Increased evidence in recent years has shown the efficacy of psychometric data for assessing consumer creditworthiness. Psychometrics is a proven scientific technique for measuring personal competencies, which are predictive of specific human performance. Psychometric tools usually take the form of professionally developed behavioral questionnaires, which are now being designed for financial services, as predictors of loan performance. In lending, psychometrics provides an interesting opportunity, which many financial institutions are beginning to explore, especially as a means to promote financial inclusion. However, in order to be effective in this new setting, psychometric tools need to be properly implemented into the lender’s loan application process and decision model. We have outlined a few steps to help clarify this process.
1. The initial audit – The first step in the process of implementing a psychometric-based credit tool is for the lender to audit its current process, in order to evaluate whether psychometric data might be effective. Lenders already use very sophisticated credit models, and it may be hard to believe that more data can make a significant difference. Indeed, additional financial data may, in some instances, be “more of the same”, and not always impactful. Psychometric data, however, by its very nature, is not merely more credit data, but a different type of data, which can be meaningful. The key, however, is to identify the right “pain-point” in the lender’s credit model that would benefit from this type of new data. It may be, for example, that the lender’s current approvals are sufficiently accurate, and don’t require more data. On the other hand, there may be many thin-file and borderline applicants for whom additional data could allow the lender to increase approval rates.
2. Setting realistic objectives – Based on the lender’s current situation, stakeholders should set clear and realistic objectives for the potential utility of the psychometric tool. Since psychometric tools nearly always complement the lender’s existing model, KPIs may vary between lenders who use different models and who tap different consumer segments in different geographies. In all events, KPIs should be measurable metrics. While perhaps an obvious KPI lending metric is reduced default rates, lenders may wish to consider approval rates and acceptance rates as well. Often times, for example, lenders approve less than 30% of their loan applicants, and increasing approval rates by even a small margin can have a substantial impact on the lender’s bottom line. Our research, for example, has found that the right psychometric tool can help to increase over 20% new approvals among the underserved.
3. Choosing the right tool – Once the lender’s needs and objectives are clear, the next step is to find the right tool that can meet these objectives. Several psychometric-based credit solutions are already available on the market, and choosing the right one can be difficult. Among other factors, the right tool is considered to be one that has been: designed for the intended purposes (and not adopted from another setting), used successfully in similar situations (with empirical evidence), appropriate for the target demographic, and has defensible/explainable scores. Furthermore, the relevant technical documentation supporting these issues should be readily available. Additional issues, such as whether the questionnaire is multi-lingual, timely, costly, customizable, asks domain relevant questions, and is user friendly, should also be considered. These issues deserve more space, and will be discussed in greater detail in a forthcoming article.
4. Positioning is key – Once the right tool has been found, it is important to position the tool in line with its objectives. For example, psychometric tools that are used to identify good credits and approve more loans among otherwise declined candidates should be positioned in the application process to facilitate this result. Specifically, such a psychometric tool should likely be administered towards the end of the application process, after the lender pre-qualifies the applicant based on its current credit model. In this way, the questionnaire can be administered to a targeted segment of applicants. Finally, the tool should be presented to applicants in a fair and transparent way. Applicants should be explained the purpose of the survey, and that it may be used as partial consideration for their lending application.
5. Piloting the tool – Carrying out a controlled pilot study before starting to use the tool is imperative. Even if the tool has been used and proven elsewhere, there are unique aspects to every lender’s own credit model that should be tested in advance. Pilot studies can be designed to measure the tool’s KPIs, and make sure the results are in line with the organization’s expectations from the start. Because psychometric scores are typically non-transient, tools can be administered to a sample of existing customers and correlated with their past payments, and can also be administered to a sample of new customers and correlated with their future payments. Furthermore, lenders may wish to carry out multiple pilots and/or compare results from different segments of their customer base.
A pilot is also a good opportunity to highlight potential logistical and technical issues associated with questionnaire administration, while still in controlled use, as well as collecting applicant feedback regarding the overall experience. Ideally, a senior manager should take ownership of the pilot to make sure it is given the deserved priority and meets all of its milestones. And, once completed, the results of the pilot should be well communicated to its stakeholders for deciding on the next steps.
6. Using the tool operationally and monitoring performance – Following a successful pilot, psychometric solutions may be customized or recalibrated based on the pilot data, before being launched. When the lender is ready to begin using the tool operationally, it should be integrated seamlessly into the lender’s platform as part of the overall loan application process. It is also important that all relevant team members on the credit team and the customer support teams be briefed on the nature and usage of the survey. In particular, topics such as fairness, privacy, validity, administration, interpretation and decisioning should be clear. Going forward, the tool should continue to be monitored periodically for performance, and its normative scores updated as needed. In general, it is advisable to work closely with the tool’s publisher for all such technical matters.
We hope these brief guidelines provide some important practical information to consider when implementing psychometric-based credit tools, and wish you the best of luck!