Scope & boundaries
This guide covers discovering how a process actually runs from the event logs your systems already write, before anyone decides what to automate.
It does not cover automating it once you know (Robotic Process Automation (RPA) Software).
Executive Summary
Choose a process mining platform by where your process data lives, whether the work leaves a system log, and what your team will do with a finding.
Process mining software rebuilds the path a business process actually takes from the event data that ERP, CRM and other systems record as work happens. The IEEE Task Force on Process Mining states that the starting point for any process mining effort is a collection of events commonly referred to as an event log. The software turns that log into a map of every path cases took, how long each step waited, and where cases left the intended process.
No process mining platform is the right choice in every situation. Three questions decide the fit: where the process data lives, whether part of the work happens on desktops that leave no system log, and what your team will do with a finding, whether that is fixing a configuration, changing a policy or automating a step. Section 4 matches each situation to the approach built for it. The vendor landscape names eight vendors across seven approaches.
Most organizations also buy process mining services, because extracting and preparing event data is one of the largest parts of the work. Section 3 separates the work to buy as software from the work to buy as help. Automating the tasks a process mining project finds is covered in the robotic process automation guide.
What Business Process Discovery Tools Do
Business process discovery tools build a process model from data instead of interviews. Process discovery techniques take an event log or event stream as input and produce a process model without using any a-priori information. The model shows every variant in the data, including the ones nobody drew in a workshop.
Discovery is one of several analyses on the same data. The IEEE Task Force on Process Mining states that process mining includes automated process discovery (extracting process models from an event log), conformance checking (monitoring deviations by comparing model and log), organizational mining, automated construction of simulation models, case prediction and history-based recommendations.
Buyers use two of them most after discovery. Conformance checking can be used to check if reality, as reflected by the event data, conforms to a predefined process model, and vice versa. Process mining can also be used to extend process models with performance-related information such as bottlenecks, waste and costs. Conformance checking shows where cases skip an approval or repeat a step. Performance analysis shows where cases wait, and for how long.
Event data formats decide how portable the work is. The IEEE Task Force on Process Mining states that the goal of the eXtensible Event Stream (XES) Standard is to standardize a language to transport, store and exchange event data, for example for process mining. The IEEE Task Force on Process Mining initiated a community process to co-design Object-Centric Event Data (OCED), the successor to XES.
Object-centric process mining follows several linked objects at once, such as an order, its deliveries and its invoices, instead of forcing every event into a single case. The OCEL 2.0 site from RWTH Aachen University states that object-centric event logs form the basis for object-centric process mining, and that OCEL 2.0, unlike the first standard, can depict changes in objects and provide more information on object relationships. Ask each vendor whether it imports and exports XES or OCEL 2.0 files, and whether one analysis can follow more than one object type.
Desktop work is the gap in system data. Steps done in spreadsheets, email and screens that write nothing to a system of record leave no event log to discover. Task mining records that work on the desktop, and section 4 covers the vendors that sell it.
Process Mining Services: What to Buy as Software and What to Buy as Help
The IEEE Task Force on Process Mining states that process mining and related services have become a sizeable business for software vendors, professional service firms and commercial end users alike. Most of that service work is data preparation. The IEEE Task Force on Process Mining states that process mining data typically originates from systems that were not designed for it and requires substantial transformation, and that this bottleneck is one of the predominant effort drivers in process mining projects.
Process mining services fall into three groups. Data services extract records from source systems and build the event log. Analysis services run the first discovery and conformance work on chosen processes and size the improvements. Enablement services train an internal team to run the platform without outside help. Vendors, their partners and independent consulting firms sell all three.
Vendors package these services in different ways. ARIS lists a Consultant Mentoring Service that embeds expert consultants in the customer’s team to fast-track deliverables, master process mining and build internal capabilities. mindzie states that its enterprise pricing can include implementation support, training, onboarding and guidance from its experts. Apromore states that it supports customer engagement through a network of international service and technology partners.
A data science team can build process mining on open-source libraries. The connectors, the event-log pipelines and an interface that business analysts can use take most of the effort, and those are what the platforms sell. The usual decision is how much of the data work your own team does and how much a partner does.
| Service | What it delivers | Ask before you sign |
|---|---|---|
| Data extraction and event-log design | Connectors to source systems, case and activity definitions, and a refreshed event log | Who maintains the extraction when a source system is upgraded |
| Proof of value | Discovery and conformance analysis on one or two chosen processes | Whether it runs on your data or on sample data, and who owns the findings |
| Improvement program | Analysts who size and track the changes the analysis finds | How the value is measured, and by whom |
| Training and enablement | Courses, mentoring and certification for your own team | When your team can run an analysis without the partner |
| Managed service | A provider runs the platform and the analyses for you | Who can see your process data, and where it is stored |
Which Process Intelligence Platforms Fit Which Situation?
Each situation below matches one approach. The vendor facts come from each vendor’s own pages, and the table after them lists what to test in a proof of value.
Most of the process runs in SAP. The ERP vendor’s own tool reaches that data with the least extraction work. SAP Signavio states that customers can connect their SAP system and start receiving insights in less than 24 hours. SAP Signavio describes a process mining solution that goes deeper across SAP and non-SAP systems alike. SAP Signavio describes mapping every business object across the landscape to visualize how all processes interact.
The process spans many systems. Cross-system platforms are built around their own data layer. Celonis states that its platform is made up of three layers, the first being the Data Core, which extracts data from any source and queries billions of records at speed. Celonis states that the Celonis Context Model, the heart of the Celonis Platform, creates a dynamic, system-agnostic, real-time representation of the entire business. Automation vendors also extract from several systems. UiPath states that native extraction is available for SAP, Oracle, Salesforce, ServiceNow and Workday, with a connector framework for custom applications.
A finding should become an automation. UiPath states that findings convert directly into automated workflows on the same platform, with no export-import cycle. UiPath states that UiPath Process Intelligence is a full discovery and analysis platform on its own, and that integration with agentic orchestration is available when wanted, not a prerequisite. IBM states that process insights can trigger corrective actions when specific business logic is met, such as alerting users on a KPI or activating payments for overdue invoices. IBM lists conformance checking, simulation and what-if analysis, an object-centric data model, action triggering and alerting, and automated RPA bot generation. The robots that run automated steps are covered in the robotic process automation guide.
Your apps and flows already run on a low-code automation platform. Microsoft states that the process mining capability in Power Automate extracts event data from systems of record and visualizes the processes, allowing users to customize reports, compare processes, determine the root cause of inefficiencies and monitor KPIs. Microsoft lists Power Automate Premium at $15.00 per user per month, paid yearly, including process mining for individual users. Microsoft states that each per-user license comes with 50 MB of process mining, up to a maximum of 100 GB per tenant. Larger analyses need the tenant add-on described under pricing.
Process models and governance are part of the job. ARIS describes its platform as combining process mining, process modeling, simulation, workflow orchestration and governance in a single unified platform. ARIS describes comparing actual execution with intended process design and quantifying the impact of deviations.
Much of the work happens on desktops. UiPath states that UiPath Process Intelligence combines UiPath Process Mining, which analyzes event logs from enterprise systems, and UiPath Task Mining, which captures the desktop work that system logs cannot see. UiPath states that its task mining includes built-in personally identifiable information (PII) redaction. Skan AI states that its first version launched in 2020, and describes the category that emerged, dynamic process intelligence, as built on computer vision and observation rather than interviews, workshops and assumptions. Skan AI states that it observes work, not people, covering in-scope apps only with all data encrypted and aggregated. Desktop capture records what employees do on screen, so agree its scope with employee representatives and the privacy and security teams before a pilot.
The team wants methods it can trace to published research. Apromore states that its software is the result of over ten years of research and innovation at the University of Melbourne, the University of Tartu and several other universities.
| Situation | Approach to shortlist | Test in the proof of value |
|---|---|---|
| Most of the process runs in SAP | Process mining in the ERP vendor’s suite | Time from connection to a first analysis, and coverage of the steps outside SAP |
| The process spans many systems | Cross-system process intelligence platform | A connector for every source in scope, and how cases are joined across them |
| Findings should become automations | Process mining in an automation suite | One finding turned into a running workflow during the trial |
| Apps and flows already run on a low-code platform | Process mining added to a low-code automation platform | Data limits per license and per tenant |
| Process models and governance are part of the job | Process modeling and mining suite | Mined behavior compared with your documented process models |
| Much of the work happens on desktops | Desktop observation and task mining | Which applications are recorded, what is redacted, and who sees the results |
| Methods must trace to published research | Research-built process mining | Which published methods the product implements |
Vendors in Process Mining & Process Intelligence Platforms
Representative vendors include Apromore, Celonis, IBM Process Mining, Microsoft Power Automate Process Mining, SAP Signavio, Skan AI, Software AG ARIS (now ARIS) and UiPath (Process Intelligence). The list is alphabetical, is not a ranking, and is not for sale. How we choose vendors
The approaches below follow what each vendor built before it sold process mining. That origin still decides which systems connect easily and what happens after a finding.
Several of these products came from acquisitions. SAP announced on March 5, 2021 that it had completed the acquisition of Signavio. The release states that Signavio's products become part of SAP's Business Process Intelligence portfolio. Microsoft announced the acquisition of Minit on Mar. 31, 2022. IBM announced a definitive agreement to acquire myInvenio, a process mining software company based in Reggio Emilia, Italy, on April 15, 2021.
Software AG announced on January 7, 2025 that Adabas & Natural and ARIS launch as standalone businesses. Software AG states that Silver Lake appointed executives to lead Software GmbH, which continues to be the holding company for ARIS, Adabas & Natural and Software AG’s central functions. ARIS states that it became an independent company for the first time in 2025.
Celonis states that it was founded in 2011, bringing process mining from academia into the boardroom. Celonis states that the platform provides capabilities to analyze, design and operate AI-driven processes. Celonis lists zero-copy, bi-directional integrations with Microsoft among its announcements.
Apromore states that it began in 2009 and became part of Salesforce in November 2025.
Other vendors start from the desktop. mindzie states that its task mining captures desktop-level activity to reveal manual work, repeated steps and productivity gaps that system data alone cannot explain.
| Vendor | Approach | Where it fits |
|---|---|---|
| Celonis | Cross-system process intelligence platform | Founded 2011; Data Core extraction, the Celonis Context Model, and tools to analyze, design and operate processes |
| Skan AI | Desktop observation and task mining | Computer vision and observation of in-scope apps; data encrypted and aggregated |
| Microsoft Power Automate Process Mining | Process mining added to a low-code automation platform | In Power Automate Premium at $15.00 per user per month with 50 MB each; tenant add-on $5,000.00 per month |
| IBM Process Mining | Process mining in an automation suite | From the myInvenio acquisition; SaaS from USD 4,250 a month, on premises from USD 2,885 a month |
| UiPath (Process Intelligence) | Process mining in an automation suite | Process Mining and Task Mining together; native extraction for SAP, Oracle, Salesforce, ServiceNow and Workday |
| SAP Signavio | Process mining in the ERP vendor’s suite | Part of SAP since March 2021; insights within 24 hours of connecting an SAP system; non-SAP systems too |
| Software AG ARIS (now ARIS) | Process modeling and mining suite | Standalone since 2025; process mining, modeling, simulation, orchestration and governance on one platform |
| Apromore | Research-built process mining | Research from the University of Melbourne and the University of Tartu; part of Salesforce since November 2025 |
Eight vendors are named here across seven approaches. The approaches were written before any vendor was selected, and a row in this table cannot be bought. Process mining vendors, named here or not, can describe their own products in the Spotlight below.
How much should you budget for Process Mining & Process Intelligence Platforms?
Process mining is priced on some mix of users, data volume and deployment. US dollar list prices for two of the vendors here are on their own pricing pages. Microsoft lists Power Automate Premium at $15.00 per user per month, paid yearly, including process mining for individual users. Microsoft states that each per-user license comes with 50 MB of process mining, up to a maximum of 100 GB per tenant. Microsoft lists the Power Automate Process Mining add-on at $5,000.00 per tenant per month, paid yearly, available as an add-on for the Power Automate Premium plan only.
IBM lists the SaaS platform from USD 4,250 a month, sold on an annual basis and hosted on Amazon Web Services, starting with 20 GB storage, 3 business users and 1 analyst user. IBM lists the on-premises platform from USD 2,885 a month on Red Hat OpenShift, starting with 3 projects, 20 million events, 3 business users, 1 analyst user and 2 task mining agents.
A free edition suits a first analysis on one laptop. mindzie offers a free Desktop Edition for local process mining that supports up to 1,000 cases.
For the other vendors, ask for a written quote on the same basis: users by role, data volume or events per year, number of source systems, deployment model, and any services in the price. Ask for the price at renewal, and for each process or source system added later.
| Meter | You are charged for | Grows with | Where it goes wrong |
|---|---|---|---|
| Users by role | Analysts who build analyses and business users who view them | Rollout beyond the first team | Viewers priced like analysts |
| Data volume or events | Storage, or the number of events processed | Processes, history and refresh frequency | A second process doubling the event count |
| Tenant or platform fee | A flat fee for the capability | Renewal terms | Paying the fee before anyone uses it |
| Connectors and source systems | Extractors for each system | Each new system in scope | Custom systems quoted as services |
| Services | Extraction, analysis and training | Number of processes | Extraction maintenance left out of the quote |
How long does implementation take for Process Mining & Process Intelligence Platforms?
Start with one process, and add the next when the first finding has led to a change with an owner.
Pick one process with a named owner, a measurable problem and records in systems you can reach, such as order-to-cash or purchase-to-pay. Write down the question the analysis must answer.
Extract case IDs, activities and timestamps from each source system. Check the log with the people who run the process before any analysis is shared.
Run discovery and conformance checking, confirm the findings with process owners, and turn at least one finding into a change with an owner and a date.
Refresh the data on a schedule, track the measure the change was meant to move, and add the next process only when the first has an owner acting on it.
What should you ask vendors about Process Mining & Process Intelligence Platforms?
Ask each vendor to answer in writing and to demonstrate on an extract of your own data.
-
Does most of the process run in one ERP?Yes Include that ERP vendor’s process mining tool, and compare it on the steps that run outside the ERP.No Start with cross-system platforms, and test a connector for every source in scope.
-
Does part of the work happen on desktops with no system log?Yes Add task mining or desktop observation, and agree its scope with the privacy and security teams first.No Process mining on system logs covers the first project.
-
Should findings become automations on the same platform?Yes Shortlist automation suites with process mining, and turn one finding into a workflow during the trial.No Give more weight to analysis depth, process modeling and data connectivity.