Monday 7 July 2014
Data mining is an analysis process used by forensic accountants and internal auditors to examine data sets or metadata to identify patterns, anomalies, and trends to answer business queries and provide predictive value for future events. Data mining software incorporates algorithms to explore, analyze, classify, relate, and partition data sets that are then used to develop different models to achieve the business objective. In the example presented, the company may develop several models using different algorithms - including a predictive model, a classification model, and an exploration model - to identify the types of transactions, vendors, or personnel likely to be associated with purchasing fraud.
If you would like to find out more information about the measures your company, governmental entity, or non-profit organization can take to enhance its fraud detection programs, please note that IASeminars offers the following important events:
When you choose a modeling strategy, you need to understand how the algorithm works. For example, knowing that a neural network provides very limited feedback on how data inputs relate to the target data being sought, you probably would not choose a neural network model to analyze purchasing information. Instead, you may elect to use a regression model or a decision tree. If the input data set is quite limited, you would likely choose a decision tree model over a regression model. Once you have developed and validated your chosen model, you would analyze the model's results using lift, gain or profit charts, confusion matrices, threshold charts, and clustering analyses.
Forensic accountants can use data mining software to perform a variety of analyses often used to detect purchasing fraud. For instance, data mining software can be used to detect patterns of purchase orders being placed just below purchase limits (referencing the purchase agent, vendor and/or product); price variances for identical and similar products; period-end transactions; duplicate payments; adherence to Benford's law; and the appearance of round numbers in invoices or payments. In addition, order quantities for inventory or supplies can be compared to those ordered in prior periods, analyzed by vendor, and evaluated against sales increases or decreases. Price trends, including historical to current comparisons and high-low price variances over quarterly, annual or bi-annual periods, can also be considered. P-card use can be matched against master vendor lists (seeking to identify use of unauthorized vendors), compared by user for price and volume differences, and evaluated for purchasing and/or return patterns.
Because large and mid-size businesses often rely on master approved vendor lists to specify acceptable sources of supply, forensic accountants and internal auditors should consider using data mining tools to periodically evaluate data sets related to the master vendor list. Analyses in this area may involve identifying individuals and supervisors who added vendors to the master list (seeking to identify persons not normally tasked with these duties); comparing vendor addresses to employee or customer addresses; comparing tax identification numbers to government or private databases and to year-end tax reports submitted to government tax authorities; and identifying any accesses to the master vendor list related to modification of existing vendor information or addition of vendors without necessary approvals.
Metadata, popularly defined as "data about data," is the means by which digital information is classified, organized and tracked. Among other things, an organization's metadata might identify the names of the persons who created, modified or accessed data, the size and location of the file in which the data is stored, and the dates of access.
Metadata related to accounting ledgers, spreadsheets, or transaction files can be of great interest to forensic accountants and internal auditors seeking to identify patterns associated with purchasing fraud. For example, by analyzing access patterns to data in account, transaction or vendor files, the forensic accountant can rapidly identify preferred vendors used by specific purchasing agents and the frequency of change orders associated with a particular vendor. That information may be particularly useful in determining if change orders are being used to circumvent purchasing limits or to provide a vendor with a large number of "no-bid" contracts.
Although less common, searches of emails and text messages sent through company servers can occasionally flag a person's intent to commit fraud or expose an actual fraud in progress. Using search terms such as "make our numbers," "lose my job," "off the books," "we'll fix it later," "he/she told me to do it," or even "consulting fee," forensic accountants and internal auditors may occasionally get lucky and find a suspect email or text message. In these cases, the accountant or auditor will review the entire flagged email or text message, messages sent prior and subsequent to the communication, and seek to establish the context in which the message was sent. If further investigation is warranted, the accountant or auditor may then use other data mining tools to examine associated data sets for changes, additions or deletions that appear to relate to the suspect email or text message. By way of example, if a suspect email included language that implied the organization's controller had been pressured to change the accounting for a particular transaction, the accountant or auditor could then analyze the controller's accesses of the general, sales or other ledgers in the days following the email's receipt to identify what actions the controller had taken in response.
If asset misappropriation frauds were limited to purchasing schemes, there is no doubt that the job of the forensic accountant and internal auditor would be far simpler than is the case. Unfortunately, asset misappropriation schemes can take on a variety of forms such as ghost employees and payroll fraud, cash diversions and register disbursements fraud, check tampering, and expense reimbursement fraud, just to name a few. And this doesn't include financial statement fraud, a source of large losses for victim organizations. While robust controls and monitoring can reduce the likelihood of an organization experiencing significant fraud losses, controls alone will never prevent the commission of a well-crafted fraud. For this reason, forensic accountants and internal auditors are wise to leverage their limited budgets by using data mining techniques to detect the patterns, trends and anomalies in all relevant areas and to direct further investigative efforts.
If you would like to find out more information about the IASeminars courses in this area, please .