Planet Money just ran an interesting story about Xerox using predictive models to determine best hires for call centres. Motivated by high turnover rates, Xerox contracted with Evolv to determine employees best suited for call centre work. Prospective employees take an online test that asks questions that may seem unusual for a typical job application. The example given on the podcast was choosing between:
a) Most managers are motivated to make ethical decisions
b) I always ask for help in difficult situations
While these questions do not seem mutually exclusive, a series of these types of questions indicate a prospective employee’s disposition toward time management, friendliness, problem-solving, punctuality, and dependability. Evolv field tested the exam in Xerox call centres and used response trends among top employees to develop their predictive model. Among collection agents, for example, Evolv found that persuasion was not as critical as creativity for top performers, so those scoring higher in creativity are better hires.
Big data doesn’t necessarily equal useful information, yet Evolv’s test draws upon performance predictors from several call centres to target the type of data that indicates strong performance for call centre tasks. In addition to responses to the test, Evolv collects additional metadata on response time and number of times a respondent changes his or her answer. For call centre tasks, in which speed and accuracy are critical, this additional data may prove very useful in identifying potential employees.
Dr. David Ostberg reports that through their testing, Evolv has made interesting discoveries about predictors of call centre employment longevity. First, prior experience in a call centre is not an indicator that a worker will stay on. Also, people tend to hire those like themselves and this does not necessarily lead to hiring top performers or long-term employees. Ostberg reports that performance and employment longevity have significantly improved since implementing Evolv’s test.
It remains to be seen how beneficial this testing will be in the long-term. While there seem to be clear gains for individual companies, there is a tension between what is good for companies and good for the economy as a whole. Further, what is good for companies may not be good for its employees (and those not hired).
Twenty years ago, before most call centres were re-located out of the U.S., I worked at the call centre of a Santa Barbara-based catalog company, the Territory Ahead. My colleagues ranged from other college students to mothers and teachers looking for part-time or holiday work to those recently completing drug recovery programmes. I wonder how many of us would have passed Evolv’s online test?
The colleagues I keep in touch with have gone on to be attorneys, immigration advocates, fashion designers, and elderly care workers. Yet Evolv’s test selects those who will remain call centre workers for a longer period of time. This isn’t necessarily a negative development, but call centre work subsidized higher education for some of us and lower paying, but more rewarding jobs for others. Of course, the job itself did not encourage our efforts, we were not allowed to study during downtime, and schedules were not intentionally drawn to accommodate other jobs, though they did. Perhaps we weren’t the best call centre workers, either, organizing relay races around the cubicles during the night shifts, answering the phone with a different catalog name (Land’s End, J. Peterman, etc.), taking long breaks on the rooftop when there were fireworks shows or parades to be watched. But we were also able to fill in beyond our jobs. With less than an hour notice, one of us (a future teacher) ran a two-week training programme when the trainer was ill and another colleague (future designer) advised on graphic design for the catalog to help meet deadlines. So perhaps we contributed in ways not counted in Evolv’s study, or perhaps this type of adaptability isn’t as valuable to profitability as other qualities.
Jobs teach us so much more than what they mean to and it’s difficult to predict how the job contributes indirectly to society by the people it employs. So, I wonder what we lose when computers start to choose who to hire. I imagine these predictors are great for the bottom line and possibly may even improve customer’s experience with call centres. And, to be fair, the U.S. lost most of its call centre jobs a decade ago. But for the countries where the call centres are, are there would-be attorneys, advocates, teachers, carers not being hired? Thinking long-term, how else will these tests be used? Are we moving toward a time when interviews are unnecessary, where for example, college applications will be another test and not require human review? Or, are we moving toward better placement for employment, education, etc., that are informed by strong predictors of success and longevity and perhaps even well-being?