Automating Fraud Detection in

Financial Statements

Rajiv Gupta, CFE, CA, CISA, CCSA

Vice President and Chief Internal Auditor
Diageo India
Rajiv Gupta.
CA, CFE, CISA, DISA, CCSA
Vice President & Chief Internal Auditor
ACFE Asia Pacific Conference
Diageo India
Skype: rajiv.gupta.tccc Hong Kong, Sep 2017
E: Rajiv.Gupta@Diageo.com
 What we will discuss?


 Setting the context: Case Study & Statistics

 Automation of Data Analytics: What and Why?

 Fraud Detection Toolkit- How CFEs can leverage the same?

 Detection of Fraud at Financial Statements Level: Beneish Model

 Take Away!
Setting the Context

Gowex case


 Red Flags in Gowex Case


 GOW’s audit fee is €40,000, which makes sense if Gowex’s actual revenues are
only 5%-10% of reported revenue.

 Gowex told some investors that New York City was paying them €7.5 million.
GOW told Gotham that it is €2 million. The real number was <€200,000 only,
according to New York City.

 Gowex does not publicly disclose basic metrics, such as hotspot count or wireless
revenue segment details.

 The 5th June 2014 Mall Plaza press release (in English) made demonstrably false
claims. The same was omitted from press release in the native Spanish.
In 2014, Livent’s Auditor ordered to pay USD
84.8 million for failing to detect fraud


 Auditors has been ordered to pay damages to the
theatre company’s creditors after an Ontario judge
ruled the accounting firm failed to detect
fraud at the company, even though there were
plenty of warning signs that something fishy
was going on in the 1990s.

For years
 The judgment is a very rare situation.
creditors and investors have been
looking for ways to sue corporate
auditors.
 Initial Detection of Fraud

Source: 2016 ACFE Report to the Nations

 Automation
of

Data Analytics

- What & Why ?
 Automation of Data Analytics: Why?


Centralized Data Analytics: for Governance & Compliance related controls.

Testing complete sets of data, rather than sample testing.

Cost Reduction: 20%-30% annual internal audit cost reduction by eliminating manual
effort in data extraction , transformation, and analysis.

Complex Analytics: Perform complex calculations or logics that would otherwise be

difficult to re-perform manually.

Controls Testing: Test controls that otherwise might be inefficient to test or are not
testable manually.

Aiding risk assessment through identification of anomalies and trends, perhaps even
through comparison to industry data, pointing auditors toward items they need to
investigate further.
Automation of Data Analytics: What? (1/2)


 Identify and investigate invoices that did not pass the three-way-match (purchase order,
goods receipt, and vendor invoice).

 Identify payment terms on the purchase order that do not match the vendor master
information.

 Check the vendor’s invoice numbering. Almost sequential numbering means the vendor has no
or few other customers than you. This is an indicator that the vendor has been set up
specifically to do business with your organization.

 Check for split contracts with vendors to identify orders that are below approval or bidding
thresholds.

 Entries by unexpected users:

- Pick the users who have passed only 1 or 2 entries.
- Entries by admin, IT, Internal Auditor etc.
Automation of Data Analytics: What? (2/2)


 Compare invoice dates to shipping dates for large gaps in time or shipping date
prior to invoice date.

 Identify active Blacklisted Customer/Vendor every morning.

 To identify if a vendor code has been changed at time of invoice processing.

 PO date after invoice date.

 Duplicate purchases (same vendor same invoice number, same amount same GL
account).
Fraud Detection Toolkit


- How CFEs can leverage the same?
 Fraud Detection Toolkit


Fraud

Fraud at transaction-level Fraud at financial statements level

Detection mechanism Detection mechanism

Benford’s Law Relative Size Factor Beneish model

 Detection of Fraud
at

Financial Statements Level


Beneish Model
 ACFE suggests the use
World bank uses of the Beneish model -
and advises the use “Fraud Magazine” in the
of the Beneish March/April 2005 issue.
model.

Why
Beneish Success rate:
Successfully
Model? identified 76% of
manipulators
Used by analysts at
Merrill Lynch and
Prudential
Securities, among
others
Taught at universities
such as Berkeley,
Chicago, Cornell,
Dartmouth, New York,
and Indiana.
 Beneish Model
S. Type Mean-Non- Mean- Enron Satyam-
No. Manipulators Manipulators 2008

1 Days Sales
Index
in

2 Gross Margin Index

Receivables

1.031

1.014
1.465

1.193
0.630

1.450
1.068

0.984

3 Asset Quality Index 1.039 1.254 1.310 0.092

4 Sales Growth Index 1.134 1.607 1.526 1.463

5 Depreciation Index 1.017 1.166

6 Sales General & Administrative 0.650 1.090

Expenses Index

7 Leverage Index 1.040 1.191

8 Total Accruals to Total Assets 0.018 0.031 0.012 0.037

Classified - Internal use
M Score 7.80 7.41
 Beneish Model- Components

• This sales index measures whether receivables and revenues

DSRI 
are in or out of balance in two consecutive reporting periods.
A material increase in the index could indicate a company’s
receivables are phony.

• This index is designed to alert you that the risk of earnings

manipulation is higher when gross margins drop. But, if the
GMI company is already engaging in attempts to inflate earnings,
gross margins will be just the opposite: higher than normal.

• This index measures the proportion of total assets for which

future benefits may be less certain. For the purpose of
AQI evaluating earnings manipulation, an increase in the asset
quality index may indicate a company’s propensity to capitalize
costs.
 Beneish Model-Components…contd.
• Growth does not imply manipulation, but growth firms are viewed by
professionals as more likely to commit financial statement fraud as
SGI

their financial position and capital needs put pressure on managers
to achieve earnings targets. In addition, concerns about controls and
reporting tend to lag behind operations in periods of high growth.

• A DepI greater than 1 indicates that the rate at which assets are
depreciated has slowed down, raising the possibility that the firm
DepI has revised upwards the estimates of assets useful lives or has
adopted a new method that is increasing income

• This index is used because analysts would interpret a

disproportionate increase in sales as a negative signal about firms
SGAI future prospects. If sales increase faster than expenses there
needs to be an explanation. If not, the SGAI may be pointing to
overstated revenues.
 Beneish Model- Components…contd.

• This index captures increasing reliance on debt financing, as

LevI 
this increases the firm’s financial risk and the likelihood of
earnings manipulation (e.g., debt agreement constraints).

• An increase in accruals from one period to the next may indicate

management is attempting to manipulate earnings through its
TATA discretionary authority over accrual policy. The presence of higher
accruals and a corresponding decrease in cash often can be an
attempt by management to internally finance its losses.

• M = -4.84 + 0.92*DSRI + 0.528*GMI + 0.404*AQI + 0.892*SGI +

M- 0.115*DepI – 0.172*SGAI + 4.679*TATA – 0.327*LevI

Score • A score greater than -2.22 (i.e., less negative than this) indicates a
strong likelihood of a firm being a manipulator.
 Important Tips

 DSRI > 1 = Possible revenue inflation.


 GMI > 1 = Gross margin is deteriorating and company is more likely to manipulate
earnings.

 AQI >1 = Tendencies of capitalizing and deferring costs that should have been
expensed.

 SGI > or too < 1 = firms under possible pressure manipulate earnings to keep up
appearances.

 DEPI > 1 = Tendencies of assets being depreciated at a slower rate to boost earnings.

 SGAI < 1 = Possible manipulation of earnings to defer costs, especially when

coefficient is negative.

 TATA > 1 = Accruals possibly used to manipulate earnings.

 LVGI > 1 = Reflecting pictures of increase in leverage.

 Committing Type I or II Error


 Type 1 error  possibility of viewing a manipulator company as a non
manipulator

 Type 2 error  possibility of viewing a non manipulator company as a

manipulator

 Daniel Beneish had earlier revealed in his publication in 1999 that more of type 1
error is usually committed by users of this model than the type 2 error.

 Absolute care must be taken by fraud detection experts when making conclusive
decision on the years with M-scores signifying “No manipulation” tendency. This is
because it could possibly lead them to commit either type 1 error.

 Hence, further efforts must be made by the model users to closely examining
the indicators of the 8 factored variables against their above individual decision
rule, even on non-manipulation M-scores indicators before drawing their final
report on the state of the company’s financial reports.
Take Away!


Classified - Internal use

 Take Away…

 The profession needs to achieve a “quantum leap” to redesign audit processes
using today’s technology rather than using information technology to
computerize legacy audit plans and procedures.

 Traditional auditing is getting less effective day by day to keep a check over
the employees and top management. Time has come for us to recognize that
accounting and finance world need fraud-specific methodologies in order to
be effective in fighting against fraud.

 Transforming the Audit: Big data and analytics are enabling auditors to
better identify financial reporting, fraud and operational business risks, and
to tailor their approach to deliver a more relevant audit.
 

Rajiv Gupta.
CA, CFE, CISA, DISA, CCSA
Vice President & Chief Internal Auditor
Diageo India
Skype: rajiv.gupta.tccc
E: rajiv.gupta@Diageo.com
 Automating Fraud Detection in
Financial Statements

Rajiv Gupta, CFE, CA, CISA, CCSA

Vice President and Chief Internal Auditor
Diageo India