Harvesting’s Automated Credit Risk Solution for lenders serving rural agriculture communities in emerging markets
We started building Harvesting’s Credit Risk Solution (CRS) more than 12 months ago and have been fine tuning it as we hear feedback from partners, large and small.
We thought it would be a good time to share more details on the system itself:
Harvesting’s Credit Risk Solution
Harvesting’s Credit Risk Solution allows lenders to leverage (1) historical client data (transaction, savings, KYC,etc.), (2) advances in machine learning algorithms, and (3) alternative datasets (climate, earth observations, agronomy, etc.) to build a robust credit risk model.
Our goal is to help lenders develop a CRS that is efficient, accessible, accurate and affordable. With our solution, clients can build multiple models, back test them using historical data, and go live with our integrated agriculture lending suite in minutes.
Leveraging machine learning in credit scoring
There is a trove of material on the web on how and why machine learning is changing our lives, including how much more precise these algorithms can be in terms of predicting a borrower’s credit behavior.
Harvesting’s CRS allows a lender to choose the appropriate machine learning algorithm to fit its risk appetite and portfolio needs.
Leveraging alternative datasets
Harvesting’s CRS leverages the location information of the borrower (if available) to bring in alternative datasets such as remote sensing, agronomic, and geo-spatial data. In frontier markets, where traditional data on potential smallholder borrowers does not exist, alternative datasets can provide lenders important insights into the repayment capacity of a potential borrower.
Harvesting CRS — How does it work?
Harvesting’s Credit Risk Solution is a two part solution: (i) Part 1: we help the lending organization build a model using our secure online tool and (ii) Part 2 : we host the model online and provide credit scores in real-time as new loan applications come through.
Harvesting CRS — Building the Credit Risk Model
Building a credit risk model in Harvesting’s CRS is relatively straight forward as outlined in the ‘pipeline’ described below.
Model Details — The lender uploads historical data on borrowers. If the data includes geo-location information (available in Harvesting’s loan origination app), Harvesting’s software will automatically upload remote sensing, geo-spatial, and other data to enrich the dataset.
Choose Features — CRS starts crunching the data and presents a detailed analysis of each variable. This step allows the lender to see the ‘importance’ of each variable in the model and to decide which variables it wants to include.
Data Partitioning — To run and test the model, a lender has the option to partition its entire dataset into a (i) training bucket and a (ii) test bucket or to upload a new dataset to be a test dataset.
Select Algorithm — At this point, the lender chooses which model it would like to use out of several available on Harvesting’s cloud-based CRS.
Harvesting CRS — Measuring Model Performance
Harvesting’s CRS provides two standards for measuring the performance of the resulting model:(a) Gini Cofficient and (b) Confusion Matrix.
Gini Coefficient — Once a model is built, Harvesting’s CRS automatically back tests it against the data the lender has provided and calculates the Gini coefficient which indicates how accurate the model is. You can read more about the Gini Coefficient and how it gets computed here.
Confusion Matrix — Harvesting CRS also automatically creates a Confusion Matrix which allows the lending organization to further validate the credit risk model it just built. You can read more about a Confusion Matrix here: https://en.wikipedia.org/wiki/Confusion_matrix
Harvesting CRS — Setting Thresholds
While automatic scoring is important, each lending organization is different and so is its risk appetite. Harvesting’s CRS allows each organization to define the thresholds of what constitutes a ‘good’, ‘bad’ or ‘moderate’ score or risk based on its own risk profile and portfolio objectives.
We believe Harvesting’s Credit Risk Solution (CRS) is one of the most advanced credit risk scoring systems available to lenders targeting agricultural borrowers in developing countries today.
If you are interested in learning more about how Harvesting’s Credit Risk Solution can help reduce risk for your lending organizations, please email us at info@harvesting.co
