TransUnion presentation at the Chief Analytics Officer, Africa 2016

CORPORATE MARKETING

Improving the view of thin file customersFrans PotgieterAlternative Data in Credit RiskSeptember 2016

# | TransUnion. LLC All Rights Reserved

1

AgendaDefinitionsThin File PopulationAlternative Data SourcesCase Studies

2016 TransUnion LLC All Rights Reserved | #

2

DefinitionsWhat is alternative data in credit risk?

What is a thin file consumer?

Lets look at two examplesData that is not included in a traditional credit reportA consumer with no payment profile lines, but possible enquiriesA consumer with only recent payment profile history or no open trades12


3

Thin File Population - Meet John


4

Thin File Population - Meet Jack


5

Thin File PopulationDataJohnJackAge2626GenderMaleMaleMarital StatusSingleSingle# Trades00# Enquiries00Owner/TenantTenantTenantOccupationManagerManager


6

Thin File Population South Africa47%


7

47%

Thin File Population South Africa4.6m


8

47%

Thin File Population South Africa3.6m


9

What do we know about the consumer without alternative data? From a bureau perspective:AgeGenderEnquiry InformationOther Demographic Information

Additional Data Sources


10

What do we know about consumers with alternative data?Additional Data SourcesWhere they live and we can determine the risk within a specific group, for example the risk associated with the suburb they reside inPublicly available informationMobile informationCommercial informationMultiple Choice/Psychometric credit scoringEducational dataSocial Network dataClub memberships


11

Additional Data Sources Challenges PoPI ActLegalStabilityOn-going availabilityHistorical data


12

Software utilised to gather information on an ongoing basis.

Case Study - Brazil Alternative data and enquiry information used on the whole population, no payment profile information.

SOURCES USED:

Public informationPrivate data sourceTransUnion enquiryObtained a Gini of 41.25


13

Approach:Case Study South Africa Obtained data not yet available to prove the value within credit riskCreated aggregated views to draw conclusions from similar groups of the populationApplied different methods to existing demographic dataIncluded alternative data not historically considered for credit scoring


14

An account origination sample was used during the analysisThe case study focused on an industry where high volumes of applications do not have traditional bureau information, i.e. consumers that have thin bureau profilesDuring the analysis a distinction was made between two types of thin bureau consumers:THIN: no trades on file, but may have information such as demographics and/or enquiries

THIN2: some trades on file, but all accounts are either very new or there are norecent updates on any trade (i.e. only inactive trades)

Sample Used For Analysis


15

The performance of the following scores are showcased:

Score Performance Analysis A score on CreditVision V1 combined with alternative dataA score on alternative data onlyA new-generation bureauscore - CreditVision V1An existing champion bureau score - Traditional Bureau


Score Performance Comparison

210% Improvement


17

Assessing thin customers using CreditVision V1 and alternative data:Same Size on Low Risk

Same Bad Rate on Low Risk

The value of alternative data for thin bureau customers


18

Assessing thin customers using CreditVision V1 and alternative data:Same Size on Low Risk

Same Bad Rate on Low Risk

The value of alternative data for thin bureau customers


19

Large portion of SA population is classified as thin file

True portfolio growth should come from: Thin file customers and new market entrants

Being the first to provide credit to a consumer results in long-term loyalty

Alternative data is very predictive in credit risk and assists in identifying your future good customersConclusion


20

QUESTIONS?

vvv


TransUnion presentation at the Chief Analytics Officer, Africa 2016

Data & Analytics