Skip to main content

Celonis Product Documentation

KPI driver analysis

KPI driver analysis overview

The KPI driver analysis is an automated analysis of the drivers of poor performance in your process. It outputs these drivers ranked by the impact of fixing them (actionability) on your process. This helps you focus on where you can make the biggest improvements to your process.

For example, the analysis might flag that many cases (invoices) in your process with a specific attribute (vendor) and an attribute value (Celonis) are linked to poor performance. Digging into the invoice data and processes for that vendor and fixing any issues would impact your process; the higher the actionability ranking for that driver, the greater the overall process improvement.

Performing a KPI driver analysis

  1. Select any attributes that could be driving low performance.

    The analysis runs automatically and outputs a table including drivers, their actionability, the number of cases affected, and the potential impact of fixing issues with this driver.

    Note

    If you don’t select an attribute, the analysis outputs data for all attributes which you can then filter on.

  2. Review the drivers

  3. Drill down into the driver data and look at driver combinations.

  4. Refine your analysis by selecting attributes that interest you.

  5. Share your findings with your team by creating an Insight.

Explaining actionability

An actionable driver is an insight that has a significant impact if addressed and is addressable". An antonym is a "gravity problem." As in, "fixing gravity would make flying easier, but it's impossible." Therefore, the insight that gravity is a root cause that prevents us from flying is not actionable. It would have very low actionability.

Celonis uses a statistical algorithm to determine whether a driver is actionable by balancing the number of cases (impact) that have a specific attribute value against the difference in the distribution if the case meets or doesn’t meet a specific KPI target.

For example, if an attribute value accounts for 80% of cases, it has a large improvement potential. However, if 90% of these cases are on track to meet a KPI target, the algorithm will give the driver a lower actionability rank, as fixing the remaining cases will most likely be difficult.