Most organisations don't fail at process mining. They fail at the second process. A pilot in accounts payable produces a clear win, then nothing repeats; the tool sits with one team, no one owns the next use case, and the initiative quietly becomes last year's project.

‍

This is a structural problem, not a technology one. A 2022 joint survey of 214 Celonis customers by Celonis and the Fraunhofer FIT Institute found that customers with a dedicated Centre of Excellence (CoE) were 8.8 times more likely to report positive ROI than those without one. The gap between a successful pilot and an enterprise capability is largely a question of how the effort is organised, not how good the tool is.

‍

What a CoE Actually Is

A process mining CoE isn't a bigger deployment of the same software; it's the operating model around it: a defined team, governance, a way of prioritising the next process, and a mechanism for turning findings into action. Fraunhofer's Center for Process Intelligence notes that scaling is where most process mining initiatives stall, since "many companies do not make it beyond the proof-of-concept phase." A pilot proves the technology works; a CoE makes the value repeatable.

‍

Why the Case Is Stronger Than It Looks

Beyond the headline ROI multiple, the same study found: 92% of CoE customers reported greater process transparency, 75% reported higher cost savings, 79% reported better use-case prioritisation, and 85% of organisations that had connected five or more processes said their CoE investment had paid off. At the time, over 200 Celonis customers with an established CoE represented more than $700 million in total realised value. The pattern is consistent: value comes from mining enough processes in a structured way, not from mining one process well.

‍

Three Phases of Building One

Fraunhofer describes CoE build-out in three stages:

• Analysis: establish current maturity, identify candidate processes, and define what success looks like with a concrete project plan.
• Conceptual design: define the operating model: governance, reporting lines, tool strategy, and a roadmap for new use cases. This is also where tooling gets critically re-evaluated for fit at scale.
• Implementation and evolution: the CoE goes live, with iterative integration, internal training, and ongoing support. This phase doesn't end; a CoE is a permanent capability, not a project.

‍

What Actually Drives Success

The research points to organisational factors over technical ones. An executive sponsor is the strongest predictor of success: 94% of customers with positive ROI had one, most often the head of technology (26%), a COO, or a CFO (19% each). Business buy-in follows early wins rather than being mandated upfront. Scale is a deliberate target: the 85% payoff rate is specifically tied to reaching five or more connected processes, not to running one process indefinitely. And governance plus community, not headcount alone, correlated most strongly with higher returns.

‍

Structuring the Team

A typical CoE includes a lead plus subject matter experts, data analysts, data engineers, and IT application owners, with change managers, product owners, or a business value architect added as programmes mature. Larger CoEs scale this considerably, often blending internal staff with external consultants. About two-thirds of surveyed organisations chose a centralised model over a hybrid one, with hybrid (a central hub plus function-specific satellite teams) typically emerging later as individual functions mature.

‍

Where This Connects to the Bigger Picture

A CoE is where organisations establish process mining as a discipline, but scope tends to expand, often into object-centric process mining once the CoE starts mining complex, multi-entity processes like order-to-cash. It's also worth being precise about boundaries: a CoE governs the diagnostic work of finding out how processes run. It doesn't execute or automate anything by itself. Mature organisations define explicitly how mining findings hand off to whatever fixes the process, whether that's automation, RPA, or a process orchestration platform, rather than assuming the CoE owns that too.

‍

What Leaders Should Do Next

• Secure a genuine executive sponsor before building anything else; it's the single strongest predictor of ROI.
• Plan for five-plus processes from the outset, even if the first deliverable is one pilot.
• Decide centralised vs. hybrid early, since retrofitting governance later is harder.
• Budget for change management and community-building, not just analyst headcount.
• Define the boundary between mining and execution now, before the organisation assumes the CoE owns both.

A process mining Centre of Excellence is an organisational commitment before it's a technical one. The enterprises getting outsized value aren't running the most sophisticated algorithms; they're the ones that took scaling as seriously as the pilot.

‍

─────────────────────────────────────────────────

‍

FAQs

‍

How many processes should a CoE aim to connect before expecting ROI? 

Survey data from Celonis and Fraunhofer FIT found that 85% of organisations that connected five or more processes reported their CoE investment had paid off, making five a reasonable initial benchmark rather than stopping after one successful pilot.

‍

Who should sponsor a process mining CoE? 

An executive sponsor, most commonly the head of technology, followed by the COO or CFO, was the strongest single predictor of CoE success in available research, present in 94% of organisations reporting positive ROI.

‍

Should a CoE be centralised or embedded in individual business functions? 

Most organisations start centralised (around two-thirds in the Celonis/Fraunhofer survey), then move to a hybrid model: a central hub plus function-specific satellite teams, sometimes emerging later as individual departments mature.

‍

Does a process mining CoE also handle automation and execution? 

Not typically, and you shouldn't assume it does by default. A CoE's core mandate is diagnostic: mining and understanding processes. Execution, whether through RPA, orchestration platforms, or broader automation, is usually a separate but connected capability, and mature organisations define that handoff explicitly.

‍

Is object-centric process mining relevant to a CoE just getting started? 

Not usually at the outset. OCPM becomes more relevant once a CoE is mining complex, multi-entity processes (like order-to-cash) rather than the simpler, linear processes most CoEs start with, such as procure-to-pay or accounts payable.