtors.
[0064] FIG. 10 shows a process 1000 for determining a collections
score for a debtor in accordance with an embodiment of the
invention. The collections scores, as generated by collections
models, enable revenue agencies to better align workload with
workforce and other available resources. Enhanced efficiency is
accomplished by prioritizing accounts based upon the collections
score. Accordingly, the most likely to pay receive "softer"
collection approaches and the least likely to pay receive more
assertive treatments earlier in the collections process. The
prioritization of accounts identifies the most difficult debtors to
collect accounts. These accounts can be forwarded to private
collections services at the onset when these accounts are still
fresh. It is expected that using the collections score to
prioritize and assign accounts may increase revenue derived from
accounts receivable collections by 3% to 7%.
[0065] Procedure 1001 obtains a credit score for a debtor after the
collections model has been constructed by process 200 (as shown in
FIGS. 2 and 3). In an embodiment of the invention, NA201TOT is a
credit score that is built and produced by TransUnion (TU) and that
is utilized in an embodiment of the invention. (TransUnion is a
credit bureau as previously discussed.) NA201TOT is also called TU
New Account Score. As performed by procedure 1003, a tax filer is
classified into one of six segments on the basis of their NA201TOT
score. Each of the six segments (score bands) has a separate model
equation (scoring expression). Procedure 1005 uses the associated
scoring expression to determine the collections score. If a debtor
is assigned to segment `2` on the basis of debtor's NA201TOT score,
then collections model `2` equation is used to determine the
collections score for the debtor. With an embodiment of the
invention, procedure 1007 determines the collections treatment type
that is based on a debtor's collections score (also called ATCS
score), irrespective of the debtor's segment score band)
assignment. In an embodiment, if two debtors have the same
collections score but are assigned to different segments, the
collections treatment type is the same. (However, embodiments of
the invention may associate different collections treatment types
for the same collections score for different score bands, i.e., the
collection treatment type may be dependent on the score band.) As
an example, debtor.sub.13 1 has an ATCS score of 0.88.
Debtor.sub.--2 has an ATCS score of 0.14. Debtor.sub.--1 has a high
score, i.e., is very likely to pay any owed amount, so the revenue
agency just sends a notice letter (Treatment Type A). Corresponding
action actions are initiated from the determined treatment type.
Debtor2 has a low score, i.e., is not likely to pay, so the revenue
agency sends the debtor a strongly worded letter. If no payment is
received within 21 days, for example, the revenue agency sends
another strong letter. If payment still not received after second
reminder, the revenue agency refers debtor.sub.--2 to a debt
collector. (Treatment Type C) An exemplary collection rule set is:
[0066] If ATCS>=0.75 then initiate treatment A [0067] If
0.4<=ATCS<0.75 then initiate treatment B [0068] If
ATCS<0.4 then initiate treatment C Collections score bands and
treatments may continuously change and improve over time. For
example one may "tweak" treatment type A. As another example, one
may change the cutoff from 0.75 cutoff to 0.7). With the above
embodiment, NA201TOT is used for scoring any debtor. Using NA201TOT
provides additional power to collections models. However,
embodiments of the invention may build models without NA201TOT. For
example, a collections score may he determined from a combination
of tax form data and raw credit data. Procedures 1001-1007 are
repeated if additional debtors are to be processed as determined by
procedure 1009.
[0069] FIG. 11 shows process 1007 (as shown in FIG. 10) for
determining a collections treatment type from a collections score
in accordance with an embodiment of the invention. In step 1101 of
the collections score (as determined by procedure 1005) is greater
or equal to 0.75, collection treatment type_A 1103 is selected. In
step 1105, the collections score is between 0.75 and 0.4,
collection treatment type_B 1107 is selected. Otherwise, collection
treatment type_C 1109 is selected.
[0070] FIG. 12 shows apparatus 1200 that analyzes raw credit data
and tax form data to initiate a collections treatment action in
accordance with an embodiment of the invention. Model analyzer 1201
constructs a collection model using historical tax data performing
process 200 as previously discussed. Model analyzer 1201 provides
the configuration for a plurality of score bands (segments) and
associated scoring expressions to scoring analyzer 1203. Scoring
analyzer 1203 consequently determines the collections score for the
debtor being processed. Treatment analyzer 1205 determines the
collection treatment type from the collections score. Consequently,
treatment generator 1207 initiates treatment action (e.g., letters
to debtors and instructions to a debt collector) to the directed
debtor.
[0071] As can be appreciated by one skilled in the art, a computer
system (e.g., computer 100 as shown in FIG. 1) with an associated
computer-readable medium containing instructions for controlling
the computer system may be utilized to implement the exemplary
embodiments that are disclosed herein. The computer system may
include at least one computer such as a microprocessor, a cluster
of microprocessors, a mainframe, and networked workstations.
[0072] While the invention has been described with respect to
specific examples including presently preferred modes of carrying
out the invention, those skilled in the art will appreciate that
there are numerous variations and permutations of the above
described systems and techniques that fall within the spirit and
scope of the invention as set forth in the appended claims.
* * * * *