In a year hallmarked by un- certainty, accurately predict- ing what may lie ahead isn’t just wishful thinking – it’s an
important business strategy. This
type of intelligence can separate
the best-performing credit unions
from the pack in the low-rate envi-
ronment where greater threats of
credit losses loom.
Understanding where thetrends are pointing doesn’t comefrom lucky guesses, gut instinct orfeelings based on past experience.The most powerful insights andforecasts are derived from dimensional, data-driven analysis.
The power of predictive analytics is in advanced statisticalmodels and machine learningalgorithms that deeply analyzemember data and see what thehuman mind can’t. Especially intoday’s climate, credit unions’interest in predictive analytics isgrowing, particularly in three keyareas: Pinpointing risk, understanding changes in behavior patterns and finding the right placesto increase income.
Identify the Greatest Areasof Risk
Many credit unions rely on a com-
bination of intuition and single,
specific metrics to assess areas of
risk. Analytics-driven models take
risk analysis to the next level with
processes that are much more
thorough, objective – and even
more exciting – automated.
A predictive analytics model
looks at historical data in mul-
tiple dimensions, and then stud-
ies the relationships between
those dimensions. Models that
are powered by machine learn-
ing continuously “learn” from
the data, and update and adjust
forecasting scenarios automati-
cally. This means that predictions
are not only more precise and
detailed, but they
continue to become
more accurate and
enriched over time.
Credit unions willreap significant benefits by using predictive analytics to forecast delinquenciesand charge-offs. Thisgives the credit uniona fuller understanding of its membershipand allows segmentation based on riskprofile.
Recently at CURise, we created adelinquency riskmodel for a largeMidwest credit union with morethan 250,000 members. The credit union wanted to build out separate collection models for eachof its product types and generatemember-level risk prediction.The goal wasn’t just to producea list of members that were likelyto pay late, but to identify exactlywho had the greatest propensityto fall more deeply into the delinquency cycle in the next sixmonths. This model will greatlyoptimize collection efforts, dramatically improving both efficiency and effectiveness.
Understand Changes inSpending Patterns
The pandemic has introducednew unpredictability into spending patterns this year. To be clear,even the best credit union predictive analytics model couldn’t anticipate a global pandemic (notyet, anyway …), but it can continuously assimilate new information to understand shifts in behavior patterns and predict howit will impact a credit union’s keymetrics.
Consumer spending plummeted in the spring, but beganto recover over the summer,buoyed by recovery aid andstimulus measures. Groceryspending remains up, but certain categories remain hard-hitby coronavirus fears and economic hardship, such as accommodation, food service,recreational entertainment andtransportation.
Now after its short upwardtrend, the signs of recoverylook uncertain again. Millionsof Americans are still out ofwork, but enhanced unemployment, stimulus payments andthe small business PaycheckProtection Program have ended. There aren’t many signs ofprogress toward new aid packages. This, coupled with the possibility of renewed outbreaks asthe weather turns colder, leavemany heading into the fall andwinter with worry and unease.
The key takeaway for creditunions when it comes to spending patterns is that there aremany factors influencing behavior and the factors are in continual flux. The situation is highlydynamic. High-level trends indicate the likelihood of continued decreases in cash and checktransactions, coupled with ongoing demand for online shopping and contactless payment.But, when it comes to understanding the crucial national, local and individual circumstances driving how and where yourmembers are spending (or not),predictive analytics is critical fortracking and synthesizing thecomplex situation.
Recognize StrategicOpportunities to IncreaseIncome
The combination of credit lossesand low interest rates have mostcredit unions bracing for slimmerearnings. Cost-cutting and creating efficiencies can help – but onlyso much. Though the coronavirushardship has been widespread,it’s not a blanket effect. More advanced predictive analysis can reveal the best pockets of untappedopportunity.
Though many credit unionleaders may have a sense of whereopportunity lies, today’s financialclimate doesn’t afford the luxuryof following a hunch, only to bewrong. At the same time, leadersdon’t need to train themselves onthe nuances of clustering, clas-sification-based machine learning techniques or market-basketanalysis to form data-driven strategies. Instead, predictive modelscan be implemented to continuously mine and study data to lookfor associations, patterns and likely outcomes. These insights makeit clear where targeted efforts willbe the most rewarded.
When a southeastern U.S. creditunion was looking to support itsauto loan portfolio after originations plummeted in the spring, itdidn’t simply roll out a new promotional offer to the entire membership. First, the credit unionemployed predictive analyticsin a valuable effort to determinewhich type of auto offer madesense for which members. Usingthe list of members with a “highlikelihood” to act on a loan offer,the credit union was able to createa series of intelligent email-basedcampaigns using insights from thepredictive models.
Traditional notions of “makingpredictions” evoke speculativeguesses and gut feelings – not thesort of things that inspire confidence or make for sound strategy.The field of predictive analyticsturns those ideas completely ontheir head, using the tools of science and technology to help credit unions analytically, accuratelyand confidently assess what thefuture may hold. n
Predicting the Future Is a Strategy, Not a Wish
Lead Data Scientist
CU Rise Analytics
CU Rise Analytics