June 25, 2019
What is Ocrolus?
Ocrolus is a smart automation and analytics company that transforms images of financial documents into machine-readable data with over 99% accuracy. We power underwriting processes for leading online lenders like BlueVine and Kapitus, and modernize a variety of document review workflows across the financial sector.
What was the inspiration behind Ocrolus? How did you first recognize a need or opportunity to focus on this area of fintech?
We started with a problem: image recognition technology can’t analyze bank statements and other semi-structured documents with perfect accuracy. As we were conceiving our product, we were inspired by Google’s clever use of humans around the world to help digitize old texts. Google set out to archive decades worth of books and magazines using Optical Character Recognition (OCR). However, the machine-only approach resulted in many words being read inaccurately due to variance in fonts and image quality. A crucial breakthrough occurred when Google began leveraging CAPTCHA tasks (i.e. ‘prove you’re not a robot’) to fix potential errors via crowd labor. The combination of OCR, human review, and algorithmic quality control enabled Google to transform any image that could be read by the human eye into perfectly accurate digital data.
Ocrolus took the same concept and applied it to financial services. Our solution blends artificial intelligence and crowdsourced data verification to analyze documents of any format or quality with over 99% accuracy. Then we use the verified data to train our systems to become smarter, mitigating the need for human intervention and continuously improving analysis speed.
Our initial customers were accountants and attorneys, but we knew from the onset that the product had many potential applications. In mid-2016, we had the opportunity to pitch the CEO of Kapitus (previously called Strategic Funding Source), who helped us learn about the underwriting process in small business lending. It became immediately apparent that Ocrolus had the potential to modernize the way lenders evaluate borrowers.
We got to work right away on an API product designed specifically for lenders, complete with analytics that we developed in collaboration with the Kapitus team. For the last three years we’ve focused on enhancing underwriting automation, and have started expanding beyond our core small business use case into consumer, auto, and mortgage lending.
How is Ocrolus changing the underwriting process? What is unique about the technology to enable this?
I’m a big sports fan and outspoken supporter of the ‘Moneyball’ movement toward analytics-based management. For many years, pro teams (1) did not collect much data, and (2) didn’t build a culture around leveraging data to drive actionable insights. When record-keeping became more robust and teams invested in analytics for the first time, the learnings were powerful. For example, in baseball, it’s effective to shift the position of your fielders when facing certain hitters. In basketball, a long two-point shot is a bad shot. Assertions like these used to be debatable, a matter of speculation or philosophy or preference. Now, they are science.
Ocrolus is changing the underwriting process in a similar way to how Moneyball changed sports. Historically, lenders used a limited number of data points to evaluate credit. Underwriters performed “spot check style” analysis, skimming documents and manually calculating a handful of credit model inputs. Lenders sacrificed thoroughness and data accuracy in favor of speed and reductions in labor costs. Our platform makes it cost-effective to quickly retrieve 100% of the data from financial documents with over 99% accuracy. Using Ocrolus, lenders no longer need to work with incomplete or inaccurate datasets.
We are the first data capture company to achieve dead-accurate results for every submitted file, thanks to our unique human-in-the-loop validation engine. The data and incisive analytics we generate provide a foundation for more robust data modeling than ever before. In partnership with Ocrolus, lenders can run retroactive analyses to determine which inputs are the best predictors of creditworthiness. We enable our customers to build next generation credit models that ingest and analyze more data, by orders of magnitude. Ocrolus is bringing the Moneyball movement to the lending industry!
What’s next for Ocrolus?
We still have a tremendous amount of growth potential in our core use-case; optimizing loan underwriting operations. We’re excited to deploy our solution more broadly in the lending space, across various asset classes. Simultaneously, we are beginning to pursue new opportunities more aggressively. There are two in particular that I’m particularly excited about:
- If I weren’t a CEO, my dream job would be…
Head of Basketball Operations at an NBA franchise. We’d invest heavily in data science to inform our decisions. We’d build a roster of positionless players who can defend and shoot 3s, and win many championships.
- What is your favorite source for news…
I’m a big fan of our company’s “industry news” Slack channel. I’m consistently impressed by the quality, relevance and breadth of news stories posted by our team.
- What’s a great piece of professional advice you’re receiving…
If something comes across your desk that will take less than sixty seconds to complete, do it immediately. Your backlog of tasks will pile up faster than you realize; it’s critical to knock out simple tasks with the utmost efficiency!