Follow Datanami:
May 14, 2019

How Machine Learning Prevents Employees from Churning

Travis Grubbs


According to Driver Knowledge, there is an average of 6 million car accidents in the U.S. every year (about 16,438 per day), 72% of these crashes result in property damage, and the driver is 23 times more likely to crash if texting. At the end of the fourth quarter of 2017, there were more than 229,000 motor vehicle repair and maintenance facilities in the United States, averaging about 14 dented, banged-up and crumpled-fender cars for every repair shop in the U.S. per day.

With all these accidents happening, you would think that repair and maintenance facilities couldn’t keep enough staff employed to handle the workload. However, for many repair shops, it’s not the complementary K-Cups from the customer waiting room that are disappearing, it is the staff that is becoming scarce from employee turnover that are putting the brakes on their bottom line. Operationally, turnover causes capacity challenges, reduces revenue, and shrinks the bottom line. Culturally, turnover erodes team membership, decreases employee engagement, and flattens overall morale.

One of the nation’s largest collision repair companies leads the industry by setting new standards in customer service, and by implementing cutting-edge technology. Unfortunately, their mechanics, painters and customer support staff were turning over as much as 40% a year across its hundreds of locations. The problem was, the specialized skill sets required in the collision repair industry are in high demand and some employees were easily tempted with offers from competitors.

With turnover negatively impacting their operational and cultural domains, the organization sought to better understand how to control the costly effects of regrettable turnover through an interactive analytics solution. To solve their employee turnover problem, they decided to add a new, cutting-edge technology gadget to their toolbox and turned to machine learning (ML).


High turnover rates are not the right direction for any company to be taking. Josh Bersin of Deloitte believes the cost of losing an employee can range from tens of thousands of dollars to 1.5–2.0x the employee’s annual salary. That includes the cost of hiring, on-boarding, training and ramp-up time to peak productivity. Like connecting an OBD2 professional automotive diagnostic tool to receive engine management lights, airbag/SRS and other car data, the organization was already familiar with how data analytics was changing the industry, so the concept of turning to an ML solution for employee diagnostics was not unknown territory.

The collision repair company reached out to Sparkhound, a leading digital solutions firm providing advanced analytics. Sparkhound experts knew that insight from an interactive analytics solution would reduce the cost of turnover but first, they needed to obtain a thorough understanding of the current state of turnovers, why it was happening, and determine the best method for building an action plan to reduce turnover risk.

The ML process pulled data in real-time from a data warehouse with functionally enforced row-level security. In addition, the analytics solution for their turnover challenges included a turnover risk indicator based on a custom-built regression model embedded in a PowerBI dashboard. The data was then processed into an interactive data visualization dashboard including:

  • A visual, user-friendly, roll-up-drill-down, actionable solution.
  • A turnover risk indicator and score, per employee, dashboard with persona-based views and row-level security.
  • A collection of statistical and time-based trends for predictions of future turnover risks.
  • Correlation-based charts and Actual vs. Target charts for turnover analysis.
  • Time and Geo-Location filtering.

Using the new ML solution, they now had a process for efficiently detecting and reducing an estimated $1 million annually in turnovers as they continued to scale their $2 billion business. The overall ML diagnostics were astounding:

  • The cumulative effect of managing and reducing turnover was estimated to be ~12-14% of HR operational costs annually.
  • There was a 7-9% impact on top-line results by decreasing turnover and avoiding lost sales, lost capacity, and/or inability to deliver quality products and services.
  • Difficult to quantify, but potentially more important, was the impact on culture… keeping and growing leaders and high performers have a huge ripple effect.

Increasing employee retention and decreasing operational, financial, and cultural costs associated with turnover could mean a 5-10% impact on top and bottom line results. The net result: ML solutions are not all overhyped. Finding the right solution to address turnover issues and improve the bottom line is a practical application of ML algorithms and improves employee morale by keeping all mechanics happy and engaged.

About the author: Travis Grubbs is the general manager of Sparkhound. He joined the company in 2013 and is responsible for managing operations and growth across the Dallas/Ft. Worth market. Travis started his career as a consultant more than 20 years ago in the Microsoft channel, focused on information management and business process optimization. Travis earned a bachelor’s degree in history and a Master of Business Administration (MBA) degree at Texas Christian University.

Related Items:

How to Build a Better Machine Learning Pipeline

It’s Still Early Days for Machine Learning Adoption