Process mining program

A process mining program is a standing capability that reads the event logs your systems already write and shows how work actually flows. It is not a single study. It runs on a schedule, against named processes, with an owner and a budget.

Mapping is testimony and mining is evidence. Process mapping asks people to describe the process and draws what they say, while mining reads the record the systems kept and surfaces the paths nobody mentioned.

The initials cause trouble. A process mining program (PMP) is not the Project Management Professional certification, a credential awarded to individuals; this is an operating capability owned by a team.

One constraint kills most attempts before they start. Mining needs case level data with timestamps — a case id, an activity name and a time for every step — and plenty of systems simply never wrote it.

Key takeaways

• Mining reads event logs, while mapping records what people say the process is.
• Every step needs a case id, an activity label and a timestamp, or there is nothing to mine.
• Three questions drive the work: what happens, does it match the rule, and where is the delay.
• The payback arrives on the second cycle, when a fix is re-measured against the same log.

How it works

A process mining program runs in cycles. You extract an event log, build the real model from it, compare that model with the process you believed you had, measure the delays, change something, then read the log again.

The reference site for the field, processmining.org, counts more than 50 commercial process mining vendors and frames the work as three questions.

A mined model still has to be drawn in a notation people can read. The Object Management Group describes its Business Process Model and Notation as the de facto standard for business process diagrams, in a flowchart form independent of any implementation.

Scope matters more than tooling — mining one department shows you a queue, while mining an end-to-end process from order to cash shows you the handover that created the queue.

Two measures turn a chart into a case for change. Process cycle time says how long the whole run takes, and process cycle efficiency says how much of that time was real work.

A program also needs a home. Most sit inside a business process management function, which already owns the process inventory, the process owners and the route to change.

Start with one process and one question. Programs that open with a tool selection stall, because the first real task is getting a clean log out of a system whose owner has no particular reason to help you.

Expect three rounds before anyone trusts the output. The first log is wrong, the second is incomplete, and the third is the one the process owner stops arguing with — which is when findings start turning into changes.

Governance decides whether anything changes at all. A finding needs a named process owner with the authority to alter a rule, or mining just produces a very detailed description of a problem that stays exactly where it was.

Privacy is a real constraint rather than a formality. Event logs identify whoever performed each step, so most programs mask the actor field before analysts see it and agree up front that findings are about processes, not people.

Examples

Finance shared services. A shared service centre in Manila mines its invoice handling log and finds a second approval step that the written policy never required, applied only to invoices above a threshold nobody could name.

Motor insurance claims. An insurer mines intake, assessment and settlement together, and learns that reopened claims follow a path the training manual does not describe at all.

Hospital patient flow. Admission, imaging and discharge each stamp a time. Mined together, they show where a patient waits for a decision rather than for a clinician, which is a scheduling fix rather than a staffing one.

Telecoms order fulfilment. Activation runs across sales, provisioning and field visits in separate systems. Stitching those logs together with a shared order id is most of the work, and the delay usually turns out to sit in a handover rather than in any one team.

Related terms

• Robotic process automation (RPA): what you automate once mining has shown which step is worth the effort.
• Process analysis: the wider study of how a process performs, of which mining is the data driven form.
• Process mapping: the drawn description of a process, where mining is the measured record of it.
• End-to-end process: the scope that makes a mining program worth funding.
• Process cycle efficiency: the ratio mining lets you calculate from real timestamps rather than estimates.

FAQ

What is a process mining program?

It is an ongoing capability that extracts event logs, builds models of how processes really run, and feeds the findings into a change route. The word program means it has an owner, a schedule and a budget.

How is process mining different from process mapping?

Mapping draws the process people describe in a workshop. Mining reconstructs the process from the timestamps systems recorded, so it finds the variants nobody mentioned.

What data does process mining need?

Three fields per event: a case id, an activity name and a timestamp. Anything else, such as cost or the person who acted, improves the analysis but is not essential.

Does PMP mean a process mining program or the project certification?

In operations writing it usually means a process mining program. Most search results for those initials point to the Project Management Professional credential, which is unrelated.

How long does a process mining program take to pay back?

That depends on log quality more than on the tool. Extraction dominates the first cycle, and the savings normally land on the second.

Shortlist providers with process mining and automation teams in the Outsource Accelerator directory before you scope the first process.