Minutes of Houston Chapter ACL User Group Meeting July 25, 2008

CHECK OUT WWW.TEXASACL.COM

Rand Sluder from Price-Waterhouse spoke to about using data analysis in detecting fraud. He first outlined some of the major types of frauds and then discussed some ways of using data analysis in fraud detection.

Fraudulent Finical Statements are not the most common types of fraud although they involve the most money. Some of the types include:
Overstatement of revenues and assets.
Understatement of expenses and liabilities.
Fictitious revenues.
Altering test documents.
Producing false reports.
Timing differences.
Fraudulent timing.
Improper disclosures.
Improper asset valuations.

Asset misappropriation is the most common, but involves the least amount of loss. It includes:
Embezzlement.
Larceny.
Skimming.
Payroll fraud.
Misuse of company assets.

Data analysis is necessary.

Data is at the heart of any fraud risk assessment or forensic accounting investigation.
imperative to identify fraud quickly.
It is difficult to identify patterns or clusters in data in large data bases. This is where data analysis, using ACL, comes in. Using the proper algorisms, you can detect transactions that merit investigation.

Ideas for using ACL to check for frauds:

Compare list of vendors with known fraudsters using 3rd party lists.
Compare vendor addresses with addresses like Mail Box Etc.
Look for ‘round trip’ transactions or relationships between vendors. Loans made to vendors that never come back.
Look for transactions with round dollar amounts, especially ones with ambiguous reasons.
Sequential  invoice numbers from a vendor. If it looks like you are the only customer, this could be a red flag.
Compare vendors to OFAC (list of persons the government has listed as suspect).

Fraud in Accounts Payable:

Some of the tests for accounts payable include:

Payments to employees and related parties. Some companies reimburse employees as vendors.
Look for “companies” with the same address as an employee.
Where there is no bid when there would normally be a bid.
Look for manual overrides of the system like cutting manual checks.
Look for an other operations outside the system.
Incorrect sequence and amounts
Checks.
Purchase orders.
Expense reports.
Improper authorization.
Inadequate support.
Void and manual checks.
Duplicate payments.


Some of the ideas from the group included:

Use ratio analysis. Take the vendor average (of invoices) and then look at the 1st, 2nd, and 3rd highest invoices. Is there a pattern? Are these outliers?

Look at the line item detail by vendor. Look for dropped decimals where the amount is 10X or 100X what it should be.

Use Benford’s Analysis on payments.

Porter used the following to ID a fraudulent vendor billing:
   1) Summarize on Vendor, Item_desc, Invoice
   2) Duplicates on Vendor, Item_desc OTHER Invoice
   3) Summarize on Vendor, Invoice
   4) Look for vendors who have repeated large count as this may indicate vendors billing the same items multiple of times.

There are up to five pieces of information in an AP billing:
1 -Who Provided            the vendor name/number
2 -the amount
3 -Date of Service
4 -What provided
5 -Who Received the services

Summarize on Who, What, and invoice. Subtotal amount
Do a duplicate test on who, what, and amount.
Summarize on who and invoice.

Look for potential conflicts of interest for couples working in the company.

Hash Values.

Get information on purchase cards from the bank and look for irregularities.

Look at vendors whose first name is 3 characters or less.

In looking for fraudulent employees look for:
Person with no benefits.
Person with the maximum number of exemptions.

Look at the forum for finding out how to determine the mapping of a field if you do not know it already.