Analysis Featured

Using Machine Learning for Risk Management in Online Lending

online lending

The financial services industry has entered into an arms race fuelled by machine learning and AI. Artificial Intelligence is an umbrella term for computers able to execute on a higher level of human mental functioning. Machine Learning is part of that AI; it allows the computer to learn from its functioning without being explicitly programmed. The machines will get better at their jobs without humans needing to code that improvement.

AI has wide-ranging applications across the business spectrum, but nowhere is it being exploited with as much zeal as in FinTech. According to CB Insight reports, more than $5 billion has been raised by AI start-ups globally in 2016. This dwarfs the over $3 billion raised in 2015. CrowdProcess is a young data science startup from Lisbon with a new take on machine learning and transparency.

The Genesis of CrowdProcess

CrowdProcess was founded in 2013 by Pedro Fonseca, Jaoa Menano, and Samuel Hopkins.

Fonseca, head of data science and CEO, started his career in FinTech at the age of 25. He also co-founded ODMLX, one of the largest open data organizations in Lisbon. Menano, co-founder and CFO, has technical expertise in data science and worked with Fidelidade, an insurance company in Portugal as project manager. Hopkins, CTO and co-founder, is proficient in machine-learning. He has worked at three different start-ups performing different roles as an application programmer, system administrator, and applied data scientist.

CrowdProcess’s headquarters are located in Lisbon and New York. The company has raised $1million in seed capital funded by Seed Camp, Rising Ventures, Kima Ventures and Frontline Ventures. It is currently in a new round with 50% of capital already committed by investors.

How CrowdProcess Taps Into AI Innovation

CrowdProcess offers a SaaS analytics and data decisioning model for the credit space. They help develop, deploy, and monitor credit scoring models. Using AI and machine learning, they can reduce default rates by almost 30% or improve acceptance rates by over 10%. Its USP is the ability to operate without black boxes.

A black box is a system where the inner components, or logic, is not available for inspection. Due to its existence in machine learning models, it hinders the process of regulation. Adverse Action regulation requires the lender to inform the borrower that an adverse action has been initiated on their credit application, and why. This provides more transparency throughout the process.

James, its flagship solution, is a one-stop shop for credit risk management. Lenders gain access to James around the clock. To ensure reliability and avoide vulnerabilities, it is fully secure and highly encrypted. James’ cloud infrastructure is scalable, and its potential can be enlarged to accommodate hockey stick growth.

James’ features include:

  • Scorecard: Traditional scorecards have been augmented with a random decision factor, gradient boosting, and neural networks.
  • Validate: It provides precautionary alerts according to the model`s metrics and access to data necessary for validation. James has an edge over other solutions in the market, which may not enable regulators to validate.
  • Deployment: James has superiority with regards to the timeline of deployment. The model can be run automatically or in conjunction with other models. So a 6-month integration time is essentially reduced to zero and the company can compare multiple models in parallel.
  • Monitoring: James monitors its own performance in real-time. The system not gives out early warnings and helps ensure data sets are used appropriately. When a material change occurs, the software is tweaked accordingly.

CrowdProcess charges a fixed fee, but its flexible depending on the volume. The company invests a lot on UI and UX removing the need for hiring coders and engineers. A financial analyst with sound knowledge of credit models will be able to use the model without any external support or requirement to code. This young startup has been able to snap established clients like Evo Banco (owned by Apollo Global Management, a bank in Spain) and Cofidis (a consumer financing house, in Paris).

The Value of Machine Learning

In order to help customers know the real value of its model and smooth out the onboarding process, CrowdProcess has launched a 1-day onboarding workshop known as Jumpstart. Customer data is cross–validated with the company´s predictive analysis. Through this back testing, customers are guided toward the proposed benefits of saving time and money in preference to traditional predictive machine learning models. CrowdProcess has undertaken 15 jumpstarts that have resulted in 100% positive outcomes and has beaten existing models.

James has the ability to outperform incumbent models in this sector. Realizing the potential, CrowdProcess is on an aggressive sales roadshow in the U.S. Its USP is not only the ability to offer a better model, but a solution which is transparent and will not turn out to be a regulatory nightmare for its clients. The Portuguese company has major tailwinds and has the chance to capitalize on this massive market.

Author:

Written by Heena Dhir.

Testimonials

default gravatar

"Your daily letter is great!" , Ram , Founder and CEO, PeerIQ

default gravatar

"Hi George - just want to tell you that you are doing a great work with Lending Times;-) Brgds, Kasper" , Kasper, Partner and Co-founder at Dansk Faktura Børs A/S

default gravatar

"I've been following your newsletter for some time now and have been very impressed with the content." Charlie,Co-Founder | Bolstr

default gravatar

"Hey George, I must say I really enjoy your site. It has inspired me to do some changes at our platform and we are the biggest consumer lender in Sweden." , Ludwig, CEO @ Savelend Sweden AB

default gravatar

"Your daily email is very useful. It gives quick update on what's going in the market. Thank you very much for all that info." Yann Murciano, Head of Base Metals Trading at Morgan Stanley

Our daily p2p news digest