First Look: Startup Offers On-Demand Data Scientist for Small Banks

Print
Email
Reprints
Comments (3)
Twitter
LinkedIn
Facebook
Google+

Students of Singularity University have a tough mission: build a company that helps one billion people within 10 years.

The "University," which is located inside NASA Research Park in Silicon Valley and was co-founded by Google executive and futurist Ray Kurzweil, is not an accredited four-year college but a supplemental learning institution that aims to "educate, inspire and empower leaders to apply exponential technologies to address humanity's grand challenges."

Eli Mohamad, a 2012 class graduate of the program, is working toward that grand ambition by building a program that aims to change the way consumers are evaluated for creditworthiness and get pitched offers. The program analyzes anonymized transaction data from credit unions and community banks to provide feedback on pockets of people. The surprising twist Mohamad is working on at his California startup, Walkmore, is that the program will also seek to crunch physical fitness data so that customers who work out receive better interest rates and/or discounts.

Indeed, Walkmore's models claim to show the more a person moves tracked through smartphone apps or wearable computing devices the more likely he is to pay bills on time.

To be sure, the fitness data won't be crunched in a vacuum it will be analyzed alongside more traditional data sources that relate to environment and finance. (All the data will be anonymous.)

The startup seeks to pilot its software with credit unions and community banks for six months.

Most of those tests will center on Walkmore identifying credit decisioning or cross-selling opportunities within partners' transaction data. (Think of the pilot as an outsourced data scientist for a smaller institution that may lack mathematicians on staff.) Walkmore also wants a couple of its piloting partners to collect fitness data.

For those that do the latter, the experience would work something like this: A consumer would enroll in the white-labeled bank Walkmore program and provide access to the stream of data from his self-tracking devices, like FitBit. To incentivize a person to cough up fitness data, he could be offered cash back, better interest rates and discounts from local businesses, suggests Mohamad.

He stresses that people would never be penalized for being sedentary. "You still have access to the original product even if you sit on the couch or simply cannot for some reason increase your physical activity," says Mohamad.

Then, a person in a wheelchair could use a Runkeeper app, for example, to log his physical activity. The smartphone app uses the phone's GPS to track physical activities. "We transform all activities into 'steps,' which is the in-game currency for redeeming cash rewards or discounts," he says.

To be sure, the ambition is bold. Banks are reticent to change underwriting practices or take on more risk as they struggle to make profits while customers could have privacy reservations. Then, there are regulation question marks, too.

Still, Walkmore's concept offers an example of a growing trend of young companies emerging to build software designed to transform underwriting, or at the very least, improve prospecting.

The startup has already tested some of its models with credit unions. It has analyzed transaction data from five credit unions in the U.S. and Canada with the credit union think tank Filene Research Institute's backing to identify groups of people to pitch products to, for example.

In the three-month pilot, Walkmore searched for correlations in gender, credit score, and income. All told, Walkmore applied its models on partners' transactional data on 500,000 people, and reported fewer delinquencies and a slight uptick in loans issued among participants that followed its suggested action items.

The startup aims to give community banks and credit unions insights that show people can be a better risk than their credit score implies. It also intends to clue partners to what customer clusters are likely to need a specific product.

To accomplish this, Walkmore sends a bank a flash drive with Greenbox, Walkmore's analytics software. The bank uploads its transactional data and a Walkmore file and then clicks "run." The anonymized results are sent back to Walkmore to analyze and translate into recommendations for banks.

"It looks like a Windows 95 program," says Mohamad. "We just see the clusters."

Clusters represent groups of people who are more similar to each other than to those in other clusters. Walkmore provides a bank with a report within 30 days that offers recommendations such as this: Cluster Q is incurring fees from Chase by withdrawing money from the ATM at the corner of 1st and 2nd street. By placing an ATM there you could provide a better service to 3,897 members, and save them $14,000 per year. Walkmore's prediction accuracy for such conclusions is 98%.

The time it takes Walkmore to review the data depends on variables such as the amount and quantity of data, the bank's internal procedures, the types of products and the demographics of consumers, Mohamad says.

After completing the upcoming pilot, Mohamad aims to commercialize the technology. The Walkmore vision, according to Mohamad, is nothing less than turning banks into local health promotion centers to decrease occurrence of preventable diseases and enable access to finance to people in need.

JOIN THE DISCUSSION

(3) Comments

SEE MORE IN

'The Law Penalizes the Consumers It Set Out to Protect': Comments of the Week

American Banker readers share their views on the most pressing banking topics of the week. As excerpted from the Comments sections of AmericanBanker.com articles.

(Image: Fotolia)

Comments (3)
I can see a lawsuit coming on this. I am heavy and out of shape. My credit and success are not related to my physical health. Granted I was a good athlete in my younger years, but I believe one has nothing to do with the other.
Posted by robrose | Thursday, December 12 2013 at 4:09PM ET
I suspect Robrose is not alone in his view.
As described, this sounds like it will grab the attention of those responsible for ECOA compliance. Based on info from the website and what is described in the article, disparate impact for some protected classes is at least a possibility (I'm not making any conclusions). In a perfectly objective world, blind to all things that make up the packages we come in, the only things that should matter are our relevant actions. I do not question that the Walkmore intention is anything but good. However, I do think the program could have an unintended negative impact. I hope Walkmore can work out the details within fair lending regulations.
The idea that actively engaged and disciplined people maybe be more likely to follow through on their obligations is not far fetched. I think the problem may be in the narrowness of the metric being applied and how that metric correlates with protected class criteria.
Things like this challenge how we all (regulator, lender, borrower) think about ECOA compliance in a fast expanding, big data world. Thanks Mary, very thought provoking.
Posted by PatrickReily | Friday, December 13 2013 at 12:26PM ET
Thanks for your comments, Rob and Patrick. We've developed Walkmore with compliance in mind, and have worked with domain experts diligently from the design phase in late 2012. We have paid special attention to ECOA, Regulation Z and the fair lending regulatory package, as well as Frank-Dodd. There is only so much you can share in an article, but feel free to approach me - I will be happy to share details with you.

Your bank is not going to penalize you if you aren't physically active. Machine learning simply enables them to use physical activity data and transaction history to provide better fitting products to each customer segment. We've written about applying data analytics to transactions in detail in the Filene Machine Learning Project study mentioned in the article, and you can find more here http://filene.org/research/report/big-data-credit-unions-machine-learning1

Robrose, I hear what you are saying. Our models do break down at higher wealth categories, so you may be in that segment. However, the strength of health and finance correlations and accuracy of predictions hold very well for more than 200 million Americans.
Posted by EliMohamad | Saturday, December 14 2013 at 10:55PM ET
Add Your Comments:
Not Registered?
You must be registered to post a comment. Click here to register.
Already registered? Log in here
Please note you must now log in with your email address and password.

The FinTech 100

FIS and Tata once again top the annual FinTech 100 list of vendors, ranked by revenue; IBM and Hewlett-Packard lead the pack of tech companies serving multiple industries; and Bionym and Silver Tail are among the 10 Tech Companies to Watch.
DAILY ENEWSLETTER UPDATE

A Newsletter featuring Bank Technology News' top stories plus special reports and data

This feature displays payments industry news and analysis from American Banker sibling brand PaymentsSource. Registration is required; for more information contact customer service.

TWITTER
FACEBOOK
LINKEDIN
Already a subscriber? Log in here
Please note you must now log in with your email address and password.