Guidelines for interpreting performance data
Interpreting performance data correctly will help you properly identify and fix the factors responsible for any turbulence.
Consider the following four practices for deciding how to respond to your performance data:
- Looking for trends can help you identify the bigger picture of how performance is changing over time, as reflected in the data you're gathering. If your actual performance is trending in the desired direction, you may not need to intervene.
- Considering the inherent variability in the process being measured will help you keep perspective. You don't want to overreact to a variation in performance measures that is due to normal fluctuations.
- Thinking about what's causing any variations in the data will help explain the causes beneath variations. Ask yourself what events or forces might underlie the variations you're seeing in your performance data.
For example, sales on a particular line of products may vary across seasons. Some companies set ranges of normal variation for such metrics and respond only when actual performance is outside that range.
For instance, suppose you see a major jump in error rates during month 3 of your data-tracking period. You'd want to investigate what was going on during that time that might have affected the error rate. Was a new piece of manufacturing equipment introduced at that time, and did the production staff have difficulty using it? Maybe the shop floor took on several new hires that month, and they hadn't yet mastered use of the equipment. In each of these cases, you might decide to hold off taking drastic action.
- Asking whether your targets or metrics need to be changed will help you determine if you need to reconsider your targets or metrics. Sometimes, when you see an abrupt change in your performance data, it's a signal that you need to reconsider your targets or metrics. Such signals can occur if your organization has changed an important process. An abrupt change in error rate may, for example, result from an employee scheduling change. Perhaps after starting your measurement program, you stopped the practice of allowing employees to vary their shifts, assigning each person a permanent shift. That could cause error rates to drop and then stabilize—suggesting that you can reasonably set lower targets for that metric.
Here's another example. Suppose that over the past 12 months, you've been tracking the number of weeks it takes to fill vacant positions in your group, with the goal of bringing new hires on board faster. At month 6, the number of weeks decreases sharply—and stays at roughly the same level during months 7 through 12.
When you investigate, you discover that your company's HR department overhauled its recruiting processes during month 6—by installing an online job-posting and job-application module in its human resource information system (HRIS). This new technology has vastly accelerated the hiring process—so that many new hires can now take days instead of weeks.
Suddenly, your metric "Number of weeks" has less meaning than it did before. If you still want to accelerate the hiring process, you might change your metric to "Number of days." Or, you may decide that the new technology has helped you improve the hiring process sufficiently—and you now want to measure some other aspect of your group's performance.
Clearly, interpreting performance data can be complex and challenging. Your company may have a unit devoted to analyzing all performance data. If it doesn't, and if you're struggling to make sense of a particular set of data, seek help from your boss or from an expert in your company who specializes in statistical analysis.