Process Mining: What It Is and How to Get Started

Process mining is one of the most exciting trends in process optimization. But what's behind it? Is it just another buzzword - or a real game-changer? In this article, we explain what process mining is, how it works, and whether it's relevant for your business.

What Is Process Mining?

Process mining is a technology that automatically analyzes how processes actually run in your organization - based on data from your IT systems.

The idea:

Your software systems (ERP, CRM, etc.) log every action. Process mining uses these logs to reconstruct and visualize the actual process flow.

Traditional process management:

• You ask employees: "How does the process work?"
• You get the theory (or what people think they do)
• The documented process ≠ the actual process

Process mining:

• You analyze what actually happened
• Based on hard data
• The real process becomes visible

How Does Process Mining Work?

The Data Foundation: Event Logs

Every IT system generates logs. An example from an ERP:

Timestamp
Case ID
Activity
User
2024-01-15 09:00:00
PO-1234
Order created
Miller
2024-01-15 09:15:00
PO-1234
Approval requested
Miller
2024-01-15 14:30:00
PO-1234
Approval granted
Schmidt
2024-01-16 10:00:00
PO-1234
Order sent
System
2024-01-18 08:00:00
PO-1234
Goods received
Weber
2024-01-18 09:30:00
PO-1234
Invoice booked
Fischer

The Algorithm Reconstructs the Process

From thousands of such entries, process mining identifies:

The standard process (how it usually runs)

Variants (deviations from the standard)

Anomalies (errors, loops, exceptions)

Cycle times (where does it take long?)

Bottlenecks (where does it get stuck?)

Visualization

The result is a process graph:

[Create Order]
        |
  [Approval needed?]
    /        \
  Yes         No
   |            |
[Request    [Send
 Approval]   Directly]
   |            |
[Approval       |
 Granted]       |
   |            |
[Send Order] <------
        |
[Goods Receipt]
        |
[Book Invoice]

Plus: Each arrow shows frequency and cycle time.

The Three Types of Process Mining

1. Discovery

Question: "How does the process really work?"

The algorithm reconstructs the process from the data - without prior knowledge.

Benefits:

• Make process visible
• Discover unknown variants
• Compare reality vs. documentation

2. Conformance Checking

Question: "Does the process follow the rules?"

The actual process is compared to the target process.

Benefits:

• Find compliance violations
• Identify process deviations
• Audit preparation

3. Enhancement

Question: "How can the process be improved?"

Use insights from mining to optimize.

Benefits:

• Eliminate bottlenecks
• Find automation potential
• Reduce process times

What Process Mining Reveals

Typical Insights

Variants:

"We have 47 different variants of the ordering process. The top 3 cover 80% - but what about the other 44?"

Loops:

"23% of invoices are reviewed 3 or more times before approval."

Wait times:

"Between approval request and approval, an average of 2.3 days passes. For Manager X, it's 5.7 days."

Exceptions:

"15% of orders skip approval entirely."

Automation potential:

"40% of all activities are pure status changes - automatable."

Example: Order-to-Cash Process

Expected process:
Order → Delivery → Invoice → Payment
(7 days)

Reality according to process mining:
35% follow the standard
25% have 2+ complaint loops
20% have invoice corrections
10% have dunning runs
10% have other deviations

Average cycle time: 23 days (not 7)

Process Mining Tools

Enterprise Solutions

Mid-Market Solutions

Open Source / Free

Is Process Mining Relevant for You?

Yes, if:

• You have ERP/CRM systems with transaction data
• You have >1,000 process instances per month
• Processes span multiple systems/departments
• You suspect inefficiencies but can't locate them
• Compliance/audit is a concern
• You're planning RPA/automation and want to know where

Probably not, if:

• You have few, simple processes
• Your systems don't generate usable logs
• Fewer than 500 process instances per year
• Processes are already well documented and controlled
• No budget for enterprise tools

How to Get Started with Process Mining

Phase 1: Proof of Concept (4-8 weeks)

Step 1: Select a process

Ideal candidates:

• High transaction volumes
• Suspected inefficiency
• Multiple participants/systems
• Good data availability

Step 2: Extract data

Typical sources:

• SAP: Tables BKPF, VBAK, EKKO, etc.
• Salesforce: Opportunities, Cases
• ServiceNow: Incidents, Changes
• Custom: Database logs

Step 3: Create event log

Required fields:

• Case ID (unique process instance)
• Activity (what happens)
• Timestamp (when)
• Optional: User, department, cost

Step 4: Perform analysis

Test with Apromore (free):

Upload event log (CSV/XES)

Generate process map

Analyze variants

Document insights

Phase 2: Pilot Project (2-3 months)

After successful PoC:

Make tool decision

Automate data connection

Set up dashboards

Train stakeholders

Implement first optimizations

Phase 3: Scaling

• Connect additional processes
• Set up continuous monitoring
• RPA/automation based on insights

Process Mining + Automation

The Connection

Process mining shows WHERE to automate.

Automation implements it.

[Process Mining]
      |
"40% of activities are manual data entry"
"Every 5th order requires manual rework"
"Approval process has 2 days wait time"
      |
[Automation Measures]
      |
API integration instead of copy-paste
Validation at entry, not at the end
Automatic approval below threshold
      |

[Process Mining Again]
      |
Measure improvement

Example

Process mining finds:

"80% of invoices under $1,000 are approved without changes. Average approval wait time: 1.5 days."

Automation solution:

Invoices < $1,000 from known vendors → Automatic approval

Process mining after automation:

"Average approval time for automated invoices: 2 minutes. Manual invoices: Still 1.5 days."

Cost-Benefit Analysis

Investment

Enterprise (Celonis/UiPath):

• License: $50,000-200,000/year
• Implementation: $50,000-150,000
• Maintenance: 10-20% of license

Mid-Market (Minit/ARIS):

• License: $10,000-50,000/year
• Implementation: $10,000-30,000

Entry Level (Apromore/PoC):

• License: $0-5,000
• Internal effort: 40-80 hours

Typical ROI

Experience values:

• 15-30% process time reduction
• 10-20% cost reduction
• 50%+ fewer compliance violations
• 3-6 months to break-even

Example calculation:

Process analyzed: Procurement
Volume: 10,000 orders/year
Cost per order: $50

Process mining investment: $30,000/year

Improvements found:
20% less rework: $15/order
10% faster cycle time: $5/order
Automation potential: $10/order

Savings: $30 x 10,000 = $300,000/year
ROI: 900%

Frequently Asked Questions

What data do I need?

Minimum: Process ID, activity, timestamp. Better: + User, cost, attributes.

How many data points do I need?

Recommended: 1,000+ process instances for meaningful analysis.

How long does a project take?

PoC: 4-8 weeks. Full project: 3-6 months.

Do I need developers?

Helpful for data access. The analysis itself is usually no-code.

GDPR concerns?

Yes, relevant. Anonymize/pseudonymize user data.

Conclusion

Process mining is more than a buzzword. It's the data-driven answer to "How do our processes really work?"

For large enterprises: Almost indispensable. The insights are too valuable. For mid-sized companies: Selectively useful. Yes for complex, high-volume processes. For small businesses: Usually overkill. Classic process mapping often suffices.

The best entry point: Free PoC with Apromore and a clearly defined process.

---

Want to know if process mining makes sense for you? We analyze your situation and show you where data-driven process analysis delivers the greatest value.