Through my time spent on a data-science team, I’ve helped uncover interesting stories for customers through the data that was already at their fingertips. In this article, I’m going to focus on examples that come to mind that I believe many agencies could relate to. The purpose of these examples is to help inspire new ways of looking at workforce data to confidently make decisions around complicated problems related to the workforce.
Using Labor Data to Optimize Schedules
The transportation maintenance department in a mid-sized county was having issues. They had recently switched to an experimental 4/10 workweek, which means they worked 10 hours a day for just 4 days a week. It appeared to be going well until supervisors noticed that there was a large gap in coverage for maintenance requests on Fridays and Saturdays. When the 4/10 schedule was introduced, employees were supposed to stagger their workweeks. It wasn’t fully evident that this wasn’t happening until analyzed and the graph of worked hours spiked on Mo/Tu/We/Th. There were 80% less hours-worked on Fridays compared to Tuesdays or Wednesdays. This fairly-simple analysis prompted the department to switch back to a 5x8 workweek, and the problem was instantly resolved. Afterwards, there was less than a 10% variance in worked-hours for any given weekday and constituents were able to benefit from even, consistent coverage. The trial wasn’t a failure, however. The county also benefited from knowing a 4/10 would work if they were willing to stagger the days that employees were able to take off.Analyzing Unplanned Absences
Unplanned absences are always top-of-mind for most employers. In 2018, the U.S. Department of Labor (DOL) estimated that over 3.2 percent of a government’s workforce was absent on any given day2. That’s higher than almost every private-sector industry listed. This can cause ripple-effects for entire teams as well as gaps in coverage and unplanned overtime. A mid-sized county that I worked with had a department struggling with both these issues (coverage gaps and OT). If unplanned sick days are truly unplanned, one would expect the rate to be level month-to-month, with few exceptions.When the whole county was analyzed, the unplanned absence rate was a near-constant 5%, which is in line with many comparable counties. If they took that at face-value, they wouldn’t see an issue. However, we visualized this with basic analytics to dive deeper and view unplanned absence rates by specific departments by month. What they found was alarming: One department was ~5% every month, except for July, October, and December. Rates didn’t just double in these months, they saw them tripling and quadrupling!
This visibility gave supervisors the data that they needed to take immediate action and drill down even deeper to figure out which employees were driving these high rates. It turned out that management wasn’t correctly enforcing their attendance policy, and refresher courses were provided which nearly eliminated this issue going forward. Issues like this aren’t hard to resolve, but the hard part is finding where they exist. Easily-accessible data can be a valuable tool for organizations trying to assess their sick and attendance policies.
Reducing Overtime by Hiring
A mid-sized county noticed an unusually large spike in payroll without an increase in employee count. This wasn’t budgeted for so further examination was needed. The payroll department could tell that overtime was a higher proportion of payroll dollars than usual but didn’t have a quick way to visualize what was driving it. After running their data through an analytics tool, what they found was very surprising to them. Almost twenty-five percent of their total overtime wages was being paid out to 1% of their employees. By isolating those employees, they were able to see that they were all working above 1,200 OT hours. One was even over 2,000 hours! Further investigation of these employees showed that they were all working understaffed positions that required specialized skills. Analytics helped the employer justify opening requisitions for many of these positions that ultimately reduced overtime and provided more jobs for the county.Creating an Analytics Team
In 2016, Gartner estimated that 60% of data projects fail. A year later, Gartner analyst Nick Heudecker said the estimate was "too conservative" and believed that it was realistically closer to 85%. That may seem worrisome, but with the proper planning, software and the correct stakeholders, any agency can utilize their data to solve problems.Tod Newcomb, a technology contributor for Governing.com, states that “It’s not a technology project that should be run by the IT department, though it will need input from CIOs and their staff to manage the databases and networks that underpin it. It’s also not about data. Rather, it’s a way to predict future strategies and support decision-making. That’s why the right stakeholders need to be involved.”3
To find insights like the stories above, there needs to be multiple stakeholders involved. Without domain experts’ input on an analytics project, analysts won’t know what they’re looking to solve. Building the right combination of domain knowledge, expertise, and technical knowledge is crucial for an analytics team to succeed.
1 https://media.erepublic.com/document/GOV17_HANDBOOK_Optum_V.pdf
2 https://www.bls.gov/cps/cpsaat47.htm
3 https://www.governing.com/columns/tech-talk/gov-data-analytics-failure.html
