Process mining measures. Magnet pulls the process.

“Isn’t this like Celonis?” — the question comes up in almost every conversation. The short answer: no, and the difference isn’t cosmetic. It decides how much of your real process you get to see at all.

What does process mining do?

Process mining reads a system’s event logs — usually an ERP’s — and reconstructs the actual workflow from them. For highly structured, transactional processes that run cleanly in one system, that’s powerful: cycle times, loops and deviations become measurable. If a purchasing process lives entirely in SAP, mining shows precisely where it stalls.

Where does mining stop?

Exactly where the event log stops. The real process rarely lives in a single system. Take a complaint: it arrives by email, is logged in the ticket system, coordinated internally on a Teams call, the goodwill approved by email, the credit booked in the ERP. Mining sees only the last step — the one in the ERP. The coordination, the exception, the decision in between appear in no log. What isn’t logged, mining can’t see.

What does Magnet do differently?

Magnet doesn’t measure one system; it pulls the process together from all relevant sources — including what happens between systems. Not an event log, but lived reality, condensed into a coherent view. And without a data lake you first have to feed everything into, and without anyone typing anything up.

When do you need which?

• Process mining, when you want to quantitatively optimize a single, ERP-centric process.
• Magnet, when you want to know how the company really works across systems and teams — as a basis for compliance, scaling or AI.

The two don’t exclude each other; they answer different questions. Mining tells you how efficiently a known workflow runs in one system. Magnet tells you what the workflow across all systems even looks like — often the prerequisite for asking mining the right question.

The term is expanded in the glossary entry on process mining; what the broad view enables is in The knowledge layer before the AI layer.