AIMultiple ResearchAIMultiple Research

Use Multi-level Process Mining to Tackle 3 Arduous Challenges [2024]

93% of business leaders showed a willingness to leverage process mining. Yet, 80% of these leaders are concerned about the efforts and time spent on selecting, extracting and transforming process data. Although AI-enabled process mining can automatically extract data from IT systems, it struggles to work with granular and multi- level data. 

Multi-level process mining (MLPM) is proposed to overcome these problems to increase precision and accuracy for process models and facilitate the adoption of process mining among different sectors. However, most vendors vaguely define MLPM capabilities of their solutions and buyers don’t know whether the solutions they are considering have MLPM capabilities.

Therefore, we’ll explain what MLPM is, how it can be identified, how it works and the top 3 challenges it can tackle. 

What is multi-level process mining?

Multi-level processes are complex processes containing many-to-many relationships, such as procure to pay (P2P) and order to cash (O2C).

Multi-level processes are challenging for traditional process mining software since they ignore the interaction between entities within processes This simplifies analysis results but leads to a loss of information relevant for analysis.

Multi-level process mining, also known as hierarchical process mining, employs an algorithm to tackle this challenge. The algorithm can define multiple different entities and track their interactions in the given process. As a result, multi-level process mining can provide a more detailed analysis taking into account relationships between entities.Therefore, users can identify activities in an entity that leads to deviations and bottlenecks in another entity.

How does multi-level process mining work?

Let’s demonstrate with an example from P2P:

  • Purchase requisition: A buyer sends 2 purchase requisitions (i.e. demand for purchase approval) for 4 items
  • Order: Procurement department creates a single order with one line for each item
  • Goods receipt: The items are received and registered in a single receipt with one line for each item in the warehouse
  • Invoice: The vendor shares the invoice, accounting department processes these received items and pays out the invoice.

An important thing to understand is the cost of this process and the cost depends on calculating the correct numbers of purchase requisitions, orders, goods receipt and invoices.

Traditional process mining tools generate a case for each requisition and calculate the cost of invoices for four distinct invoices. On the other hand, multi-level process mining can detect the association among these interrelated events, so it registers them as a single case, correctly calculating the cost.

How can you verify MLPM?

Use sample data from a multi-level process in the vendors’ tool during the PoC process. Look into especially complex cases. For example in case of P2P, analyze cases such as

  • Multiple requisitions merged into one order
  • A requisition that is split up into multiple orders
  • Multiple orders merged into a single invoice
  • Multiple orders that are split up into items and covered by multiple invoices
  • An order that is covered by multiple invoices

If you applied one of these cases and tracked the multiple orders created from one requisition on the model or simply zoom into a sub-process by opening a new tab, it is a sign that the tool deploys MLPM algorithm in process discovery. 

1. MLPM can overcome data granularity issues

One of the major issues for process mining is the high volume of data that event logs might contain. In fields like healthcare, education and user behavior studies, event logs include detailed information about the events, known as the granularity issue. 

The traditional process mining requires breaking these large event logs into smaller sections and then clustering them to analyze. However, analysts often complain that they lose a great deal of information which could have affected the process analysis. Thus, they often do not know if the levels they ignore will be relevant in the middle of the analysis. On the other hand, when they try to include all information, they end up with messy and chaotic process models. 

Multi-level process mining (MLPM) aims to overcome this issue since it can include multiple levels for an entity within the same model. As a result, the models become more user-friendly than the traditional approach. This way, analysts that want to understand user behavior by utilizing process mining can trace activities along the lines of sub-processes within an activity tree. 

2. MLPM can help measure the impact of changes

Companies often implement changes or automation for certain tasks and activities executed manually, aiming to improve overall process performance. Yet, such changes risk affecting other sub-processes that will also be affected by the change. Since sub-processes interact, it is unclear how increasing efficiency for one sub-process, like requisitions, could create a new bottleneck at the invoice level.   

Users can simulate and measure the impact of such changes at one level over another with the help of multi-level process mining. 

For example, in a case study from the UK, analysts applied multi-level process mining to the clinical pathway of endometrial cancer patients.1 Their analysis discovered the activities for each patient for 15 years to monitor the impact of the change in one year over the following years. The method allowed users to detect the change, localize, discover and compare it with the previous years. Scholars found 3 different periods where a critical change was observed. 

The image represents a process model discovered by Multi-level Process Mining. The model starts from referral and terminates with the diagnosis which ends with discharge or surgery.

Source: A multi-level approach for identifying process change in cancer pathways

The figure above is a directly-follows graph showing the process model of the pathway from the referral to the end. Each number on arrows shows the number of patients moving to the next activity. The researchers compared this process model against each model obtained in different years with conformance checks capability. 

3. MLPM can improve auditing 

Auditing is a major use case for process mining since process mining can reduce the time and manual efforts allocated for internal checks and comparisons. Process mining can map the process flow and compare it with the uploaded rules or ideal models. Without MLPM, multi-level processes can only be audited at a high level while MLPM enables precise auditing of all steps involved in the process.

What are the challenges of multi-level process mining?

MLPM can take longer to analyze processes than other process discovery algorithms since multi-level computation is more resource intensive. For especially complex processes, logs contain more traces and events for each trace, prolonging the analysis time. In such cases, a prior data cleaning phase is required to downsize the sample. 

Further reading

Explore more on process mining and automated process discovery:

If you decided to leverage process mining but do not know which vendor to choose, start reviewing our data-driven and comprehensive vendor list.

Check out comprehensive and constantly updated list of process mining case studies to understand multi-level process mining better with real-life examples.

And if you still have questions, you can always let us know:

Find the Right Vendors
Access Cem's 2 decades of B2B tech experience as a tech consultant, enterprise leader, startup entrepreneur & industry analyst. Leverage insights informing top Fortune 500 every month.
Cem Dilmegani
Principal Analyst
Follow on

Hazal Şimşek
Hazal is an industry analyst in AIMultiple. She is experienced in market research, quantitative research and data analytics. She received her master’s degree in Social Sciences from the University of Carlos III of Madrid and her bachelor’s degree in International Relations from Bilkent University.

Next to Read


Your email address will not be published. All fields are required.