The following article was written by professor Christian Terwiesch, chair of Wharton’s operations, information and decisions department and co-director of Penn’s Mack Institute for Innovation Management. Terwiesch is also professor of health policy at the Perelman School of Medicine.

As labor is becoming more costly and simultaneously emerges as a major bottleneck for many manufacturing and service industries, improving labor productivity becomes an obvious priority. Be it the preparation time it takes for a restaurant worker to cook a meal, the activity time of an assembly line worker in a car company to install a component, the call duration it takes a customer service representative to resolve a customer problem, or the service time it takes a healthcare worker to administer a vaccine — when labor is scarce, time is critical. In such cases, improving labor productivity is not just a matter of reducing costs. Higher labor productivity in capacity-constrained operations has a direct effect on service level, revenue, and growth.

Accurately measuring the productivity of workers has become something of a lost art. Benjamin Franklin famously said, “Lost time is never found again.” So how are we supposed to keep track of how much labor time it takes to perform a particular task? Here are three ways to measure it.

Using a Stopwatch. Though the approach might look archaic, manually measuring processing times of employees performing a repetitive task is still the most common approach to measuring labor productivity. When dealing with processing times that are in the range of seconds to minutes, an observer simply “peeks over the shoulder” of the employee to create a sample of observations. For example, many processing times in the assembly of standardized products such as cars or computers range between 30 seconds to 600 seconds. Using a stopwatch to measure labor productivity is easy and can be done quickly at low costs.

Output Tracking. Many modern workflow systems leave a trail of data behind as a unit goes through the process. Time stamps stored in the cash register of a retail check-out operation, connection times for customer requests in a contact center, bar code scans and GPS information for a delivery driver, or physical flows in a manufacturing facility that are tracked via RFID tags — all of these examples leave a trail of data behind. Such “digital exhaust” is a true treasure chest for those interested in analyzing worker productivity.[1] Output tracking happens automatically as a byproduct of the ongoing operations. Data collection, therefore, is cheap and does not have to be limited to a sample of observations. It can be performed for every worker and every single task they perform.

Visual Tracking. Recent advances in computer vision and artificial intelligence have shown the promise to combine the best elements of the methods above: non-obstructive data collection in real-time for all workers, in addition to spotting specific opportunities for improving the process. Cameras capture activity on the frontlines, be it the kitchen for a fast food restaurant or the assembly line of an automotive supplier. Deep learning methods help read the video data and identify objects, like a cup of coffee or an operator in a production plant. The software then uses a set of rules to determine processing times. This enables the generation of real-time reports to management. Real-time analysis can also be used to enforce adherence to pre-defined business rules and quality standards.

From Measuring Productivity to Driving Performance

Collecting accurate data about how labor spends its time is the foundation for any active measurement of labor productivity. High-performing operations take this data as a starting point to launch the following next steps.

Coach, Don’t Punish. Observing workers can easily be regarded as a “big brother” approach, encroaching on the individual privacy of the worker, and creating a culture of fear and anxiety. However, when such concerns arise, this is typically not a reflection on measuring productivity specifically, but is more generally indicative of a negative organizational culture and a lack of trust. So, first and foremost, the organization needs to create a psychologically safe work environment in which workers trust that whatever data about their productivity is collected will not lead to negative consequences.

Study How Your Workers Differ. We once conducted a labor productivity study of nurses in an emergency department, tracking the treatment times of a group of 189 care providers working in emergency care.[2] The difference between the top performing providers (the 10% with the highest productivity) and the bottom performing providers (the 10% with the lowest productivity) was enormous. After controlling for all external variables (such as patient conditions and time of day), we found that the top performers were able to see twice as many patients per shift than the bottom performers. We have seen similar levels of variation in other settings, including loan processing tasks, call durations, and assembly operations. Variation across workers is the norm, not the exception, and should always trigger further analysis.

Find the KPI’s and Focus on What Matters. As the advancement of technology makes measuring labor productivity easier, we face the risk of drowning in data generated by automated reports, leaving us confused about what actions to take. Just like how a medical doctor should start with careful diagnosis and only order the necessary tests for a patient, we should start implementing measurement systems where they matter the most. For this, we first need to go through the traditional steps of process analysis and improvement, including mapping out the workflows, identifying bottlenecks, and spotting inefficiencies. Such efforts generate the key performance indicators (KPIs) that spotlight the biggest levers for improving the performance of the system.

Build a Learning Organization. Monitoring labor productivity alone will not lead to improvement. Instead, we have to think about the process leading from data collection to process improvement. Any form of learning — be it in school, sports, or on the job — is built on the principle of providing feedback. We find it helpful to articulate three feedback loops:

  1. Workers should get feedback about their current performance. Such a direct and immediate feedback loop works best if fully automated and the feedback is provided in real-time, making output tracking and visual tracking the most useful technology to use. Automated feedback allows workers to correct errors themselves or seek help from their supervisor if needed.
  2. Share worker data with “Kaizen circles,” or groups of workers who are empowered to improve the process without explicit involvement of management. This equips workers with the data needed to come up with good ideas for improvement.
  3. Keep in mind that some problems are more structural in nature and might not be solvable at the worker or the team level. This includes problems of coordination among multiple processes or with suppliers and customers. More radical process changes such as the introduction of new technologies cannot be done at the worker or the team level, and instead, require high-quality data to be analyzed by management.

Thanks to the use of modern accounting systems, companies track and record every dollar, Euro, or Yuan that goes through their operation. It is hard to waste money without triggering some form of an alert.

The same level of rigor is often lacking when it comes to the measurement and analysis of labor productivity, even though one could argue that the time of our employees is one of the most valuable resources that exist in an organization.

A new generation of measurement tools holds great promise in improving labor productivity. These tools have the potential to significantly improve the skills of our workforce, reduce the current problems of labor shortage, and make sure that “labor time is never lost again.” 

References

[1] Terwiesch C, “OM forum—empirical research in operations management: From field studies to analyzing digital exhaust,” Manufacturing & Service Operations Management, 21 (4), 713–722, 2012.

[2] McCarthy ML, R Ding, JM Pines, C Terwiesch, M Sattarian, JA Hilton, “Provider variation in fast track treatment time,” Medical Care, 43–49, 2012.