Starting process mining without a data team
No data engineer, no warehouse, one ERP and two spreadsheets. That is the situation of most industrial SMEs, and it is not an obstacle to a first analysis.
Process mining carries a reputation for heavy projects, inherited from large rollouts where a warehouse gets built before anyone looks at anything. On an SME scope, the first useful analysis takes one query and one afternoon.
Pick a process, not the company
The first scope must fit in one sentence: from customer order to shipment, from purchase request to receipt, from quality report to closure. A process one person can describe from memory, and that someone already complains about. Trying to map the whole company is the surest way to deliver nothing.
The real prerequisites
- Read access to the ERP database, or a regular export if the vendor will not open it.
- Someone on the business side who can say what a status means. That is the only indispensable role.
- Twelve months of history, so that seasonality does not read as a drift.
What takes time, and what does not
Extraction and computation are not the issue: they are mechanical. What takes time is vocabulary: deciding that "on hold" and "suspended" are the same thing, or that they are not. That discussion cannot be delegated to a tool, and it is often where the first realisation happens.
Count on a week between first access and a first readable map. The rest (comparing against the written procedure, quantifying the gaps, automating a step) comes afterwards, and only if the first map showed something.
