As discussed previously, KPIs need to be specific, measurable, attainable and relevant. Good KPIs are insightful, immediately intuitive and add value to business processes when measured over a longer time frame. KPIs that never change are of no real value because they don’t represent a metric in flux. The same goes for data that offers no ability to monitor it over the long term. As a KPI example, a binary data point indicating whether the network is up or down may be a useful piece of information, but it is not a KPI. Instead, “Daily network uptime” expressed as a percentage is far more useful, as it gives management the ability to monitor this information and spot trends as they are developing.
Good KPIs are either obvious in their relevancy or have some type of quality rating attached to them. “CPU utilization of 70%” is somewhat meaningless on its own, as is “CPU utilization: high.” “CPU utilization of 70%, high threshold status” ties these two pieces of information — a detailed performance metric and qualitative analysis — together. “CPU has been at 70% utilization, above high threshold, for the last 45 minutes” makes that KPI even more valuable.
One common problem with KPIs can occur when they are designed in a way that makes them easy to manipulate. On the surface, a KPI for “number of help desk requests” sounds promising, as it seemingly ensures your IT service department is not idle. However, these types of metrics can be susceptible to manipulation, especially when they are tied to compensation or job status. Examples of better KPIs would include the average time required for tickets to be resolved, average time before first response, or Net Promoter Score for service desk operations.