Machine Learning Times
Machine Learning Times
AI Success Depends On How You Choose This One Number
 Originally published in Forbes, March 25, 2024. To do...
Elon Musk Predicts Artificial General Intelligence In 2 Years. Here’s Why That’s Hype
 Originally published in Forbes, April 10, 2024 When OpenAI’s...
Survey: Machine Learning Projects Still Routinely Fail to Deploy
 Originally published in KDnuggets. Eric Siegel highlights the chronic...
Three Best Practices for Unilever’s Global Analytics Initiatives
    This article from Morgan Vawter, Global Vice...

10 years ago
Using Predictive Analytics to Improve Hire Quality


To learn more about talent analytics, attend Predictive Analytics World for Workforce

As the economy improves, demand for new hires is rising as more departing and retiring employees leave openings to fill. With increased voluntary turnover, organizations face growing pressure to hire quickly, often resulting in an increase in poor hires. A good indicator of quality of hire is turnover during the first year of service. An employee leaving — voluntarily or involuntarily — within the first year typically indicates poor selection, poor onboarding or both.

Through the use of advanced analytic techniques, modeling frameworks and data from various sources, predictive services can be built to link employee behavior, activities, traits and performance to desired business outcomes. And hiring is one area where predictive analytics is already producing benefits.

According to PwC Saratoga’s “Human Capital Report,” first-year turnover has increased from its historic low of 21.5 percent in 2011 to 22.6 percent in 2012 to 24.1 percent in 2013. These are small increases, to be sure, but the two-year reversed trend may indicate breakdowns in hiring and onboarding.

Even small improvements in turnover can have a significant financial impact. PwC Saratoga (Editor’s note: The author is a director at the firm) calculates the average cost of turnover for new hires is equivalent to one to 1½ times the annual pay of the departing employee. The figure is based on direct costs and indirect costs. Costs directly attributable to hiring include recruiter costs, third-party advertising agency, formal onboarding and signing bonus. Indirect costs include lost productivity due to various factors, including the new hire’s learning ramp-up time and effort spent by hiring managers.

Consider a hypothetical example. At the 2013 median external hiring rate of 13.7 percent, a 10,000-employee organization would have hired 1,370 new employees in 2013, but would have lost 24.1 percent — or 330 — of them in the first year. Assuming an average annual salary of $50,000, this turnover would have cost at least $16.5 million. Even a 25 percent improvement — 82 fewer leave — would have been a savings of at least $4.1 million.

PwC’s benchmarking database, which covers more than 350 U.S. organizations, shows that an HR organization spends about $357 per employee on recruiting and staffing each year — or $3.57 million in a company with 10,000 employees. The “savings” in the above hypothetical example more than covers this figure.

By adopting predictive analytics, organizations can improve effectiveness of recruiting by understanding who is successful in its environment and which sources produce better hires. This approach allows an organization to build a robust data-driven analytic framework that enables it to rank effectiveness of different hiring sources by employee roles and use resulting heat maps and optimal hiring profiles to effectively channel future investment in recruiting.

First, prioritize relevant areas that could drive effectiveness of a hiring program. Consider identifying pivotal roles and digging into each independently rather than taking a holistic approach. Develop hypotheses for why some hires do better than others.

Second, collect relevant data about successful hires to test hypotheses. Merge disparate sources of data to evaluate performance of past hires who have stayed with the firm and consistently delivered on performance.

Third, use statistical modeling to understand why those hires were successful. What drives better quality of hires and which data points tell this story? Look beyond mere correlations to multivariate analysis of different factors within one modelling framework.

Finally, implement actions based on the results from predictive models into recruiting, hiring and onboarding. There are several ways to do this: develop hiring quality predictors by employee roles, create optimal hiring profiles, or build dynamic heat maps or dashboards that track effectiveness of hiring sources by business, region or employee roles.

The objective of predictive analytics applied to recruiting is to help organizations more effectively deploy resources and drive higher quality and greater longevity of new hires.

Successful organizations use predictive analytics in their candidate selection process by using advanced analytics to isolate characteristics of success, then using those markers to create optimal hiring profiles and tailoring assessment surveys that evaluate every job candidate for those characteristics — a combination of attitudinal, demographic and experiential factors.

It is important to note that even if the assessment instruments have common factors across roles, the optimal level for each factor should be tied to specific roles so the instrument produces a “go/no go” decision based on an evaluation of the data and the candidate’s background.

Consider another hypothetical example. A company that provides part- and full-time temporary workers to various retailers hires about 13,000 workers annually, but retained only 2,200 beyond 90 days. The company identified six practices on the retail floor for which it provided workers. It noted a possible difference between rural and urban environments, so the six practices were doubled to 12.

The company then developed 33 hypotheses to explain the high attrition rate. Next, it compiled data that allowed it to systematically test those hypotheses — data that included information like educational background, past jobs, job performance, as well as results from candidate experience and engagement surveys. Using this data, it ran different models to find out what is important for each of the dozen defined roles.

The company discovered that workers with a high degree of personal flexibility and greater comfort with the uncertainty of where and how long they would work were more likely to last beyond 90 days. They also learned that the commute distance to the job was a huge factor, especially for the urban workers.

From this analysis, the company created 12 characteristics profiles for the six roles for each setting, rural and urban. It then developed interview questions based on those profiles and is now working on a pre-assessment survey to identify the higher-quality job candidates.

Ranjan Dutta is a director in the PwC Human Resource Service Saratoga practice. He can be reached at editor [at] talentmgt [dot] com.

By: Ranjan Dutta, director, PwC Human Resource Service Saratoga practice
Originally published at

Leave a Reply