Which process mining tool shows exactly where the bottleneck is
Every serious tool on this page will show you exactly where orders wait, if you can give it an event log, one row per step per order with the order number, the step and a timestamp. It draws the real path of every order with the waiting time on each arrow, and the longest wait is your bottleneck, usually a hand-off such as credit release, stock allocation, release to the warehouse or invoicing.
So the answer depends on where your order data lives. On SAP, Celonis and SAP Signavio ship ready-made order-to-cash analyses. On Microsoft, Power Automate Process Mining is the cheapest credible start. With a CSV export, Disco draws the map in an afternoon. If orders arrive by email and live in spreadsheets, no tool can see the bottleneck yet, so read the event log section first.
I build and run production AI agents for operational teams, so my bias is towards removing the manual step once you have found it, not buying a bigger map. No vendor has paid to be here, and prices are what vendors publish, checked on 25 September 2026.
1. Celonis
Celonis is the best-known name in process mining and now calls itself a process intelligence platform. It extracts event data from ERP and CRM systems, rebuilds the path every order took, and shows where orders wait, loop back or go off route, with order management content aimed at on-time delivery.
It wins in large companies with high order volumes on SAP or Oracle, where a small gain in on-time delivery covers the licence. It is the wrong buy when orders arrive by email and live in spreadsheets, because there is nothing to mine. Celonis publishes no prices, so expect an enterprise sales process, though it still advertises a self-serve Free Plan.
2. SAP Signavio Process Intelligence
SAP Signavio Process Intelligence is SAP's own process mining, sold in a suite with process modelling. It connects to SAP and non-SAP systems, and SAP publishes ready-made order-to-cash accelerators for SAP ECC and S/4HANA, so SAP customers see order cycle times without building the analysis from scratch.
It wins for SAP customers, especially before a move to S/4HANA. It is the wrong buy if you are not on SAP or want a vendor independent of your ERP supplier. SAP's own price sheet on the UK government's G-Cloud 14 marketplace, dated April 2024 and still listed, prices Process Intelligence per block of 200,000 records at between 4,500 and 30,600 pounds per block per year, minimum three blocks, and the listing offers a 30-day free trial.
3. UiPath Process Mining
UiPath sells process mining and task mining together as Process Intelligence inside its automation cloud. It extracts data natively from SAP, Oracle, Salesforce, ServiceNow and Workday, shows where cases wait, and feeds what it finds into UiPath's automation tools.
It wins when mining feeds an existing UiPath automation programme, because finding the slow step and automating the manual work there happen in one platform. It is the wrong buy as a standalone mining tool. Process Intelligence is only in the Standard and Enterprise plans, priced by sales, with a 60-day free trial of Standard. If you need task mining on your own servers, check first, because UiPath removed it from its self-hosted Automation Suite in November 2025.
4. Microsoft Power Automate Process Mining
Power Automate Process Mining, formerly Process Advisor, is Microsoft's process and task mining inside the Power Platform, with process maps, variants, root cause analysis and a desktop app. Object-centric mining became generally available on 5 June 2026, and Dynamics 365 Supply Chain Management ships a warehouse material movement template for it.
It wins for companies already on Microsoft 365, as the cheapest credible way in. Power Automate Premium, at 15 dollars per user per month paid yearly, includes process and task mining with 50 MB of data storage. The Process Mining add-on costs 5,000 dollars per tenant per month paid yearly, adds 100 GB and requires Premium. There is a 30-day free trial. It is the wrong buy when your order data is large and sits outside Microsoft.
5. Fluxicon Disco
Disco is a desktop application for Windows and Mac from Fluxicon, a small company run by two process mining PhDs. Load a CSV or Excel export, point it at the order number, step and timestamp columns, and it draws the process map with the waiting time on every path. I would hand it to an analyst for a first look. Version 4.1 shipped in February 2025.
It wins for a one-off diagnosis, and for testing your data before you meet an enterprise vendor. It is the wrong buy for continuous monitoring, because it works from exports. The free demo imports only 100 events per file and cannot be used commercially, but Fluxicon offers free time-limited evaluation licences and free academic licences at partner universities. Commercial licences are priced on request.
6. IBM Process Mining
IBM Process Mining discovers processes from event logs and adds simulation, so you can test a change, such as a second person on credit checks, before making it. IBM says it was a leader in Gartner's 2025 process mining quadrant.
It wins for IBM customers and regulated firms that need an on-premises option. The SaaS edition on AWS starts at 4,250 dollars per month billed annually, with 20 GB of storage, three business users and one analyst. The on-premises edition on Red Hat OpenShift starts at 2,885 dollars per month for 20 million events, three business users and one analyst. A free trial is available. It is the wrong buy for a first diagnosis, since the entry SaaS edition costs 51,000 dollars a year before implementation.
7. ARIS Process Mining
ARIS is best known for process modelling, and ARIS Process Mining sits in the same platform as modelling, simulation and governance, so you can compare the process you designed with the one the data shows. ARIS has run as a standalone business since January 2025, inside the Software AG group that Silver Lake owns.
It wins for companies that already model processes in ARIS and for regulated organisations that need evidence of how processes really run. It is the wrong buy without a modelling practice to connect it to. Pricing is by quote, with a free Process Mining Basic trial.
8. Apromore
Apromore grew out of academic process mining research and is strong on simulation, root cause analysis and compliance checking. Salesforce completed its acquisition on 3 November 2025, and it is now sold as Apromore from Salesforce, feeding Salesforce's plans for agent-driven process automation.
It wins for Salesforce-centric companies whose order journey starts in the CRM. It is the wrong buy if you want independence from your CRM supplier. Its open-source core is archived on GitHub, with no release since 2021, so treat it as a commercial product priced by quote.
9. PM4Py and ProM
PM4Py is an open-source Python library maintained by Process Intelligence Solutions, a Fraunhofer spin-off, under the AGPL-3.0 licence, with a release on 1 September 2026. ProM is the long-standing academic workbench, a Java application extended by plug-ins, and version 6.15 shipped in September 2025. Both are free.
PM4Py wins when an analyst who writes Python wants the process map and the waiting time per stage without a licence fee. ProM wins for research and teaching. Both are the wrong buy for business users who need shared dashboards and support. Check the AGPL terms before building PM4Py into software you offer to others.
The event log problem, and what to do first
Every tool above needs an event log, one row per step per order with the order number, the step and a timestamp. Microsoft's own documentation uses the order ID as its example of a case. ERPs record these as a side effect of work, which is why process mining shines in large SAP and Oracle estates.
Most mid-size companies I work with run much of order fulfilment outside the ERP. Orders arrive as email attachments, someone keys them in, exceptions live in a shared spreadsheet, and the chase to the warehouse is another email. None of that leaves a timestamp, so a tool maps the part it can see, while the real wait often sits in the part it cannot.
Do three things first. Pull the timestamps your ERP already holds for last quarter's orders, typically created, released, shipped and invoiced, and compute the median wait between each pair. Run a two-hour workshop with the people who touch an order, walk five recent orders end to end, and note every step outside the system and where it waits. Then build a cycle-time dashboard on those timestamps, in Power BI, Metabase or a spreadsheet, showing the median and slowest tenth of orders per stage, refreshed weekly.
Give any step that matters a timestamp, even a dated status column or an inbox label. Buy a process mining tool when the dashboard shows something averages cannot explain, such as rework loops, many different paths or orders that behave differently by customer or site.
What changed in 2025 and 2026
In May 2026 Gartner replaced its Magic Quadrant for Process Mining Platforms with one for Process Intelligence Platforms, which also covers task mining, modelling and monitoring. The vendors' own announcements say it named ARIS, Celonis, Pega and SAP Signavio as leaders and QPR as a visionary.
Ownership moved too. Salesforce now owns Apromore, ARIS stands alone inside Software AG, Fortino Capital bought MEHRWERK, the maker of mpmX, in July 2025, and Celonis agreed in May 2026 to buy Ikigai Labs.
If you run SAP, follow Celonis v SAP. Celonis sued in a US federal court in March 2025, alleging that SAP uses its ERP position to shut rivals out of customer data. In June 2025 SAP agreed not to block the Celonis extractor or charge customers extra for it until the case ends, and in October 2025 the judge let most antitrust claims proceed. Ask any vendor how it will extract your SAP data and what that costs.
Object-centric process mining went mainstream. It follows an order that splits into several deliveries and invoices without forcing it into one case, which suits fulfilment. Microsoft made it generally available in June 2026, and QPR supports the OCEL 2.0 standard.
Two smaller European vendors work inside your data platform. QPR ProcessAnalyzer runs natively in Snowflake, with a 30-day free trial, then a fixed monthly fee plus capped usage. mpmX runs inside Qlik, Snowflake or Databricks, with a working capital trial of up to three months from 3,900 euros per month plus typically five to twelve consulting days of setup.
Comparison at a glance
| Tool | Best for | Typical cost | Wrong buy when |
|---|---|---|---|
| Celonis | Large SAP or Oracle estates, high order volumes | Custom quote, free plan to try | Orders live in email and spreadsheets |
| SAP Signavio Process Intelligence | SAP customers, especially before S/4HANA | Custom quote, G-Cloud sheet 4,500 to 30,600 pounds per 200,000-record block a year | You are not on SAP |
| UiPath Process Mining | Existing UiPath automation programmes | Standard or Enterprise plan by quote, 60-day trial | Bought as a standalone mining tool |
| Microsoft Power Automate Process Mining | Companies already on Microsoft 365 | In Premium at 15 dollars per user per month, add-on 5,000 dollars per tenant per month | Large order data outside Microsoft |
| Fluxicon Disco | One-off diagnosis from an export | Price on request, free demo up to 100 events per file | You need continuous monitoring |
| IBM Process Mining | IBM customers, on-premises needs | From 4,250 dollars per month SaaS, 2,885 dollars per month on-premises | A first diagnosis on a small budget |
| ARIS Process Mining | Firms already modelling in ARIS | Custom quote, free Basic trial | No modelling practice to connect to |
| Apromore | Salesforce-centric companies | Custom quote | You want independence from your CRM supplier |
| PM4Py and ProM | Analysts who code, research, teaching | Free, open source | Business users need dashboards and support |
How to pick
If your orders live in email and spreadsheets, or you have fewer than about 50 people, do not buy a process mining tool yet. Do the timestamps, workshop and dashboard first. If you want a map anyway, the Disco demo or a few lines of PM4Py will draw one from a sample.
Mid-size and on Microsoft, start with Power Automate Premium and add the tenant add-on only when your data outgrows 50 MB. For a one-off diagnosis, use Disco. With data already in Snowflake, Databricks or Qlik, look at QPR or mpmX.
In a large SAP estate, SAP Signavio is the natural commercial fit and Celonis the deepest independent option, so test both on your own order data. If you already run UiPath, IBM, ARIS or Salesforce, their process mining is usually the easiest to add.
At every size, the tool shows where orders wait. It does not remove the wait. That takes someone who owns the fix, which is often taking manual keying or chasing off the team at that step.