How ML is creating winners in the personal loan space 🥇

Machine learning is rapidly growing as more businesses find ways to use it to automate and improve their decision making. This is especially true for personal lending companies – including online lenders, branch-based lenders, banks, credit unions, short-term lenders and others – where the large volume of available data and the importance of accurate decisions have made it particularly valuable. Personal loans are typically unsecured and the borrower does not have to provide collateral against the loan, increasing the level of risk personal lenders take – highlighting importance of accurate underwriting decisions.

Here are a few critical areas where machine learning is being used by personal lenders. If you’re not familiar with machine learning, see this post for an introduction to the key concepts.

1. Improved Credit Underwriting

Credit underwriting is the process of assessing the risk of an applicant, determining whether to make a financing offer and at what rates and terms. This is a core competency of all personal lenders, and their approaches can be broadly bucketed into three categories:

  • Judgmental: A loan underwriter reviews the application and makes a subjective judgement as to the creditworthiness of the applicant.
  • Rules-Based: Application information, usually in the form of metrics such as debt-to-income, are run through a series of pass-or-fail rules that determine loan eligibility, rates and terms.
  • Model-Based: Applicants are scored by a predictive model that produces a score which quantifies the applicant’s credit risk. Different score thresholds determine overall eligibility and risk-based pricing.

In practice, lenders often use a combination of these methods. For instance, a lender may use a third-party scoring model, such as a FICO Score or Vantage Score, in combination with rules (e.g. FICO Score must be greater than or equal to 660). Over time, underwriting has generally gravitated from judgmental, to rules-based, to model-based, as lenders seek to more precisely risk-rank their applicants and limit exposure to any fair lending concerns (regulations which dictate that lenders must not discriminate). As risk models have become the prevalent method of credit underwriting, machine learning has emerged as the most effective way to create such predictive models.

Machine learning is most effective when many data points inform a given decision. Personal lenders generally have a large amount of data on the applicant at the time of underwriting; typically this data comes from a credit report on the consumer from a major credit bureau, such as Transunion, Experian or Equifax. Historically, lenders have relied heavily on credit scores provided by the bureaus, but increasingly lenders are leveraging a wider array of data to make better risk determinations. As the number of data points available increases and the number of potential relationships expands, the application of machine learning becomes increasingly valuable in order to discover sources of predictive power not available with rules-based underwriting.

Machine learning has enabled many lenders to go beyond just looking at basic credit metrics and allowed them to create advanced risk profiles for prospective borrowers. Particularly for lenders with a high volume of historical loan performance data, machine learning can effectively create credit predictive risk models that better assess applicants and lead to lower default rates on loan portfolios.

2. Better Fraud Detection

It often takes multiple days for lenders to approve or deny a personal loan application, and much of this time is spent verifying application information. Personal lenders typically request additional information that evidences the applicant’s identity, stated income and bank account ownership. Lenders are required under federal regulations to collect some of this information, and it helps to ensure that the application isn’t fraudulent, as losses related to fraud can be a significant concern in personal lending (especially when loans are originated over the internet).

Machine learning can dramatically accelerate the application verification process with predictive models that instantly assess fraud risk. Lenders train these models on historical application data and the models uncover relationships between application characteristics and the likelihood of fraud. The data used is typically a combination of stated application information, verification information collected during the process and third-party information from fraud data providers. With these models, it’s possible for applicants to be immediately analyzed, categorized, and approved or denied – all without any human intervention. In practice, most lenders use models to automate approvals, while flagging high risk applications for further review by a loan officer.

For personal lenders operating online, the risk of fraud is often higher, given the lack of face-to face contact and pre-existing relationship with the applicant in most cases. Luckily, the amount of data online lenders have to train fraud models is also higher, as the digital channel allows a wide range of information to be automatically collected on the applicant’s digital fingerprint, such as device type, location based on IP address, and the frequency of applications made by that device.

Whether in-branch or online, application fraud is a real threat to personal lenders, and the use of predictive models trained with machine learning is growing for this reason. Predictive models can reduce or eliminate the need for manual application review, reducing processing time from days to minutes and resulting in a much better customer experience.

3. More Effective Customer Acquisition

Personal loans are typically time-delimited instruments, with a fixed payback period of a few years. This means that it’s critical for personal lenders to have an effective customer acquisition function to originate new loans as old loans pay off. For personal lenders in the US today, customer acquisition costs – typically in the form of direct advertising expenses – are a major cost and companies are seeking ways to make their customer acquisition more effective.

Many lenders are turning to machine learning to create predictive models that help them optimize advertising spend. For example, sending letters advertising loans to consumers is a typical marketing strategy. Lenders track responses to these mailers and develop a collection of response rate data over time. Applying predictive analytics to this data allows them to determine which potential customers are most likely to respond and let lenders prioritize their contact lists to send mail to the highest potential leads.


Predictive machine learning models are improving personal lending by making it easier to assess creditworthiness, making the verification process faster and enhancing customer acquisition. Because personal lending is a data intensive business, machine learning is particularly well suited to helping lenders to create predictive models that help them make better decisions. In addition, the recent emergence of tools, such as automated machine learning, which increase the ease with which predictive models can be created, makes it increasingly attractive for lenders to adopt machine learning, benefitting lenders and borrowers alike.

About DigiFi

DigiFi is a technology company that helps businesses make better automated decisions.

Our platform lets businesses easily use automated machine learning and rules management to optimize critical decisions with no coding or technical expertise required. Repetitive work that used to take hours can now be completed in minutes, letting your team focus on what matters most.

Learn more at