Methods and Systems For Processing Electronic Checks
Methods and systems for receiving account data, where the account data includes an account number that is compared to a previously submitted account number where the previously submitted account number is associated with a successful payment. It is then determined if there is a likely match between the account number and the previously submitted account number by performing at least one of several fuzzy logic techniques. Upon determining that the likely match exists the parsed account number is automatically replaced with the previously submitted account number.
Latest CHECKFREE CORPORATION Patents:
The present application is a continuation of, and claims the benefit of priority to, U.S. patent application Ser. No. 10/925,592, “Methods and Apparatus for Processing Electronic Checks,” filed on Aug. 25, 2004. The entire content of the above-recited priority document is incorporated herein by reference as if set forth fully herein.
BACKGROUND OF THE INVENTION1. Field of the Invention
The invention relates to electronic checks. More particularly, the invention relates to methods and systems for processing electronic checks to minimize the number of returned electronic checks.
2. Brief Description of the Prior Art
All bank checks are encoded with information which allows for machine processing of the check. This information is printed at the bottom of the check starting in the lower left hand corner and often extending under the signature line. This line of printed information is referred to as the MICR (magnetic ink character recognition) line. Prior to writing the check, the line of information includes the bank routing number (ABA number), the checking account number, and the check number. After the check is written and presented for payment, the check is read by a human and the amount of the check is imprinted on the MICR line to the right of the other information. From that point on, the check is processed by machine.
An electronic check or “e-check” is like a check without the paper. The drawer or maker of the check provides the necessary information from the check, but not the paper. This can be accomplished in several ways. One popular way is referred to as electronic check conversion. Here, the drawer or maker of the check provides a void blank check which is scanned to obtain all of the information except for the payment amount which is then entered via a keyboard or keypad. Another method is to enter all of the information via a keyboard or keypad. The latter method is used when payment is made by telephone or via the Internet. In telephone payment methods, the check information is usually obtained via an interactive voice response unit (IVRU) which plays pre-recorded prompts and interprets DTMF (dual tone multi-frequency) signals generated by the telephone keypad.
E-checks are commonly used in situations where the payer chooses not to use a credit card (e.g. does not have a credit card) or where the payee chooses not to accept a credit card. E-checks are processed in the same manner as paper checks which have been fully encoded and machine read. The information is compared to databases to determine whether the checking account can be identified and whether the checking account has sufficient funds to pay the check. If either of these inquiries fail, the check (paper or e-check) is “returned” to the presenter (usually the payee). If the account has insufficient funds to pay the check, the return is referred to as an NSF (not sufficient funds) return. If the account can not be identified, the return is referred to as an administrative (ADM) return. A little over 1% of paper checks are returned, mainly NSF returns. E-checks have a much higher return rate (e.g. 4.7%) than paper checks and this is mainly due to ADM returns.
The high administrative return rate of e-checks is mostly due to human error on the part of the payer (check drawer). Although the MICR line information is standardized, the standardization only lends itself to optical reading and character recognition. It is not positional and the characters denoting a field are unknown to a person so the standard cannot be used when optical reading equipment is unavailable. There are also optional fields on some checks and this may confuse the payer when attempting to read the MICR line. Often a payer will enter too many or too few digits. In addition, a phenomenon known in the art as “fat fingers” causes erroneous keypad and keyboard entries (typographical errors). These errors are usually associated with the account identification information rather than the payment amount data. It is estimated that two thirds of e-check returns are administrative returns. Prior art
The high rate of e-check returns is a significant problem for businesses that rely heavily on e-check payments, such as utility companies. A typical utility company may have approximately one million customers who pay their monthly bill by e-check. With an e-check return rate of 4.7%, that means that 47,000 customer payments will be rejected every month. This causes an immediate cash-flow problem for the company but also causes a very expensive customer relations problem. Most of the 47,000 customers will need to speak to a customer service representative to correct the situation. At a conservative estimate of $10 per customer service call, this will add nearly half a million dollars to the company's monthly operating expenses.
State of the art systems for processing e-checks include TeleCheck® by TeleCheck Services, Inc., Houston, Tex., and StarCheck® by Concord EFS, Inc., Memphis, Tenn. These systems are primarily aimed toward reducing NSF returns and do not significantly reduce ADM returns. JP Morgan Chase maintains a data base system called “File-Fixer” which is purported to be able to reduce ADM returns, but thus far has been unable to reduce ADM returns at all. PhoneCharge, Inc. conducted an in house test and found that it did not reduce the admins, in fact it had a negative effect because valid checking accounts were locked making them ineligible for transaction processing.
BRIEF DESCRIPTION OF THE DRAWINGS
The entire content of the Appendix, including the files and source code originally submitted to the United States Patent Office by CD-ROM, of U.S. patent application Ser. No. 10/925,592, filed on Aug. 25, 2004, of which this application claims the benefit of priority to, is incorporated herein by reference as if set forth fully herein. The invention is most easily understood by reference to the flow charts rather than the C code of the incorporated by reference Appendix of U.S. patent application Ser. No. 10/925,592. Reference to the code will be made where possible when explaining the flow charts. It will be appreciated however, that the flow charts do not directly correspond to the C code.
Turning now to
Turning now to
Referring now to
Returning to
Starting at 56 in
If a Valid response is returned to 70 in
If it is determined at 76 that no ABA numbers were found in the data entered by the customer, it is determined at 80 whether there was a pervious check payment by this user. This determination is made based on information obtained from an interactive payment system within which this invention is implemented. A suitable interactive payment system is described below with reference to
Still referring to
Turning now to
The Fuzzy ABA function returns to 114 in
Continued processing of checking account information begins as outlined in
Referring now to
Turning to
If neither of the two remaining fields is the correct length as determined at 140, 142, or 144, it is then determined at 152 whether only one of the fields is less than seven digits. If only one field is less than seven digits in length, the account number is taken to be the longer field at 154 and the validate check length function is called at 156. This function is described in detail below with reference to
If it is determined at 152 that both fields are less than seven digits or both are seven or more digits, it is then determined at 158 whether the check number is known. If the check number is not known, the flag ASK_FOR_CHECK_NO is set at 160 and the routine returns to 170 in
Turning now to
If it is determined at 178 that the check number is not known, the Echeck server is called at 190 to obtain the valid lengths of a checking account number for the present ABA number. At 192 it is determined whether either the left string or the right string has the correct number of digits. If one of the strings does have the correct number of digits, that string is taken to be the account number at 194 and the validate check length function is called at 196. This function is described in detail below with reference to
If neither the left string nor the right string have the correct number of digits as determined at 192, it is determined at 198 whether both strings are less than or equal to, or greater than, six digits in length. If yes, the flag ASK_FOR_CHECK_NO is set at 200 and the process returns to 170 in
If it is determined at 174 that the ABA number is either the first or the last field, a different processing routine for more than three fields is called at 206. This routine is outlined in
Turning now to
If it is determined at 210 that the ABA number is the last field, the following variables are set at 214: “left account number” is set to all of the digits in the first field through the second from the last field, “left check number” is set to the next to the last field, “right account number” is set to all of the digits in the second field through the next to the last field, and “right check number” is set to the first field.
After these variables are set, it is determined at 216 whether the check number is known. If it is, the variable “left check number” is compared to the known check number at 218. If they are not the same, the variable “right check number” is compared to the known check number at 220. If these don't match, the flag START_OVER is set at 222 and the process returns to the start at 10 in
If it is determined at 218 that the variable “left check number” is the correct check number, the variable “left account number” is taken to be the checking account number at 224. If it is determined at 220 that the variable “right check number” is the correct check number, the variable “right account number” is taken to be the checking account number at 226. The validate check length routine is then called at 228.
If it is determined at 216 that the check number is not known, the Echeck server is called at 230 to determine the correct number of digits for the account number based on the current ABA number. If it is determined at 232 that both variables “left account numbers” and “right account number” have the incorrect number of digits, the flag ASK_FOR_CHECK_NO is set at 234 and the process returns to 170 in
The validate check length function begins at 248 in
Referring now to
Returning to
Turning now to
Referring now to
Finishing with
The Echeck server Checking Account number Validation function is outlined in
The appropriate length(s) for a business account is obtained at 338. This is obtained from a database which matches ABA numbers with correct account number lengths. The database is built over time based on successful payments. It is then determined at 340 whether the present account number is a correct length. If it is, the flag MICR_OK is set at 342 and the routine returns to 480 in
When entering
If the checking account number length is not correct, a determination is made at 344 whether this ABA number, account type and length have been associated with a threshold number of successful payments. The presently preferred threshold is fifty. If yes, the flag BAD_LENGTH is set at 346. If no, the flag LENGTH_UNSURE is set at 348. In either case, the process continues as outlined in
In the case of BAD_LENGTH, the process continues at 350 in
Turning now to
After the payment information is obtained from the customer, it is presented to the customer'bank for payment. This process is illustrated in
The bank processing of an echeck is outlined in
If all of the above tests are passed, the customer is prompted at 740 to enter the payment amount and it is determined at 750 whether that amount exceeds a transaction cap. If the amount is too high, the customer is so prompted at 760 and a loop is entered which doesn't end until the customer enters an acceptable amount. Though not shown in the Figure, a counter can be set to exit the loop and end the call after several failures. When the customer enters an appropriate amount, the invention described above is invoked at 770 starting with
If it was determined at 790 that fewer than two payments were made with this checking account, it is then determined at 850-860 whether this checking account has had any checks returned. If yes, the steps at 800-840, described above, are followed. If not, the payment is processed at 870.
Those skilled in the art will appreciate that the invention relies on several databases, in particular, a “transaction” database, a “negative” database, and a “length” database. The transaction database includes a list of all transactions made through the system. Each transaction includes customer account number, ABA, checking account number, payment amount, etc. The negative database includes a list of all customer account numbers and checking account numbers that have had declines or returns. Each record includes the customer account number or checking account number and the number of declines/returns on that number. The length database includes a list of all valid checking account number lengths associated with non-returned payments through this system. Each record includes DNIS, ABA, number of payments to the ABA for personal checks, number of payments to the ABA for business checks, a list of the personal lengths associated with the ABA, a list of the business lengths associated with the ABA.
According to the presently preferred embodiment, two counts are stored per ABA number: the number of successful payments for personal checks and the number of successful payments for business check. When determining if a given length is “valid”, “invalid”, or “unsure”, the number of payments is compared to the threshold (currently 50). If the number of payments is below this threshold, then the length is unsure if it is not in the lengths database. This process will be enhanced by including counts by each length. It will then be possible to determine if the length should be considered valid or unsure based on how many times that given length has been used for a successful payment. For example, if a length of 7 is given for ABA 338922833 and the length database has the following, number of valid 7 digit payments=2 and total number of payments to ABA 338922833=3000, it cannot be certainly determined that a length of 7 is valid due to the fact that 2 is such a small percentage of 3000. In this case, the process would return unsure.
The methods of the invention may be applied in instances where no parsing of account data is necessary, i.e. where the customer parses the data before inputting it to the system. In addition, customer data may be obtained in ways other than directly from a user. For example, data from multiple users may be collected outside of the system, by a payee or a bank, and presented to the system in the form of a batch file. Data may also be presented in paper form which must be scanned or manually input into the system. In cases where no parsing is performed, the determination of account validity can still be performed using valid account number lengths associated with ABA numbers. Also, the fuzzy logic techniques of the invention can be applied independently of parsing and account number length validation.
There have been described and illustrated herein methods and systems for processing electronic checks. While particular embodiments of the invention have been described, it is not intended that the invention be limited thereto, as it is intended that the invention be as broad in scope as the art will allow and that the specification be read likewise. It will therefore be appreciated by those skilled in the art that yet other modifications could be made to the provided invention without deviating from its spirit and scope as so claimed.
Claims
1. A method comprising:
- receiving account data;
- parsing the account data to determine an account number;
- comparing the parsed account number to a previously submitted account number wherein the previously submitted account number is associated with a successful payment;
- determining if there is a likely match between the account number and the previously submitted account number by performing at least one of a plurality of fuzzy logic techniques; and
- upon determining that the likely match exists, automatically replacing the parsed account number with the previously submitted account number.
2. The method of claim 1, wherein the plurality of fuzzy logic techniques include at least one of the following:
- if there has been a transposition of a first digit of the parsed account number with a second digit of the parsed account number,
- if there is no more than one digit of the parsed account number different than the previously submitted account number,
- if the number of digits in the parsed account number and the previously submitted account number differ by no more than one, and
- if more than a preset percentage of the string of digits of the parsed account number and the string of digits of the previously submitted account number match.
3. The method of claim 2, wherein the preset percentage is equal to or greater than 80%.
4. The method of claim 1, wherein the account number is associated with a customer.
5. The method of claim 1, further comprising storing the parsed account number in a database.
6. The method of claim 1, further comprising transmitting payment information to a financial institution associated with the previously submitted account number.
7. The method of claim 1, further comprising parsing the account data to determine a routing number;
- comparing the parsed routing number to a previously stored routing number wherein the previously stored routing number is associated with a successful payment;
- determining if there is a likely match between the routing number and the previously stored routing number by performing at least one of the fuzzy logic techniques; and
- upon determining that the likely match exists, automatically replacing the parsed routing number with the previously stored routing number.
8. The method of claim 1, wherein the account data includes at least one non-numeric character.
9. The method of claim 1, wherein the account data includes at least a portion of MICR line data.
10. The method of claim 1, wherein the step of parsing includes separating the account data into fields based on the placement of at least one non-numeric character.
11. The method of claim 10, wherein the step of parsing further includes counting the number of digits in each field of account data to determine at least a likely account number field.
12. The method of claim 11, wherein the step of parsing further includes comparing the number of digits in the likely account number field to a predetermined number of valid digits for account numbers associated with a financial institution.
13. A method comprising:
- receiving account data, wherein the account data includes an account number;
- comparing the account number to a previously submitted account number wherein the previously submitted account number is associated with a successful payment;
- determining if there is a likely match between the account number and the previously submitted account number by performing at least one of a plurality of fuzzy logic techniques; and
- upon determining that the likely match exists, automatically replacing the parsed account number with the previously submitted account number.
14. The method of claim 13, wherein the plurality of fuzzy logic techniques include at least one of the following:
- if there has been a transposition of a first digit of the account number with a second digit of the parsed account number,
- if there is no more than one digit of the account number different than the previously submitted account number.
- if the number of digits in the account number and the previously submitted account number differ by no more than one, and
- if more than a preset percentage of the string of digits of the account number and the string of digits of the previously submitted account number match.
15. The method of claim 14, wherein the preset percentage is equal to or greater than 80%.
16. The method of claim 13, further comprising storing the account number in a database.
17. The method of claim 13, further comprising transmitting payment information to a financial institution associated with the previously submitted account number.
18. The method of claim 13, wherein the account data includes at least one non-numeric character.
19. The method of claim 13, wherein the account data includes at least a portion of MICR line data.
20. A system, comprising:
- means for receiving account data, wherein the account data includes an account number;
- means for storing a previously submitted account number, wherein the previously submitted account number is associated with a successful payment;
- means for comparing the account number to the previously submitted account number;
- means for determining if there is a likely match between the account number and the previously submitted account number by performing at least one of a plurality of fuzzy logic techniques; and
- means for automatically replacing the parsed account number with the previously submitted account number when the likely match exists.
21. A system, comprising:
- a database, wherein the database includes a plurality of previously submitted account numbers;
- an interface; and
- a processor, in communication with the database and interface, wherein the processor is configured to execute software instructions for: receiving, via the interface, account data, parsing the account data to determine an account number, comparing the parsed account number to a previously submitted account number stored in the database, wherein the previously submitted account number is associated with a successful payment, determining if there is a likely match between the account number and the previously submitted account number by performing at least one of a plurality of fuzzy logic techniques, automatically replacing the parsed account number with the previously submitted account number when the likely match exists, and transmitting, via the network interface, payment information to a financial institution associated with the previously submitted account number.
Type: Application
Filed: Feb 12, 2007
Publication Date: Jun 14, 2007
Applicant: CHECKFREE CORPORATION (Norcross, GA)
Inventors: Brian Mejias (Howard Beach, NY), Thomas Cain (West Haven, CT), Garret English (Stormville, NY), Leonard DeCaro (Ansonia, CT)
Application Number: 11/673,783
International Classification: G07F 19/00 (20060101); G06Q 40/00 (20060101);