Using Big Data for small business loans | mb.com.ph | Philippine News
Home  » Business » Business Tech » Using Big Data for small business loans

Using Big Data for small business loans

Business understanding:

The countryside’s potential is hampered by lack of credit. Startups and small business owners that need the capital are denied access simply because they have no prior banking or credit history. With inherent risk of lending to the “unbanked” AND the fact that loan amounts are typically smaller makes it uneconomical for lending institutions to service them.

Things may be changing with the advent of social media, rising mobile usage and big data analytics. Lending institutions can leverage technology to service this large but overlooked segment. Innovative lenders are enjoying success by using big data analytics and unconventional data sets to assess credit worthiness — at a higher lending rate too – enabling companies like Kabbage to be among the top three small business lender in the US with over 12 billion pounds in loans.

Data Understanding:

The key is to substitute the lack of prior credit history with the borrower’s digital. These digital footprints provide lending institutions with an alternative to assess loan risks. In addition, the use of big data analytics lowers the cost and time to process such loans. Indeed, McKinsey and Co. estimates that “Big data analytics can reduce the marginal costs of providing a $200 loan in Tanzania by more than 40 percent”

Lending institutions ask for access to borrowers’ data in exchange for granting loans. This enables lending institutions to monitor the loan “across a continuum”- not just at the start, but throughout its life.

Cellphone records can be a proxy for measuring disposable income of the borrower. In addition, Eagle  showed that the diversity in the social relationships in a community was related to economic prosperity.”  A paper published in Nature   found that lifestyles can be inferred off mobile location data of users.

Shipping records from 3rd parties like UPS/FedEx show the actual business activity of the borrower and in some cases, the borrowers’ profitability simply based on the majority of destinations of the shipped items.

Social Media Profiles of businesses that are ACTIVE may indicate the level of business attention to maintain its online reputation. So companies like Funding Circle look for Customer Reviews & Engagement, Customer Service and thought leadership in social media.

Accounting records show how well a business is doing and how well it handles its cash flows. Ageing of payables reveal how promptly liabilities are paid. Loans to Asset ratios can reveal the overall financial health of a company.

Data Preparation:

Lending institutions enter the snapshots of the above data into their systems tied to a single identifier for the borrower.

Cellphone records track how many message of each type: SMS, calls and bandwidth are used each month

Shipping records track the shipment frequency, shipment destinations, and declared amounts.

Social Media tracking might include frequency of updates, number of followers, number of comments, number of shares and the number of likes.

Accounting data track frequency of entries/updates, loan to asset ratios and ageing of payables.

Modeling:

The lending institutions then use a supervised machine learning on historical loan performance to predict and create a model. Key questions/insights to answer questions such as: What are the key factors that would signal red flags? What are the key factors that increases the likelihood that a loan will be repaid?

The model may come up with insights such as the ones feature by Cisco:

Shipping patterns can be a good indicator of credit risk- online sellers that ship to California customers make better borrowers.

People who use Apple products are a good credit risk.

Businesses with active Facebook pages are less likely to default on their loans.

Loans seekers using first-generation email accounts raise red flags.

Evaluation:

While the efforts to do serve the ‘unbanked’ are daunting, the payoff is immensely rewarding. An example is M-Shwari which successfully merged non-traditional data set to serve the ‘unbanked’ segment.  It reached “7.2 million unique consumers with its bank-linked saving accounts within 24 months of its launch, which allowed it to disburse 20.6 million cumulative loans (with a 2.2 percent default)”

The author will be presenting other detailed case studies at the ROOTCON 2016 in Taal Vista Tagaytay on Sept 23, 2016. Details here: https://www.rootcon.org/xml/rc10/register