Sample work · Data & Documents
Process analysis: where orders get stuck between ERP, warehouse and accounting
Three exports with different date formats and time zones were merged into an event log. This yields the process graph, waiting times per step and the biggest bottlenecks.
Sample work (demo), not a client job. Built as a trial task with invented example data; the check figures below come from the automatic test run and are not polished.
What was built
- Exports merged with local time, UTC and US date, including across daylight saving time changes
- Warehouse and accounting rows assigned to the order despite differing spellings
- Processors pseudonymized prior to evaluation
Download kennzahlen.xlsx Download prozessanalyse-demo.zip
The data is fictitious and has a known target. Bottleneck drivers are anomalies, not proven causes; measures in the report are suggestions for clarification.
How it was checked
Automatic test run on 2026-10-31, taken from the log (test data invented, figures measured):
- 1000 orders from 3 exports (ERP-CSV local time DD.MM., warehouse-XLSX UTC with 5 reference spellings, accounting-CSV US date with free text/invoice number linkage, daylight saving time change within the period) -> 6039 events
- Linkage warehouse 1970/1976, accounting 1877/1883 = target
Limits
What this sample does not cover, so nobody expects more:
- Data quality at the customer's determines the result: missing scans, subsequently added timestamps or collective postings distort waiting times; the sample detects missing events and orphaned rows and ...
- We do not build access to the customer's systems ourselves: the basis consists of exports (CSV/XLSX) generated by the customer; field meanings (which column is 'approval', which time zone does the warehouse use) ...
- Data protection: Processors are pseudonymized before analysis using HMAC-SHA256 (MA-xxxxxx); the key remains with the customer, only they can reverse it. Personal performance evaluation ...
- Graphviz (dot) is not installed: pm4py discovers DFG and performance DFG, the image prozessgraph.png is a custom matplotlib drawing (fixed layers, edge labeling count/median), not ...
- There is only one timestamp per step: 'waiting time' is the interval to the previous event (idle and processing time combined), calendar hours without working time calendar
- Bottleneck drivers are anomalies in segments (one or two characteristics), not proven causes; measures in the report are suggestions for clarification with the customer
- Free text without year ('Order 123') is assigned to the current year; local time specifications in the repeated hour of the daylight saving time change (October 25, 02:00-03:00) are ambiguous and would be discarded and ...
- pm4py is licensed under AGPL v3: unproblematic as an internal analysis tool, code delivered or offered as a network service would have to be disclosed or requires the commercial license
- Synthetic data with a known target; real processes have more variants, partial deliveries and consolidated invoices (n:m linkage), which require custom rules per customer
Something like this for your business?
Describe briefly what you need. You get a fixed price before I start; nothing happens without your yes.