SOCIAL PERFORMANCE TASK FORCE WEBINAR USING DATA TO

SOCIAL PERFORMANCE TASK FORCE WEBINAR “USING DATA TO BETTER SUPPORT BUSINESS AND SOCIAL GOALS” JACOBO MENAJOVSKY DATA SCIENTIST FOR FINANCIAL INCLUSION SEPTEMBER 30, 2014 1

SUMMARY • Why data is important? • What is the value of collecting client level data? • My data is all over the place! • • Some ideas on how to merge and work across different data sources Mixing PPI with financial and demographic information • From raw data to better customer understanding • • Social Performance and business based segmentations Clustering, Targeting and Benchmarking Product and service design Hypothesis testing • How data and information reshapes the whole organization 2

WHY DATA IS IMPORTANT? • Lower your delivery and your operational costs • Increase your understanding about your customers • Test your hypothesis and theories of change • Measure your results and your social impact • Drive behavior change 3

WHAT IS THE VALUE OF COLLECTING CLIENT LEVEL DATA? • The Progress out of Poverty Index • Measuring poverty outreach • Understand how customers use financial products depending on their poverty situation • Adjust your organizational goals and product offering with more granular information 4

MY DATA IS ALL OVER THE PLACE! • Handling different data sources and levels of aggregation • Benefits of mixing poverty, demographic and financial data 5

MY DATA IS ALL OVER THE PLACE! PPI data a) Customer profile data b) Example Customer Transactional data c) Four different datasets containing different customer information Public data i. e. Poverty rates d) 6

YOU CAN STILL REPORT SOME RESULTS a) • 34. 6% of the individuals are living with less than $2. 50 per day. b) c) • 62. 2% are Female. • Almost half (48. 4%) of the individuals are living in rural areas. 7

YOU CAN STILL REPORT SOME RESULTS a) • Women in rural areas present the highest poverty rate. b) • And they represent 31% of the total sample 8

WHAT IF YOU MERGE YOUR DATA? ! PPI data a) Merged dataset Customer profile data b) Customer Transactional data c) Merge different datasets using a unique customer ID across all datasets Public data i. e. Poverty rates d) Add external public data (in this case official poverty rates) using the Province/Location field available in both internal and external sources. 9

WHAT IF YOU DON’T HAVE A UNIQUE ID? • In this situation the merging should be done by other common identifier across customers or beneficiaries. • • Group name, Branch, Province, Region, etc. • Be aware that when you aggregate information by a “higher” common denominator you will be loosing information. 10

FROM RAW DATA TO BETTER CUSTOMER UNDERSTANDING • Social Performance and business based segmentations • Clustering • Targeting • Benchmarking • Product and service design • Hypothesis testing 11

FROM RAW DATA TO BETTER CUSTOMER UNDERSTANDING • Social Performance and business based segmentations a) b) c) d) 12

From raw data to better customer understanding d) Poorest Less poor On the one hand, on average less poor customers are borrowing higher amounts Average saving balances c) Average loan size (1 st. Cycle) Social Performance and business based benchmarking Poorest Less poor On the other hand, poorer customers are saving slightly more than the less poor ones. 13

FROM RAW DATA TO BETTER CUSTOMER UNDERSTANDING CLUSTERING AND TARGETING Cluster #3 Total borrowed Total savings Total # loan cycles Initial savings • 9% of the customers who enrolled in the program showed poorer performance measured by the total amount borrowed and saved; the total number of cycles and the initial savings balances (see cluster 3 in blue). • Where are those customers? 14

TRYING TO UNDERSTAND THE LOW PERFORMERS (CLUSTER #3) IS IT SOMETHING RELATED WITH THE BRANCH? Cluster #3 • 91% of the customers in cluster 3 (low performers) enrolled in Branch 1. • All three clusters showed very similar poverty levels. • Why these customers didn’t perform as well as the rest? 15

FROM RAW DATA TO BETTER CUSTOMER UNDERSTANDING INSIGHT TO SUPPORT PRODUCT (RE)DESIGN Initial deposit amount a) b) c) 27% 9. 2% Average monthly increase in savings (%) Customers with at least 6 months as savers • Only 37% of the customers are actually saving money. • Is this saving product satisfying every customers’ needs? • Poverty situation doesn’t look like the reason. 16

FROM RAW DATA TO BETTER CUSTOMER UNDERSTANDING INSIGHT TO SUPPORT PRODUCT (RE)DESIGN • Are savings somehow tied to borrowing behaviors? • Most “decreasers” do not engage in more than 2 borrowing cycles. Customers with at least 6 months as savers 17

From raw data to better customer understanding Hypothesis testing • Does poverty have an effect on credit size? The answer is… YES! a) Average loan size (1 st. Cycle) The less poor customers, showed significantly higher loans. 18 Poorest Less poor

From raw data to better customer understanding Hypothesis testing • Does poverty have an effect on size of initial deposit? The answer is… YES! a) Average initial deposits The less poor showed significantly higher initial deposits. 19 Poorest Less poor

From raw data to better customer understanding Hypothesis testing • Are there significant differences between gender and age groups among my customers? a) The answer is… NO! There are similar age distributions among both male and female customers. 20

HOW DATA AND INFORMATION RESHAPES THE WHOLE ORGANIZATION IT • Essentials for data analysis • What it takes to be more data-driven • Different roles and responsibilities Action takers Customer Analysts Decision makers 21

Q&A • Thanks! 22

CONTACT INFO Jacobo Menajovsky Data scientist consultant for financial inclusion jjmenajovsky@gmail. com 23