© EZ-R Stats, LLC Duplicate Payments Slide 1 Auditing for Duplicate Payments A better way … Web CAAT.

© EZ-R Stats, LLC Duplicate Payments Slide 1

Auditing forDuplicate Payments

A better way …

Web CAAT

May 29, 2010 © 2010 EZ-R Stats, LLC Slide 2

About duplicate payments Why they occur Fraud Errors Control breakdowns

System Procedures

How to detect


Historical ExperienceState of North Carolina

Fiscal 1996 – 2004 $4.5 million recovered Approximately $500K /

year Most recent experience

About $400K/ year

“Pay and Chase”


Matching approach

Exact matching“Fuzzy”

matchingEvery possible

pair


Duplicate payments


Why check for them?Recovery feePossibility of

fraudIdentify

control break downs

Proactive checking


Cost Recoveries

“Pay and Chase”35% fee to Cost Recovery ContractorRisk of loss

Proactive Approach Identify up-frontMake control recommendations to

preventContinuous monitoring


Invoice elements Vendor number Invoice number Invoice Date Invoice Amount


Exact matchesAll four elements matchThree combinations of three way

matchVendor, invoice, amountVendor, invoice, dateVendor, amount, date


“Fuzzy” matches – invoice numbers

Levenshtein distance

Transpositions LDO (letters, digits

only) Same characters Leading characters Trailing characters


Example of Levenshtein distance

Measure similarity of invoice “12341” and “24371”.

Start 12341 24371

Step1 – delete left most digit

2341 24371

Step2 – Insert a “4” between “2” and “3”

24341 24371

Step 3- Replace seond “4” with “7”

24371 24371

Levenshtein distance is 3


“Fuzzy” matches – invoice number


Date Transpositions07/31/2010 vs. 07/13/2010

01/21/2009 vs. 02/11/2009


Data validation

Invoice date Invoice amount Vendor number


Potential duplicates?


Potential duplicaes

Menu item potential duplicates


Forms

Browser based Pull down menus “Fill in the blanks”


Results

Web based report Import into Excel


Processing volumes500,000 invoices40,000 vendorsProcess on lap-top with dual

2.2 GHzAbout two minutes per test


Road tested


Pricing


DemoExcel workbook10,417 Payments19 tests


Workbook

Excel workbook, 10,417 payments, 10 columns


19 Tests can be run “A” – “S”

Description of tests Used for identifying potential duplicate

payments Same concept applies to other areas

Journal entries Purchase orders Expense reports Fixed asset items, etc.


Test “A”

All four key values equal Same vendor Same amount Same invoice date Same invoice number Note: case insensitive comparisons


Test “B”

Same vendor,Same invoice number,Same invoice amount


Test “C”

Same vendor number,Same invoice number,Same invoice date


Test “D”

Same vendor,Same invoice amount,Same invoice date


Test “E”

Same vendor number,Same invoice amount,Two invoice numbers the same

considering letters and digits only (i.e. no special characters)


Test “F “

Same vendor,Same invoice amount,Same invoice number, if only

letters are considered


Test “G”

Same vendor number,Same invoice amount,Same invoice number, if only

digits are considered (i.e. ignore letters and special characters, blanks, etc.)


Test “H”

Same vendor,Same invoice amount,Invoice numbers are within the

specified Levenshtein distance


Test “I”

Same vendor,Same invoice amount,Invoice numbers are different

due only to a transposition


Test “J”

Same vendor number, Same invoice amount, Over 90% of the characters/digits in

each invoice are the same Can specify different percentage Characters not necessarily in same

sequence


Test “K”

Same vendor,Same invoice date,Invoice amounts are within 2%

of each otherCan specify different

percentage


Test “L”

Same vendor, Same invoice date, First four leading characters of two

invoices are the same Can use different number of leading

digits Can specify different tests (LDO, DO,

LO)


Test “M”

Same vendor, Same invoice date, First four trailing characters of two

invoices are the same Can use different number of leading

digits Can specify different tests (LDO, DO,

LO)


Test “N”

Same vendor,Same invoice date,Same invoice amountDifferent invoices


Test “O”

Same invoice number,Same invoice date,Same invoice amount,Different vendor


Test “P”

Same invoice number,Same invoice date,Similar amountMeasure as percentageAbs(invamt1-invamt2)/invamt1Auditor specifies percentage, e.g.

2%


Test “Q”

Same vendor, Same invoice amount, Same invoice number Similar invoice date Measured using Levenshtein distance Auditor specifies test distance


Test “R”

Same vendor,Same invoice number,Similar dateMeasured using Levenshtein

distanceAuditor specifies distance


Test “S”

Same invoice number, Same invoice date, Same invoice amount Similar vendor number Measured using Levenshtein distance Auditor specifies distance


Output

Output is to a text fileImport into ExcelPairs of rows


Limitations

Currently handles only: Excel Access Text files (csv,tsv, etc.) No limit on rows (other than imposed by Excel) Has been tested using about 450,000 invoices Feasible to run on PC


Processing times

File of 10,000 payments takes less than one minute

Some tests take longer:Levenshtein distanceLeading/trailing digits


Benchmark results500,000 invoices tested6,000 vendorsDone on lap-top with dual 2.2

GHzAbout two minutes per testLarger volumes require longerYMMV


Duplicate Vendors

Primary cause of duplicate payments Identified using two primary methods

Exact – Same, same, different “Fuzzy” – Name matching


Same, Same, Different

Same IRS Taxpayer ID (TIN) Different Vendor Number



Same Street Address Same City Same Zip Code Different Vendor Number



Same area code, Same contact number Same contact name Different Vendor Name



Same bank routing numberDifferent vendor name/number


“Fuzzy” matching of vendor names

Remove common terms (e.g. “corp”, “inc” etc.)

Remove all but letters and numbers Compare every combination using-

Match after removal of special characters Leading “N” characters Levenshtein distance


Fuzzy matching of TIN

TranspositionsLevenshtein distance


Fuzzy matching of Bank routing number Transpositions Levenshtein distance


Fuzzy matching of address

Letters and digits only Levenshtein distance Transpositions Same characters


Benchmark Timings

10,000 vendors Access database CPU 1.5 GHz, memory 500MB Same, same, different - < 1 minute “Fuzzy”

LDO – 20 seconds Leading – 2 minutes 10 seconds Levenshtein - ? (long time)


Continuous Monitoring

ObjectivesIdentify issues earlyVerify controls are workingQuantify areas for audit


Monitor for potential duplicate payments

Set up “duplicate payment test” directory Designate “log” file Run / refine tests Convert log file to “monitor” file Now simple to run tests on a cycle Just update file containing payments


Monitoring process

Run all testsReview outputReview for errors in current periodIdentify potential overpayments

early


Other areas for monitoring

Journal entries Expense reports P-card transactions Vendor payment trends Payroll Inventory Receivables Vendor master file


Example monitor processes

Invoice payments – regression analysis Counts Totals Averages By month, week,

quarter, etc. Policy compliance

Requirements for PO


Monitoring (cont’d)

Use of Benford’s Law Identification of credits

not taken Top “10” Discounts not taken Vendor master –

Checking for duplicates

Checking for PO Boxes/ drop boxes

Employee conflict of interest


Monitoring (cont’d)

“Impossible” transaction conditions

Data stratification Population statistics Quartiles Duplicate transactions Sequence gaps

Same, same, different


More info

Auditors Guide to MonitoringUser Guide – Audit Comman

der


ImplementationStart

smallLow

hanging fruit


Questions?

General info919-219-1622E-Mail

Thank you

© EZ-R Stats, LLC Duplicate Payments Slide 1 Auditing for Duplicate Payments A better way … Web CAAT.

Documents

invoice amounts

invoice datetest d

invoice datetest e

invoice amounttest c

different percentagecharacters

different tests ldo

considering letters

test aall