Process mining is not software. It is a method with four phases. We have held the sequence strictly — it dovetails with the SCOReX®-oriented process map we maintain on ERP-adjacent mandates.
Phase 1 — Data intake and event-log extraction. The hardest phase in the Mittelstand, contrary to vendor decks. We need three fields per activity: case ID, activity name, timestamp. In the ERP these sit across tables such as VBAK/VBAP/LIPS (SAP) or OINV/ORDR (Business One). Fraunhofer FIT describes this phase as the most critical lever.
Phase 2 — Discovery. The event log renders the as-is as a graph. Alpha miner and inductive miner yield different views; Fraunhofer IAO recommends vendor-neutral analysis. From our experience the first visualisation is rarely pretty — that is the point.
Phase 3 — Conformance check. The to-be is compared against the data-reconstructed as-is. Compliance risk surfaces, four-eyes principles are bypassed, reversal bookings distort statistics. We have found at a Dreher Consulting client in mechanical engineering with 280 employees that 18 percent of all customer orders had technically skipped approval-relevant steps — documented, but noticed by nobody.
Phase 4 — Enhancement. The model is enriched (cycle times, bottlenecks); optimisation hypotheses are prioritised on data. Process mining delivers facts; process management delivers change.
Celonis and SAP Signavio Process Intelligence bundle these four phases — convenient, but obscures the methodical core. Tools are excellent, the sequence remains decisive.
A practical example: process-mining pilot at a Mittelstand mechanical-engineering company
Anonymised case from mechanical engineering: 320 employees, three plants in southern Germany and Austria, SAP S/4HANA Public Cloud, target 20 percent cycle-time reduction in order fulfilment. Achieved 23 percent in nine months — slightly above target.
We have observed in our projects with Mittelstand companies — in over 30 projects — that once more than 250 employees fall inside scope without a process owner, the initiative tips into a pure IT exercise. Concrete example: a specialised equipment manufacturer with 320 employees, three plants in southern Germany and Austria, SAP S/4HANA Public Cloud live since 2023. Pilot target: 20 percent cycle-time reduction in order fulfilment.
Phase 1 took six weeks for clean event-log extraction — three parallel order-number ranges (standard, custom equipment, service) lacked a shared case ID. We built a mapping table with IT. In phase 2 discovery surfaced 41 process variants where three had been assumed. In phase 3 we found 11 percent of custom-equipment orders had gone into production without technical clarification. Cycle-time penalty: 17 days on average.
In phase 4 we anchored four KPIs and installed a process owner from sales. After nine months cycle time was down 23 percent — slightly above target. From our experience that speed is only reachable when three preconditions hold: process owner named first, case ID unique, and executives accept that phase 1 takes longer than phases 2 to 4 combined.
Common mistakes in process-mining roll-outs
Five recurring patterns sort Mittelstand initiatives into success or failure: tool selection before process ownership, missing end-to-end case ID, discovery without conformance, enterprise scaling without Mittelstand adjustment, and mining detached from process management.
We have observed five patterns in our projects that predict Mittelstand failure — independent of the tool.
Mistake 1 — tool before process owner. The procurement decision for Celonis or SAP Signavio is signed before role clarification. Three months later: expensive licence, no one accountable. Process owner first, tool second.
Mistake 2 — missing end-to-end case ID. Transactions rarely carry a single ID across systems. Order-to-cash starts in the CRM, runs through the ERP, ends in accounting. Without a mapping concept, discovery produces three separate sub-processes.
Mistake 3 — discovery without conformance. Many initiatives stop after phase 2. Pretty pictures shown, everybody nods, nothing happens. Only conformance checks quantify deviation from target.
Mistake 4 — enterprise scaling without Mittelstand adjustment. The Computerwoche analysis shows productive deployments concentrate in companies above 1,000 employees. Mittelstand players need leaner pilot scopes — we begin with a single end-to-end process.
Mistake 5 — process mining detached from process management. Insights evaporate without target processes, KPIs and reviews. Process mining does not replace process management; it supplies its evidence base. The IEEE Task Force on Process Mining anchors this distinction in the Process Mining Manifesto.
What process-mining vendors won't tell mid-market buyers
Three gaps between the vendor story and DACH Mittelstand reality: the 80-percent data-readiness gap in event-log preparation, the fact that discovery is the easy half and change is the hard one, and the sober question of when a classical BPMN workshop is more efficient than a mining project.
Vendor decks look identical worldwide; the Mittelstand reality in the DACH region does not. Three gaps recur in our projects — also visible in the Computerwoche market analysis on the mid-market lag.
1. The 80-percent data gap
Vendors present cleanly structured event logs from enterprise environments. In the Mittelstand, from our experience, at least 80 percent of the time one of three things is missing: consistent activity naming, end-to-end case IDs, or reliable timestamps across system boundaries. What was sold as a six-week pilot becomes 14 to 18 weeks in practice, two thirds of it data preparation.
2. Discovery is the easy half, change is the other
Process mining produces excellent diagnoses. Turning a diagnosis into change hangs on the process owner, KPI anchoring and incremental optimisation rather than radical re-engineering. We have seen initiatives where the tool produced dashboards for twelve months that no one acted on. Process discipline falls asleep without KPIs.
3. When classical BPMN mapping is enough
Not every process needs mining. If a process has fewer than 20 variants, sits with a single owner and runs in one system, a four-hour lean workshop with a BPMN sketch beats a four-week mining project. Our rule: more than three systems, more than 20 variants, at least 250 cases per month — then mining pays off.
Our take
Process mining in the Mittelstand is not a tool problem. It is a methodology problem decided by the three points above, in that sequence.
Application across Mittelstand verticals — when process mining is the right tool
Process mining delivers measurable leverage in mechanical engineering, specialised equipment, kitchen-manufacturer configuration, logistics complaint handling and finance use cases like purchase-to-pay. Thin-traffic processes are faster and cheaper handled with a four-hour lean workshop and a BPMN sketch.
We have deployed process mining across industries — fit is not uniform. Read-out from 30+ engagements.
Mechanical engineering and specialised equipment. Highly relevant: many variants, long cycle times, multiple systems in order-to-cash. Leverage: 15 to 25 percent cycle-time reduction.
Kitchen manufacturer and configuration-heavy furniture industry. At a Dreher Consulting kitchen manufacturer client we analysed mass customisation: 800 configurations per month, seven systems from configurator to dispatch. Discovery revealed two notionally parallel processes running intertwined. Effort: two quarters; effect sustained.
Logistics and trade. Complaint handling: many cases, high variance, several systems. Conformance checks regularly uncover four-eyes breaches.
Accounting and finance. Purchase-to-pay and period close are often the first productive mining use cases because data is clean inside the ERP. The Bitkom 2025 digitalisation study lists finance as the function with the highest BPM maturity.
Service and language-academy environments. Limited fit: the method assumes structured event data. Unstructured touchpoints benefit more from classical process management.
Frequently asked questions
Next steps
Process mining works when methodology comes before tooling. We examine three points with you: is there a process owner? An end-to-end case ID? Is the goal discovery, conformance or enhancement? When two answers are “not yet”, that is where we start — not at tool selection. More on our methodology on independent ERP consulting and in our overview of digitalisation services.
Initial conversation on your process-mining pilot
30 minutes with our senior consultants — we examine process-owner role, case-ID mapping and the goal of your first end-to-end process.
