Automation is the great equalizer when it comes to competing with giant lending companies.
Predictive analytics isn’t new. Machine learning isn’t new. However, data science can be complex and not something the average non-scientist online lender can manage easily. That’s changing due to a new autonomous approach to predictive analytics using artificial intelligence as a core technology allowing lenders to reduce the development of actionable and valuable metrics from months to days.
DMway Analytics provides an autonomous predictive analytics solution powered by machine learning that enables subject matter experts without data science knowledge and experience to build their own predictive models in a fraction of the time it takes traditional models. Here’s how they do it.
Democratizing Predictive Analytics
“We started a couple of decades ago,” said Gil Nizri, CEO of DMway Analytics. “Back then, a lot of algorithms were created for the financial industry. Even then, many data scientists realized that one day we’ll be able to automate algorithms and create algorithms in a couple of clicks.” But it took a while before demand for the technology caught up with the scientific developments that make it possible. “Democratizing predictive analytics took a couple of decades because it’s complicated. Now, what was done by human data scientists can be done by a machine.”
DMway’s mission is to level the playing field for small lending companies by making predictive analysis easy and available to non-scientists who are subject matter experts in their business specialties. One of their verticals is alternative lending. They also serve the financial services, marketing, insurance, telecommunications, and utilities industries.
Some of the questions predictive analytics can answer for online lenders include “who is likely to default on their loan,” “how many people will default this year,” and “who is a good credit risk?”
“Not everyone who works at a lending company has knowledge of predictive analytics, data science, or algorithms,” Nizri said. “Some lending companies simply can’t afford to hire people who can build complicated algorithms and adjust them as needed. That puts smaller lending companies at a disadvantage when competing with larger lenders like CitiBank, Wells Fargo, and other legacy institutions.”
With that in mind, DMway Analytics created a solution that allows small lenders to compete with large lenders in an area that is increasingly more essential to successful lending. By equalizing the playing field, they are democratizing predictive analytics.
Problems Solved By Predictive Analytics
Nizri breaks machine learning for predictive analytics down to three key technological techniques:
- Classification – When you want to classify a small population among a larger population. Who will likely pay on time? Who will likely default the loan? Classification involves any use case that fits into that family of problems.
- Expected Value – When you want to predict the future, what is the expected value of the thing? For instance, the lifetime value of each customer and what interest rate should be charged for each individual. These can be solved by the DMway algorithm.
- How Many Times an Event Occurs – Identify the event you want to count—for instance, loan defaults—and count the number of times it happens within a given time frame.
Once an algorithm is created based on a lending company’s criteria, it becomes automated so that loan application decisions can be made almost immediately, Nizri said. The process can also reduce fraud prevention. If a company can count how many times fraud occurs, and under what circumstances, they can devise a strategy to prevent it. These three predictive models can address 90 percent of the problems lending companies face, according to Nizri.
“By simplifying the creation of predictive models, any loan expert can do this without knowledge of data science complexities,” Nizri said.
DMway Origins and Business Model
DMway officially launched in January 2016 with $1 million in seed money from JVP Media Labs. Prior to that, the company bootstrapped itself from conception to funding. The funding allowed them to go to market with their first product, but there were alpha versions prior to 2015. Since the initial seed round, the Israeli Office of the Chief Scientist (now called Israel Innovation Authority) has invested a couple of hundred million dollars into the startup, as well, giving DMway a huge boost.
The predictive analytic solution is sold as a subscription. Companies pay an annual fee based on the number of users. While lenders are the primary target market, not all customers are lenders. The solution can predict other company data, as well, Nizri said.
Nevertheless, financial services startups were the first companies to adopt DMway’s technology, by design. Because they need to compensate for less manpower, the automated models wrapped up in the solution saves them money and makes them more competitive.
“As a startup you don’t have a lot of capital,” Nizri said. “That’s why fintech companies were the first to adopt.”
After one year, DMway has 10 lending company customers. Most of them are on the higher end of mid-size, Nizri said. Some of them turn over a couple of billion dollars per year. Among the list of clients Nizri mentioned are Direct Finance and Backed Inc. Companies use the platform to predict lending trends, loan default probabilities, and fraud. DMway also provides full underwriting automation and loan approval through its platform.
“When the entire process is done by machine learning algorithm, you can handle a lot more loan applications in a better and more secure way, then you can mitigate risk better than when loans are processed by human underwriters,” Nizri said.
DMway’s co-founders include Nizri, CEO; Professor Jacob Zahavi, chief analytics officer; and Dr. Ronen Meiri, chief technology officer. Zahavi was the first person to ever discuss machine learning algorithms and has been working with them for over twenty years. Nizri is a veteran evangelist of predictive analytics.
How to Be A Competitive Lender
Most predictive analytics tools are developer tools meant to be used by data scientists, but for small lenders who do not employ data scientists, predictive analysis may be out of touch.
“We are removing every barrier of entry for the world of predictive analytics,” Nizri said. “One of those barriers is the need for data science, machine learning, and predictive analytics knowledge. Users of our product do not need any of that knowledge.”
The DMway platform mimics the way a human scientist works and generates a state-of-the-art visual in about three minutes. Nizri said it’s as good as any human-made algorithm. The platform generates out-of-the-box reports and interfaces data science with business users so the non-scientist can better understand the root causes of problems and how to mitigate them. To prove his claims, Nizri benchmarked his company’s algorithm against human-made algorithms and found them to be as good as or better in every controlled situation.
“It’s more than automation,” he said. “It also includes intuitive, heuristic algorithms along with knowledge based on four or five decades of study by a large number of data science veterans. It would take any data scientist 10 or 15 years to reach that level of knowledge and experience. ”
It typically takes a human scientist three to 12 months to create a predictive analytics model and complete a project. By the time a company reaches a conclusion based on the data, it’s no longer relevant. By speeding up the process using machine learning algorithms, DMway levels the playing field and makes predictive analytics more relevant for everyday uses. Nizri believes that once his company has paved the way, other companies will enter the playing field to challenge them.
“If you currently provide loans and your algorithm is bad, it will take your data scientists 3-6 months to improve it, then you are in deep trouble,” he said. “You’ll underwrite bad loans everyday. With DMway, you can have a state-of-the-art algorithm and provide great loans to the marketplace and start profiting within two to three days.”
The Future Belongs to Machine Learning
While machine learning isn’t new, what is new is the rapid pace at which it is forcing innovation in financial services. More and more alternative lenders are implementing machine learning technology into every part of the process, from loan application review to underwriting. As more companies adopt machine learning technology, the more necessary the technology becomes to remain competitive. It is even more important for small lenders because every minute and dollar they can save on the process makes them more level with lending giants and ensures they remain alive in the marketplace.
“Even Lending Club is small compared to the giants,” Nizri said. “The difference they have is the ability to be agile, flexible, and creative. This is what machine learning algorithms give you.”
Nizri sees an effective machine learning algorithm as the difference between alternative lending leaders and run-of-the-mill players. In order to survive, lenders will have to have the best tools available. It’s more important than human talent, which will likely go to the larger companies that can afford to pay their salaries.
For Nizri, augmented analytics is the future, and he is proud to be the head of a company on the forefront of the avant garde. To remain competitive and grow as he sees the company doing, he’d like to see more startup funding and a strategic partner.
“That will help us boost our global sales and offer more value as an industry leader,” he said.
Written by Chris McElroy and Allen Taylor.