CP030: Detecting fraud in data using Excel – 5 techniques for you

In the 30th session of Chandoo.org podcast, let’s learn how to uncover fraud in data.

What is in this session?

In the wake of hedge fund scams, accounting frauds and globalization, We, analysts are constantly second guessing every source of data. So how do you answer a simple question like, “am I being lied to?” while looking at a set of numbers your supplier has sent you.

Here is an email Peter, one of our readers, sent me last week:

Dear Mr. Chandoo

Shown numbers are measurements from a certain industry process. They are sent to me from a subcontractor. I have to validate these numbers. But am I being lied to? Did they make up the values themselves?
Values can be between 1 and 6 per contract. They are not necessarily normal distributed because of different batches.

That got me thinking?

That is our topic for this podcast session.

In this podcast, you will learn

• Quick announcements about 50 ways & 200k BRM
• Introduction to fraud detection
• 5 techniques for detecting fraud
  • Benford’s law
  • Auto correlation
  • Discontinuity at zero
  • Analysis of distribution
  • Learning systems & decision trees
• Implementing these techniques in Excel
• A word of caution

Listen to this session

Podcast: Play in new window | Download

Subscribe: Apple Podcasts | Spotify | RSS

Click here to download the MP3 file.

Links & Resources mentioned in this podcast

50 ways to analyze data course:

• Know more & join us

Fraud Analysis theory & examples

• Benford’s law
• Autocorrelation theory, example
• Discontinuity explained, usage in hedge fund scam data analysis
• Various Excel analysis models

Commonly occurring distributions

• Normal distribution
• Poisson distribution
• Exponential distribution
• Binomial distribution

Excel formulas & features to help you with fraud analysis

• SUMIFS / COUNTIFS formula
• FREQUENCY() function
• Data Analysis Toolpak (enable this free add-in from Developer > Add-ins)
• CORREL() function

Transcript of this session:

Download this podcast transcript [PDF]

How do you detect fraud?

I never had to detect fraud. So when Peter sent that email, I learned about it by researching. Based on the little I know about this field, my personal favorites are Benford’s law, distribution analysis, discontinuity at expected value.

What about you? Do you deal with data that has certain amount of fraud? How do you detect and analyze it? Please share your suggestions & tips in comments area.

Subscribe to our podcast and get latest episodes automatically

Use below link to add our podcast to iTunes. For other options (Android, Windows phone or RSS) click the link below. Thank you.

Other options to subscribe