Fraud Patterns and Scorecards. Tony Green MSc BA Analytical Consultant

Transcription

1 Fraud Patterns and Scorecards Tony Green MSc BA Analytical Consultant Experian and the marks used herein are service marks or registered trademarks of Experian Limited. Other product and company names mentioned herein may be the trademarks of their respective owners. No part of this copyrighted work may be reproduced, modified, or distributed in any form or manner without the prior written permission of Experian Limited.

2 U.K. Application Fraud (Source National Hunter) 2

3 U.K. Value of Fraud 253,242,204 of fraud prevented over a 7 year period 3

4 U.K. Application Fraud Trends Automotive fraud has seen a 42% year-on-year increase to Q

5 A CEE Survey A survey by KPMG of 25 companies in the Automotive Industry in CEE showed that 20% experienced a fraud of more than 50,000 in % of these were committed by external parties. 5

6 Organised Fraud A dentist in Cosenza, Italy used EU funds to buy a yellow Ferrari Testarossa, which sells at around 200,000 and a Formula One car, along with 55 other luxury cars, which he stored in warehouses. He received EU money by inventing a solar-panel business that never saw the light of day. 6

7 Automotive Application Fraud Impersonation Fraud Dealer s Diligence Dealer Fraud Non Staff Dealer s Impersonation First Vehicle Disclosure Party Collusion Fraud Shipping Diligence Fraud Fraud / Broker Fraud Fraud Adverse Unscrupulous Vehicles Fraudulent Is Large increasing intensifying numbers information are dealers leaving and brokers particularly of often fraudulently and such forecourts and as undisclosed introducers before brokers involves income obtained all and are organised cars the are being employment previous necessary especially providing closed criminal false down addresses information documentation only gangs manipulation prestige to using set brands, are up in counterfeit due order again being and are to fraud to under the illegally hidden obtain more checks different documents economic shipped or have higher from trading climate that been UK commission are names and completed becoming stricter / false increasingly lending mainland loans criteria European more obtainable ports and difficult to identify Vehicle Shipping Fraud First Party Fraud Staff Collusion / Broker Fraud Non Disclosure 7

8 Preventing Fraud The only sustainable level of long term prevention is to operate rigorous review of credit approval and claims handling and the careful checking of identity and credit status both corporate and individual. The timely exchange of information about fraud and those engaged in fraudulent activities (within the constraints of our data protection laws) is also vital to fraud prevention and to the long term security of our assets. 8

9 A Starting Point? Data matching or profiling techniques Traditional scorecard and predictive techniques Both (Building a scorecard based on data matching, profiling and predictive techniques) 9

10 What do we need? Data Sources Information that the client may have about their customers Confirmed Frauds-- customers who have stolen from them before (names, addresses, phone numbers, etc) Relationship Information-- people who they have done good business with in the past. People who applied and were declined (names, addresses, phone numbers, etc) 10

11 What do we need? Information that is available to verify application information Listings of addresses-- does the property exist? What kind of property is it (a home, a business)? If it is a business, is it a prison? a hospital? a university? Telephone listings-- not only confirm that a number belongs to a person, but the address to which it is registered. Credit information Records of how the identity is being used. Are many companies receiving applications from this person? Is the same address or phone being used for many different names? What types of accounts are on the file? Things like home-loans indicate stability, retail store cards and other products that are very easy to get reflect more risk 11

12 Scorecard methodology Start with a population of verified frauds Build a scoring model based on those characteristics Identify specific characteristics, and combinations of characteristics that are common to the frauds, but less common in the goods 12

13 Benefits of Co-operating Known Frauds 300,000 Finance Frauds 35,000 Insurance Frauds Area Number of Clears Number of Frauds Frauds/Clears % SCOTLAND NORTHERN IRELAND NORTH EAST NORTH WEST EAST MIDLANDS WEST MIDLANDS WALES SOUTH WEST SOUTH EAST GREATER LONDON TOTAL An application in Greater London is 5 times more likely to be fraud than an application from the East midlands 13

14 Improving the Model - Postcodes as a Fraud Indicator Postcode Area Number of Frauds SE28 Thamesmead 3105 SE18 Woolwich 3084 E6 East Ham 2690 IG11 Ilford 2345 SE15 Peckham 2259 E17 Walthamstow 2139 N17 Tottenham 2096 E16 Canning Town

15 Profile of a Fraudster? Male 15

16 Profile of a Fraudster? Male Between 23 and 28 16

17 Profile of a Fraudster? Male Between 23 and 28 Lives within the M25 17

18 Profile of a Fraudster? Male Between 23 and 28 Lives within the M25 Works in the Fast Food Industry 18

19 Further Data Analysis FLA figures reveal that London is the UK s motor finance fraud capital, followed by Glasgow and Manchester. Haverford West? Wick? 19

20 Marketing Information? Population Variables YOUNG ELDERLY ASSET POOR ASSET RICH HIGH DENSITY LOW DENSITY LOW INCOME HIGH INCOME TRADITIONAL COSMOPOLITAN 20

21 15 Groups with 67 Subgroups 21

22 Characteristics of Selected Groups Alpha Territory Liberal Opinions Piers and Imogen Most Wealthy Individuals Wealthy foreign nationals Bespoke luxury items Travel first class Specialist advisers Johan and Freya Young, Professional, well Educated Inner City living Smart rented Flats Eat out, Visit theatre High value equipment 22

23 Characteristics of Selected Groups New Homemakers Upper Floor Living Lukas and Keeley Homes built in last 5 years Single Professionals Middle Incomes Older Downsizers Surviving on a budget Managing debts Jamal and Chantel Limited Incomes Less attractive places to live Convenience food on a daily basis Alcohol and Cigarettes are a large proportion of budget Rely on ready cash 23

24 All Groups Group % of Pop Group % of Pop Active Retirement 4.34 New Homemakers 5.91 Alpha Territory 3.54 Professional Rewards 8.23 Careers and Kids 5.78 Rural Solitude 4.40 Claimant Cultures 5.16 Small Town Diversity 8.75 Elderly Needs 5.96 Suburban Mindsets Ex-Council Community 8.67 Terraced Melting Pot 7.02 Industrial Heritage 7.40 Upper Floor Living 5.18 Liberal Opinions

25 All Groups Group % of Pop % of Fraud Group % of Pop % of Fraud Active Retirement New Homemakers Alpha Territory Professional Rewards Careers and Kids Rural Solitude Claimant Cultures Small Town Diversity Elderly Needs Suburban Mindsets Ex-Council Community Terraced Melting Pot Industrial Heritage Upper Floor Living Liberal Opinions

26 FLA Fraud Trends in Selected Groups 26

27 FLA Alpha Territory Fraud trends 27

28 Improving the Model A good model identifies a small portion of the population that contains many frauds and fewer goods. The most important part of a fraud score is what the client does with it Blocks the application and investigates further Backtracks through other applications to look for links 28

29 Fraud Ring Identification Bank Details Address Details Mobile Landline 29

30 Improving the Model Start Revisit with a current population frauds of verified frauds Build a scoring model based on those characteristics Identify specific characteristics, and combinations of characteristics that are common to the frauds, but less common in the goods 30

31 Do You Need Help? I am the man who can 31

32 2009 Experian Limited. All rights reserved.