Unlocking Business Insights in Timecards

When it comes to big data analytics in the world of human capital management, most of the attention has focused on the hiring process. But companies are now starting to use advanced analytics to inspect the performance of the workforce after it’s been set. The source of the data? The humble timecard.

It turns out there’s a wealth of potentially valuable data contained in timecards. The good stuff is not the obvious line item that shows Jane Smith worked 37 hours last week. Basic reporting tools can tabulate those numbers for workforces that number in the hundreds of thousands of employees, and slice them any way you like–by store, by region, by shift, or by season.

A richer source of data lies in the changes and edits that have been made to those timecards by the store managers. For a mid-sized company with 15,000 employees, there could be upwards of 400,000 changes in the audit trail over the course of the year, according to Gregg Gordon, senior director of the new Big Data Workforce Practice Group at Kronos, which develops labor tracking, scheduling, and recruiting software for some of the biggest companies in the country.

That audit trail is typically accessed only if there’s an employee dispute over hours worked or if the Department of Labor wants to audit the company, Gordon says. It really served no other purpose. But after running the audit trail data through a K-means clustering analysis, data scientists at Kronos realized that they could mine that data much like clickstream data for a website, and that it could reveal some interesting patterns that weren’t observable any other way.

The K-means clustering analysis revealed outliers in the audit trail, or which store managers are making a lot of edits and changes to the timecards. Further analysis could reveal whether the changes helped the store’s performance or hurt it. What’s more, by adding external data sources, such receipts from a point of sale (POS) system or work orders from an ERP system, Kronos found they could correlate those timecard changes to draw conclusions about their managers.

“Store managers edit timecards based on a wide variety of situations,” Gordon says. “The clock is the system of record but people forget to punch in or they get stuck in a line or a manager waived them over and asks tells them to service a customer right away. From store manager’s perspective, all of them are required. But not all improve the performance of the business.”

The audit trail analysis of one Kronos customer revealed a store manager who was constantly reducing labor against the recommended labor schedule provided by corporate headquarters. “At first thought, you’d say ‘He’s just trying to reduce labor to improve his performance, but he might be missing revenue because he doesn’t have enough labor to service the customer and he’s missing opportunity in the store,'” Gordon says.

“But when we mapped it against his output, his sales results, we found that the schedule was overstaffing,” he continues. “He was underscheduling against what his recommended schedule was, but his sales results were higher than what corporate expected. So he was actually doing a real good job and what it told corporate was it needed to go reevaluate the forecast of his store.”

While this big data technique can highlight the work of good store managers, it can also help to uncover store managers who do bad things, such as cheating employees out of hours or gaming the labor system.

When the K-Means clustering approach was applied to another Kronos customer in the retail business, they discovered a store manager who was making a lot of edits to time cards. “He was shaving points off his employees’ actual time punches in order to meet California meal and break laws and also to ensure that they didn’t extend working hours over eight hours,” Gordon says (employees in California receive overtime after eight hours, compared to 40 hours per week in other states). “It makes his performance look better, which is what his bonus is based on.”

Manufactures may spend less than retailers on labor, but it’s still a big chunk of the budget. One of Kronos’ early adopters applied the big data analytics technique to the timecard audit trail and got some unexpected results. A deep inspection of the timecards revealed an employee attendance problem, despite the fact the manufacture’s variance reports looked fine.

“What happened is supervisors who are tired of getting their hand slapped by bad variance reports have gamed the system and they’re slowly adjusting their labor standards and their staffing,” Gordon says. “They’re putting buffer into the workforce. So their reports look fine and they’re managing all that variability themes and they don’t have to report anything. But you’re losing utilization and productivity on the floor.”

Kronos, which launched Gordon’s big data group last month, has delivered several add-ons that allow its customers to analyze their timecard audit trails to discover these types of issues. The add-ons were written in a combination of R and other languages, and plug into the Kronos Analytics product, which is an OEMed version of Microstrategy‘s business intelligence software.

The company also offers hiring analytics in its recruiting and human resources tool, which it acquired several years ago. For now, there’s no concrete plans to deliver predictive analytics in the workforce management products and complementary BI tools, but that could be in the works, Gordon says.

Related Items:

IBM Targets Predictive Analytics at Hiring, Retention

Hiring with Hadoop: How Big Data is Helping HR

In Big Data, Jobs Search For You