Delivery & Steering

Process Mining: Benefits, Getting Started and the Path from Analysis to Improvement

Marin Niehues26 March 202611 min read

Whiteboard with columns of sticky notes

Process mining reconstructs from the event data in your IT systems how processes actually run, and shows with hard figures where waiting time and rework arise. Before automation, it is the fastest route to a robust decision for high-volume processes with many variants. This article provides an overview and does not constitute legal advice.

Most process descriptions show the intended flow as described in workshops. The actual flow sits in the systems, because every case leaves timestamps in the ERP system, the CRM or the ticketing tool. Process mining analyses these traces and replaces recollection with measurement. That matters most when an investment in automation or AI is due.

What is process mining?

Process mining is an analytical method that automatically builds a model of the actual process from IT system event logs and compares that model with the intended process. The method originates in computer science and is closely associated with Wil van der Aalst, who developed it at Eindhoven University of Technology and later at RWTH Aachen University. An IEEE working group, the Task Force on Process Mining, published a manifesto in 2011 to which more than 75 experts from over 50 organizations contributed. It sets out guiding principles and describes maturity levels for the quality of event data.

The literature distinguishes three types of analysis:

  • Discovery: a process model is derived from the event data without any prior model.
  • Conformance checking: the actual flow is compared with the prescribed model or with rules, and deviations become visible case by case.
  • Enhancement: the model is extended with times, costs or resources, so that waiting times and bottlenecks become apparent.

The data requirement is simpler than many executives expect. An event log needs three attributes per entry, and further fields such as the handler or the amount refine the analysis.

What are the benefits of process mining?

Process mining gives the business a measured view of cycle time and rework for each process, and with it the basis for decisions on improvement and automation. The benefit shows in four areas:

  • Cycle time and on-time delivery: the analysis reveals where cases wait and which variants take longer.
  • Rework: repeated steps such as prices changed several times or invoices checked again are counted and can be valued.
  • Compliance: conformance checking finds cases that breach internal rules, for example purchase orders created only after the invoice arrived.
  • Automation potential: the number of variants and the share of manual steps show which part of a process suits automation.

Typical areas of use are the large cross-functional processes with high volumes: purchase to pay, order to cash, handling of customer requests and control of production orders. They run through a few core systems and therefore produce complete event data. Instead of competing departmental views, the analysis puts a shared finding on the table. How that finding connects with the search for the constraining step is covered in the article on finding bottlenecks in value creation.

Do you need process mining before automation?

A company needs process mining before automation once a process exceeds roughly a thousand cases a month, runs through one to three systems with timestamps and occurs in many variants. For smaller processes, or for work that happens mainly on paper and in email, a process walk-through with a sample of real cases is sufficient.

The reason lies in the logic of automation. A software robot or an AI agent works reliably on the standard path, and every exception requires its own rule or a handover to a person. Research on robotic process automation therefore describes structured, high-volume processes as its economic sweet spot. Process mining shows before the investment how large the standard path is and how many cases deviate from it. If a process with many unexplained variants is automated, the exceptions move into a manual residual workload that consumes the planned benefit.

For AI applications there is a second reason. Process mining shows where people decide, check or gather information in a process, which is precisely where language models and agents come in. The analysis also supplies the baseline for the business case: cycle time and rework rate before launch. Which processes suit AI is covered in the article on process automation with AI.

How do you get started?

You get started with a single process, a clear question and an outcome metric expressed in euros or days. The approach follows five steps:

  1. Define the question: for example, why does purchase order approval take twelve days on average, and how many cases run in loops?
  2. Clarify the case notion and data source: is a case the order or the order line, and which tables do the events come from? Agreeing this with IT and the business prevents most later misreadings.
  3. Build and check the event log: data is extracted, condensed into events and reconciled with the business on a sample. In delivery practice, this step accounts for the largest share of the effort.
  4. Analyze: discovery shows the variants, time analysis the waiting times, conformance checking the rule breaches. The result is a short list of causes with their share of cycle time.
  5. Derive and measure actions: every action gets an accountable owner and a target value. The same analysis is repeated after implementation and proves the effect.

A benchmark from delivery practice applies to duration: the first analysis of a process from one core system is available after six to ten weeks. Several source systems or missing unique case IDs extend the third step in particular.

From analysis to improvement

The value of process mining arises only when the finding turns into a changed way of working. Research on organizational use, including a Delphi study with experts from academia and practice and an interview study on adoption, identifies the same hurdle: insights are gained but not translated into actions and firm accountability. The tools are mature. What is missing is organizational anchoring.

Three rules for implementation follow. First, every analyzed process has a process owner with decision rights over its interfaces. Second, actions are ranked by their impact on the outcome metric, and the sequence follows the bottleneck. Third, the metric is measured continuously and reviewed every quarter. How process ownership and organizational structure fit together is covered in the article on process management and organization design.

Data protection and co-determination

Event data frequently contains handler IDs and timestamps per person. It is therefore personal data, and the analysis falls under the GDPR. In German companies with a works council, it also touches co-determination, because such a tool is technically capable of monitoring the behavior or performance of employees. Three safeguards have proven effective: the purpose is set down in writing as processes and cycle times, the handler field is pseudonymised or aggregated to team level, and the works council is involved before the first extraction. How to plan this early is shown in the article on the works council and AI adoption.

How much does process mining cost?

Process mining costs the tool license, data preparation and the time of the business and IT, and in the first analysis data preparation is usually the largest single item. Licenses are mostly sold as subscriptions priced by data volume, processes or users. External support in the German consulting market is typically billed at around 1,000 to 2,500 euros per consultant day.

A simple calculation shows whether the analysis pays off. A company posts 120,000 supplier invoices a year. The analysis shows that 8 percent of them pass through a clarification loop costing 25 minutes on average. That makes 9,600 loops and 4,000 hours. At 60 euros per hour, that is 240,000 euros a year. If an action addressing the root cause halves the number of loops, the benefit is 120,000 euros annually, plus shorter cycle times and secured early-payment discounts. How applications are sequenced by value and feasibility is shown in the article on prioritizing AI use cases.

Niehues Consulting as your partner for process mining

Niehues Consulting uses process mining where it prepares a decision: before automation, before an AI application or before responsibilities are reorganised. The work covers business analysis and requirements, validation of the data with the business, the business case and the sequencing of actions. Marin Niehues is a management consultant with a delivery perspective: he builds organizations that deliver and brings AI into operations. His projects range from VC-backed scale-ups to DAX 40 groups, in 20+ countries on six continents.

The approach follows the Niehues Delivery Model with its phases Understand, Build and Embed. Three of its review questions matter most for process mining. Which business outcome should change measurably? What is the baseline, and how do we measure success? Which metric shows after 90 days that the solution is landing? How such initiatives fit into a managed portfolio is described in the article AI management: what executive leadership must steer. The starting point is a free 30-minute initial consultation via the contact page.

Frequently asked questions

What is process mining in simple terms?

Process mining reads from the timestamps in IT systems the order in which the steps of a case actually happened. The result is a map of the real process with all its detours and waiting times.

What data does process mining need?

Process mining needs a case ID, an activity name and a timestamp for each event. ERP, CRM and ticketing systems almost always hold this information, spread across several tables.

How long does a process mining project take?

As a benchmark, the first analysis of a process from one core system takes six to ten weeks. Preparing and checking the data takes up most of that time, and follow-up analyses are much faster because the data pipeline is in place.

What is the difference between process mining and task mining?

Process mining analyses the flow of a case across systems and departments using event logs. Task mining records work at the individual desktop, such as clicks and inputs, and shows how a single step is carried out. It affects employees more directly and calls for even more careful alignment on data protection.

What is object-centric process mining?

Object-centric process mining handles events that affect several objects at once, such as one delivery covering three orders. Classic process mining assigns each event to exactly one case and distorts such processes. The OCEL 2.0 specification from 2023 describes an exchange format for this.

Which processes are best suited to process mining?

High-volume processes running through a few core systems are best suited, such as purchase to pay or order to cash. Work that happens mainly in email and on paper does not yield complete event data, so a process walk-through with a sample is the better starting point there.

Is process mining GDPR compliant?

Process mining can be run in line with the GDPR when the purpose is defined, the data is minimized and handler IDs are pseudonymised. In German companies with a works council, co-determination must also be observed before the first extraction.

Let’s talk about your initiative.

In 30 minutes we clarify where your organization stands and which step will pay off most now.

Request a consultation

30 minutes · free · no obligation