Imagine that Chris wants to buy a house and needs a mortgage. He applies online and is sent an email by an intern asking to schedule time to discuss his interest. The intern conducts the initial screening conversation, they schedule him for an in person interview during which time he is interviewed by quite a few folks who ask many questions.
Chris is curious because nobody asks about his current job or how he expects to pay the mortgage. Nobody asks about other credit or past credit history. They seem more interested in his current zip code, engagement in his current community and other demographics. They ask if it’s ok to review his social media data – which feels invasive and perhaps irrelevant, but he agrees because he really needs the mortgage.
Chris is surprised when they approve him for the loan. He’s done his own budgeting and it seems risky. But he really wants the house. Approving mortgages is their business, so he places some value in their risk assessment process. Don’t they know the factors that signal / predict successful and unsuccessful creditors?
6 months later, Chris struggles to pay his mortgage and begins to fall behind on payments. He explains his situation to the mortgage manager. They are able to restructure the mortgage, giving him more years to pay it off and smaller payments. He works with their credit counselors and even agrees to attend training to learn more. He struggles and tries to make his mortgage payments through the next year. He means well but he just can’t pay his bills. Finally, Chris files for bankruptcy. He loses his home. The lender has a terrible financial loss but repeats the same exact process with the same person who wants a mortgage. Or maybe they’ll change it up a little and ask some different questions.
Obviously, this isn’t how lenders make decisions about extending mortgages. They’d go out of business. Lending money is a repeatable process that can be studied with data science. They decide there have to be factors that can help to signal which loan candidates have a higher or lower probability of paying their mortgage.
To stay in business and be profitable, lenders need to predict which borrower candidates are a good risk before extending an offer. Once the offer has been extended, all the company can do is restructure their mortgage, coach, cajole, support, train and hopefully manage the borrower to keep them from completely defaulting. How useful would it be to send the risk management group a prediction about the borrowers after the mortgage has been extended? Crazy, right?
It’s too late once they enter the lender’s ecosystem.
Lenders have lots of data their data scientists can study. They are able to find factors that predict the outcome they are looking for – namely, borrowers with a greater probability of paying their mortgages. Lenders have been successfully predicting human outcomes for many decades – with a high degree of reliability. It’s how the good lenders stay in business.
All day, every day, businesses extend job offers to new employees to join their business ecosystem. Like lenders, it is a massive risk to the business’s ecosystem when a new employee is introduced. Businesses are hoping (but they don’t really “know”) the new employee will add value and not be a drain on the system. By month 3, businesses can usually tell if the employee will be an asset or a liability. At that point – like the credit risk example – all the business can do once the employee is part of the ecosystem is to coach, retrain, support and cajole the employee hoping to help avoid them being a bottom-performing employee.
After an employee is hired, it is too late to find out. Like lenders, businesses need to be able to predict – before extending an offer – candidates with a greater probability of being successful in the role they’re being hired for.
It is Possible to Predict Employee Success, Pre-Hire – Like Lenders, Insurers and other Human Domain Areas
Lenders, insurers, and the like are all human domain areas where data science is being applied to predict human outcomes. These same data science approaches are immensely useful predicting employee performance – pre-hire – before it’s too late.
One thing data scientists know is that you don’t always have the data you need inside your own business. Marketing regularly licenses additional data about their prospects and customers to augment what they already have in-house. Lenders pay for credit history and credit score information about people wanting credit, insurance companies license additional data about people they are considering insuring – all with the goal of having as complete a set of data as possible.
Two Problems with Most “Predictive” HR Systems and Approaches
Talent Analytics, Corp. Stands Alone in Predicting Flight Risk and Job Performance Pre-hire – Before It’s Too Late
If your business is spending millions on employees that either leave quickly or don’t perform, reach out to learn more about how we deliver on the promise above. And, don’t settle for solutions that promise to “predict” something after bottom performers have already entered your ecosystem – when it’s already too late.
Greta Roberts is the CEO & Co-founder of Talent Analytics, Corp. She is the Program Chair of Predictive Analytics World for Workforce and a Faculty member of the International Institute for Analytics. Follow her on twitter @gretaroberts.