Process mining is software that rebuilds how work actually moves through a company, with its loops, delays and exceptions, from the time stamps that every order and invoice leaves in the company's systems. Companies use it to find where work gets stuck, before they decide what to change or automate.
Every system records what happens to each record, and when. The ERP notes the moment an invoice is received, the moment it is approved and the moment it is paid. These records are called the event log. Each line in it holds three things: a case ID, such as the invoice number; an activity, such as "invoice approved"; and a time stamp. Process mining reads millions of such lines. It joins the lines with the same case ID into the path that each case took. Then it draws all the paths together as one map, with the number of cases and the average time at each step.
Imagine a two-wheeler parts maker in Hosur. Its process manual says that a supplier invoice is approved within one day and paid within 30 days. The company loads one year of event logs, covering 48,000 invoices. The map shows something different. The median time to approval is six days, not one. 22% of invoices are sent back to the supplier at least once, because the price on the invoice does not match the purchase order. And 9% are paid late, so the company loses the discounts that suppliers offer for early payment. The map also shows the cause. Most of the price mismatches come from 15 suppliers whose prices changed during the year, while the purchase orders still carried the old prices.
This changes the decision. The company first corrects the prices of those 15 suppliers in the ERP. Only then does it automate the matching that remains, with a software robot or a workflow in case management software. After the change, process mining keeps measuring, so the company can see whether approval times really fall. These measures are often shown on business intelligence (BI) dashboards.
Process mining sees only what the systems record. Work done outside them, such as reading an email or checking a spreadsheet, leaves no event log. For that work, a related tool watches the screens instead. This is task mining. Some tools also use the history to predict what will happen next, such as which invoices will be paid late, with time-series forecasting.