projecteverest

[Proposed Experiment]: SoCon, Malawi, Existing Customers Risk Profiling Metrics

Lean Phase: Customer Segment

Assumption: Creating a Risk Profiling method will reduce the current default rate by ensuring that loans are only provided to those who will be able to pay it back.  By interviewing existing customers with a mix between defaulting and non defaulting we will be able to gain metrics which will be used in our algorithm to test if we can give out a payment plan. We assume that there will be similarities in characteristics between defaulting customers which will allow us to solidify our current weightings.

Time Box: 2 weeks (or time to profile 15 defaulted customers and 15 non defaulting)

Success Metric:

  1. % of defaulting customers would fail the risk profiling survey
  2. % of non-defaulting customers would pass the risk profiling survey.

Criteria:

Success point: More than 70% of defaulting customers are correctly identified by risk profiling tool as not being loan worthy. More than 70% of non defaulting customer would pass the risk profiling and be loan worthy.

Green Light- Proceed to use risk profiling when deciding whether to give loans to new customers 

Orange light point: -Between 50-70% of existing customers surveyed are correctly identified by risk profiling tool as not being loan worthy.

Orange Light: Change weighting of certain aspects and potentially remove or add new questions

 Failure Point Less than 50% of defaulting customers surveyed are correctly identified by risk profiling tool as not being loan worthy

Red Light- Completely overhaul risk profiling and develop new metrics for more reliable profiling. Potentially look into what else could be causing such a high default.

 

 

Experiment build:

The risk profiling tool will be in the form of an online survey, which will link to a Excel spreadsheet, which will determine from the answers given, whether the person is a suitable loan candidate

  1. 1.     Who are they?

·       Name?

·       Phone number

·       Location/Village

·       What loan plan are you interested in?

 

2.    Can they repay?

·       Occupation

·       When do you get paid?

·       Is your income impacted by seasons?

·       What is your monthly income?

·       Weekly expenses related to product

·       Do you have access to Mobile Money

 

3.    Will they repay?

·       Do you have kids who are school age?

·       Have you ever had a loan before?

·       Who was the loan with?

·       Did you repay it?

·       Do you have any loans that are outstanding right now?

·       How much is your outstanding loan for?

 

edited on 11th January 2019, 08:01 by Olivia Dressler-Smith

Fiona Aaron 2 months ago

This sounds great. Surely though that even if only 30% of risky customers are identified by the tool, it's still better than no one being detected? I almost feel that less than 50% shouldn't be a failure necessarily but a suggestion that this might work and needs greater tweaking or changes than the orange light metric.
Additionally, have you looked into how banks or other micro-finance companies assess risk and how successful their methods are?

Users tagged:

Reply 1

Share