Celonis + VynFi: Load Synthetic Process Mining Data via IBC

Celonis is the market leader in process mining. Its Intelligent Business Cloud (IBC) ingests event logs from SAP, Oracle, Salesforce, and other systems to visualize process execution, detect inefficiencies, and drive automation. The typical onboarding takes weeks: connector setup, data extraction, transformation, and validation.

VynFi generates Celonis IBC-compatible event logs directly — with metadata sidecars, case/activity/timestamp triples, and realistic process variants including rework loops, three-way matching gaps, and anomaly injection. Use this for Celonis demos, PoCs, training environments, or algorithm development without touching production ERP data.

Generate IBC-Compatible Data

import os
import vynfi
client = vynfi.VynFi(api_key=os.environ["VYNFI_API_KEY"])
config = {
    "sector": "manufacturing",
    "rows": 10000,
    "companies": 10,
    "periods": 12,
    "processModels": ["p2p", "o2c", "s2c", "manufacturing"],
    "exportFormat": "json",
    "ocpm": {"enabled": True, "computeVariants": True},
    "fraudRate": 0.03,
    "fraudPacks": ["revenue_fraud", "vendor_kickback"],
}
job = client.jobs.generate_config(config=config)
completed = client.jobs.wait(job.id, timeout=600)
archive = client.jobs.download_archive(completed.id)
# The archive contains:
# - ocel-event-log.json (OCEL 2.0 native)
# - process_mining/celonis_ibc.csv (Celonis IBC format)
# - process_mining/celonis_metadata.json (sidecar)
print("Archive files:")
for f in archive.files():
    if "process_mining" in f or "ocel" in f:
        print(f"  {f}")

Import into Celonis

• In Celonis IBC, go to Data Integration > Data Pool > Add Data Connection.
• Choose 'CSV Upload' as the connection type.
• Upload the `celonis_ibc.csv` file from the archive.
• Map columns: CASE_ID, ACTIVITY, TIMESTAMP, RESOURCE (auto-detected from headers).
• Create a new Data Model and add the uploaded table as the activity table.
• Run the data load and open Process Explorer.

What You Get

The generated event log covers realistic P2P flows: Purchase Requisition → Purchase Order → Goods Receipt → Vendor Invoice → Payment, with realistic deviations — rework loops (GR rejected, re-inspected), three-way matching failures (price variance, quantity mismatch), and anomaly-injected fraud patterns (duplicate payments, fictitious vendors) at the configured 3% rate.

Celonis will surface these as process variants. The happy path typically represents 65-80% of cases; the remaining 20-35% show the deviations you'd investigate. Fraud-labeled cases appear with distinct activity patterns that Celonis's ML features can learn to detect.

Use Cases

• Celonis PoC demos: show the platform's value without waiting for ERP extraction.
• Training environments: let analysts practice process mining on realistic data without production access.
• Algorithm benchmarking: compare process mining algorithms on a dataset with known ground-truth variants and anomalies.
• Compliance testing: validate that your Celonis-based compliance checks detect the injected fraud patterns.
• Pre-sales: generate a client-specific sector (retail, manufacturing, financial services) for tailored demos.

Scaling Up

For large-scale Celonis deployments, use VynFi's Scale tier with NDJSON streaming to pipe event logs directly into Celonis's data pool via their Push API. This avoids file-based import entirely and supports TB-scale datasets with back-pressure control.