Robotic process automation (RPA) provides a virtual workforce to automatize manual, repetitive and error-prone tasks. However, successful process automation requires knowledge about the potential for automation, effective training of the robots and continuous monitoring of their performance. Within this paper, we illustrate how process mining supports organizations throughout the lifecycle of RPA initiatives. Our process mining application provides a visual and fact-based proof for automation capabilities and enables prioritization of activities. The application further supports the user by monitoring and benchmarking robots to ensure sustainable benefits. These insights are especially valuable for process managers and miners with a special interest in process automation.
IT systems [
Within this paper, we present an approach to use the power of process mining to enable effective RPA activities as part of automation initiatives within process transformation. For a successful symbiosis, we suggest the following three-step approach for process automation via RPA:
Assessing RPA potential. The first decision within RPA refers to the discovery of
potential for process automation. Typically, within a company there are dozens of
process types and process steps with different levels of automation. For a successful RPA
implementation, processes should be scalable, repetitive and standardized [
Once processes are standardized, the highest potential for automation within an organization should be detected. Most systems have some form of automation already in place, so looking at current automation rates within processes is crucial to define RPA targets. Furthermore, there might be specific cases, by geography, vendor or material, where manual work is widespread. By comparing automation rates, users can explore, where current automation solutions might be improved and where additional automation could create benefits like reduction of throughput times or improvement of other process-related performance measures.
Developing RPA application. The next step covers the training of the RPA application. As a best practice, users should start to train robots with the existing workflow. The trained robots start their work within a pilot project and their activities are tracked by the underlying IT systems. After a sound number of executions, the generated process instances can be evaluated by using the process mining application. The performance of different robots and the non-robotic supported processes should be benchmarked to identify the most effective RPA implementation.
Safeguarding RPA benefits. After selection and implementation of the most effective RPA applications, continuous monitoring ensures tracking the impact of the RPA initiative and especially its return on investment. Process mining enables the user to see how processes change over time due to RPA. It also immediately detects when a process evolves and how robots need to adopt to an alternating business environment. 2
For the case study, we use a demo data set of the Purchase-to-Pay (P2P) process
obtained from a SAP ERP system. P2P is a common business process with many
standardized and highly repetitive transactions providing potential for automation. The demo
data covers 395,315 purchase orders created over one year. The application for RPA
evaluation is integrated in Celonis Process Mining [
Assessing RPA potential. As automation rates become transparent, process mining
allows the user to quickly and easily understand the maturity of business processes and
to identify the largest automation opportunities [
Comparing both firms, we can conclude that process maturity level is much higher for Aloha AG. From the example, which shows 99.3% of all P2P activities of Aloha AG, we see a quite linear process flow with a low number of deviations. In contrast, just 55.6% coverage for Balu AG generates a complex and non-standardized process picture. In this case, increasing the automation rate through RPA would cost much effort to train robots such that they handle all deviations and variants in the process. Hence, we can conclude from our process mining application, that the process maturity and consequently the readiness for RPA is higher for Aloha AG.
Moreover, process mining accelerates the implementation of RPA by streamlining processes prior to an RPA initiative and thereby decreasing time and cost of the robots’ training. For non-standard processes, process mining provides insights about the root causes driving complexity, so users can standardize the process first before moving on to RPA. For example, we can see from the OLAP table in the bottom right part of the analysis in Fig. 3, that the current automation rate for specific material groups, like ‘Metal processing’, is already at a good level compared to other materials like ‘Tools’. Starting with the low-hanging fruits, i.e. cases with low automation rates, often provides faster benefits from RPA than increasing the automation rate of already highly automatized process steps. In summary, the process mining application enables the user to detect the potential for automation, but also the business impact of automation by showing the change in throughput time and other process performance indicators when automation rates increase.
Developing RPA application. After selecting the target process and identifying as
well as prioritizing the best automation opportunities, various robots can be trained by
the RPA application, e.g. by recording the manual processes steps in the front-end
(here: the SAP ERP interface for P2P). As it is usually unclear in advance, which
process variant is most effective, different robots are trained by different process execution
paths. When training is finished, the process robots are deployed in a pilot project.
While common RPA applications record the performance of single robots, process
mining acts as a supervisor that benchmarks different robots, compares them with
nonautomatized executions and therefore allows to select the best performing
implementation. Via user interaction analysis as part of advanced process mining features [
Safeguarding RPA benefits. When robots are trained effectively, they can be deployed and actively used in daily business operations. While running automations, users should continuously maintain and increase automation capabilities. Our process mining application for RPA enables the user to continually monitor the processes from end-toend and to adjust the robots’ behavior when systems or other conditions change.
After successful implementation, users can start the three-step procedure again by identifying new processes for automation. 3
The RPA use case presented here is implemented in Celonis Process Mining and has
already been applied by several thousand users around the world to deploy their RPA
initiatives with the help of process mining. Also Gartner highlights in their recent
Market Guide for Process Mining that process mining is a key enabler when it comes to
RPA initiatives, since visualizing and understanding the process context as well as
spotting and prioritizing opportunities for task-level automation are huge success factors
[
A successful example for the combination of RPA and process mining can be found
at the global telecommunication giant Vodafone. Process mining alerted Vodafone that
many orders with excessive throughput times arise due to complex deviations from the
standard process before they are sent out to the supplier [
Based on the application of the RPA use case, we propose the following aspects to be key drivers of a successful symbiosis of RPA and process mining:
Selection of an appropriate use case. RPA unfolds its maximum potential on rulebased processes with high volume of manual tasks and handling time; processes with fixed procedures, that are standardized and mature; and processes where a technical integration via the backend is too costly and/or impossible. In these manual processes, it helps to eradicate human processing errors. Process mining can give guidance to RPA initiatives and points at the most eligible processes for automation. Common applications to achieve fast benefits from RPA are, among many others: transferring data from one system to another, payroll processing, customer registrations, customer profile updates, generation of standard reports, or data cleansing.
Standardization before automation. To be most successful, variation in business processes should be minimized before starting RPA. This means that process variants must be standardized to generate high volumes of transactions per variant. Through such a standardization, users can avoid expensive loops during RPA realization, speed up implementation time, increase success rates and achieve maximum return on investment.
Prioritization of activities. Usually, the number of potential use cases exceeds the resources available to do all projects at once. Thus, it is important to use available resources to achieve maximum results in a short period of time. Prioritizing is a key factor of successful RPA initiatives, as it allows to realize low-hanging fruits and to generate quick wins. Experience shows that starting RPA with processes exhibiting low automation rates can generate faster benefits than increasing automation of process that are already highly automatized. Process mining supports the structured prioritization of RPA projects.
Establishment of a central, coordinating unit in the organization. Moreover, establishing a central unit or task force within an organization helps to work out an overall RPA roadmap, to evaluate a high volume of ideas coming up, to prioritize initiatives in the best way, to accumulate knowledge and to define successful projects as benchmarks for future RPA activities.
Continuous monitoring of results. An RPA initiative is not a one-time project but requires to continuously track results and to use findings for ongoing improvement. This is where process mining provides fast and powerful insights into RPA’s impact on process performance KPIs such as throughput times. This aspect also includes to constantly benchmark projects with successful prior initiatives to gain maximum value from automation. 5
A free demo license of Celonis including the RPA use case can be accessed by
academic users after registration in the Celonis Academic Cloud [