Product Q&A

Has anyone developed a framework for creating a Universal Health Scorecard for business processes across an organisation? I started exploring this idea after reading Keung's 2000 research paper. My goal is to develop a common framework for assessing process health, regardless of the type of business process. One challenge I face is that my company has both sequential and parallel processes. While sequential processes are relatively straightforward to analyse using ARIS CCPM, parallel processes are more challenging because a single case is broken down into multiple tasks, each with its own case ID. These tasks can still be linked through their original case ID. I believe the fundamental perspectives for assessing process health should remain consistent across processes, even if the indicators differ based on their structure and the capabilities of the process mining tool. My current perspectives, drawn from a structured literature review, are: - Time: Turnaround time (TAT)- Cost: FTE re

I hate just how limited Process Mining feels beyond the Process Explorer map. How can PM consultants be expected to identify opportunities to improve a process when the people operating it may have years of experience and a much deeper understanding of its practical realities? What unique information does Process Mining provide that materially changes the decisions those people make? I have been assigned to the Process Mining Team and have been struggling ever since to find its usefulness beyond the process map. Many of the other analyses and dashboards I build seem reproducible using regular reporting and analytics tools, which makes me question what the incremental value of Process Mining actually is. I have also struggled to find case studies that clearly explain the complete chain from a Process Mining analysis to a specific business intervention and then to a quantified business outcome. Many case studies describe the problem and the eventual improvement, but I rarely see exactly

In the previous article I described why we need two hierarchies instead of one hybrid one, so let me dive deeper into what I mean by that. What is a process hierarchy?First and foremost, let’s define this and one of the most important topics here is the naming. Some people call it a process hierarchy, others call it a process framework and even some people call it a process structure, however it all comes down to the same very thing: a process hierarchy is a structured approach to collect and sort the various domains, sub domains and single business processes in order to be able to describe the way of working of your organisation with a minimum (preferably no) redundancy.As this is quite the statement, let’s unpack it.First it deals with a structured approach and this is also the most complicated topic. The challenge that organisations have to deal with is that all of the content that you document using a BPM tool has two purposes:(1) It needs to be created / maintain(2) It needs to be

Hi everyone,During data extraction, I would like to convert a string field into a datetime format.The original field format is: 2024-12-09I would like to convert it to the following format: 2024-12-09 00:00:00.Does anyone have any idea how I could do this?I would really appreciate your help. Thank you! 😊   

I’ve got a KPI requiring to check whether a user is active in a certain timeframe based on a column in a table. In the event a user is inactive, it will count against the KPI. For example, if a user is active in January and February but not in March, the KPI would calculate to a 67% usage rate. I’ve figured out I can use the TIMELINE_TABLE function to map out when users are inactive.The issue I’m running into is filtering on the timeline table. I wind up getting an error message and I’m not sure how to troubleshoot it after reading through the documentation.When I select a specific user or time period, this error populates. I was using the formula template [ “TIMELINE_TABLE”().’Column Name’ ] for dimensions, but once I select a dimension or the timestamp column, I receive the above error. Any help would be fantastic!

Hi everyone,I’m trying to connect MongoDB to Celonis. At first, I assumed that I’d need to use a custom JDBC driver (based on a thread that I’ve read here and the Celonis documentation).However, when I reached the step to select the “Database Type” in the connector, I couldn’t find a fitting option for MongoDB.Has anyone successfully set this up?Should I use a specific Database Type in this case? Or is there another workaround to get MongoDB connected?Thanks for any tips or guidance! 

One of the topics that always pops up when I discuss the principles behind BPM with our clients is that of the process hierarchy, or sometimes called process taxonomy. It is the way you have organized your business processes in a layered or nested structure. Quite often I resort to the analogy of sorting LEGO bricks. Imagine you have a large pile of unsorted LEGO bricks and you want to build all the sets from the LEGO City collection (think about the police station, the fire station, the train station etc). All the bricks you need are in the pile, but finding the right brick can take forever. Even if you scan the entire pile of bricks and apply AI to it, you would still only be capable of quickly determining if the right brick is in the pile, but it will be quite difficult to pin-point the right location. So, what do most experienced LEGO builders do? The first sort the pile into smaller piles (I used to do this using old shoe boxes) and they can do this either by color, or by function

how to get badge after completion

Problem Description: In a Celonis Studio View, I am creating a column chart to display process cycle times, which are categorized into different intervals (buckets) using a CASE WHEN statement. During this process, two main issues occur:Missing Columns/Categories for Empty Data (NULL / No Data): If no records exist for a specific interval within the current filter context, the corresponding interval is not displayed as an empty category (a bar with value 0) on the X-axis; instead, it is simply omitted from the chart. This results in missing segments along the axis labels, making interpretation difficult.Incorrect/Alphabetical Sorting of the X-Axis: Since the CASE WHEN statement returns text labels for the intervals (e.g., '-15 - 0', '0', '1 - 9'), Celonis defaults to sorting the X-axis alphabetically rather than logically/numerically. Manually overriding this via the Sorting field with a separate numeric PQL formula (CASE WHEN ... THEN 1 ... THEN 2) does not yield the desired result, a

I’m currently exploring the Celonis Machine Learning Workbench and looking for some practical examples to learn from.I found the official PyCelonis Examples repository on GitHub, which is very helpful for understanding PyCelonis and interacting with Celonis resources. However, I was wondering whether there are also repositories, sample notebooks, or templates focused more on end-to-end ML Workbench use cases.I’m particularly interested in Supply Chain examples such as:Late/on-time delivery prediction Supplier or procurement risk prediction Inventory-related predictions Lead time prediction Bottleneck or process outcome predictionI also came across the older O2C – On-Time Delivery Prediction example, but it seems to rely on the legacy celonis_ml package.Does anyone know of any up-to-date PyCelonis 2.x / ML Workbench GitHub repositories, Jupyter notebooks, or sample projects that could be used as a starting point?And if you have built something similar yourself and can share the general

Quick one: we recently ported an analysis to a view, and we had an Action Flow set up that exported the analysis as a PDF and sent it to a customer via email. Now that we've moved to a view, is there any way to do the same thing — PDF export of a view, sent via Action Flow?

Running into a confusing behavior with our Delta Load setup and hoping someone can shed some light on it.We have a Data Job set to Delta Load mode, with several tables that don't have a Delta Filter configured. According to the docs, those tables should automatically fall back to a Full Load. For most of them that seems to work fine, but for some we're getting duplicate rows instead — identical in every column except _CELONIS_CHANGE_DATE.What makes it stranger: if we manually trigger a Full Load for those same tables, the duplicates disappear and everything looks clean. The problem only comes back when those tables run through the Delta Load job again.Two questions:Why does the "implicit Full Load" within a Delta job produce duplicates rather than overwriting existing data? Why does it behave differently from a manually triggered Full Load?

The Process Explorer component keeps loading indefinitely when connected to one particular Knowledge Model. Everything else seems fine — Variant Explorer and Case Explorer work without issues on the same KM, and if I switch the Process Explorer to a different Knowledge Model in the same view it loads perfectly.Already tried creating a fresh view, adding a new Process Explorer component from scratch, and reconnecting it to the same KM — same result. The issue follows the Knowledge Model, not the component or the view.Anyone seen this before, or have an idea where to start looking?

Working on Celonis Training on “ Get Data into Celonis”. Now, I am not being able to "Create Data Pools" from "Data" Tab at Celonis Studio.Tried to send a ticket at Celonis “Help Center” but “Submit” button is not activated for me. Anyone can help ? Thanks in Advance.

I have the below two columns. I want to extract the portion of the string matching the value in the ‘Seq’ column. Example, for the first row, the seq value is 400. In BIND column, I want to extract the strings that match the value the 400.Till the semicolon. using PQL. 

Running into an issue with the translation CSV workflow and wondering if anyone has seen this.When exporting translations, some entries contain HTML and CSS that the editor generates automatically — happens as soon as you use standard formatting like bold, font color, or font size in a textbox. The output ends up looking something like:<div><strong style="color: rgb(0, 0, 0);">Some text here.</strong></div>Those inline styles contain semicolons, and re-uploading the CSV fails because of them — even if nothing was changed in the file. The only thing that actually unblocks the upload is deleting the entire row, reference tag and translation both. Clearing just the translation value and keeping the reference tag doesn't work, the upload still fails.The problem is that this affects a large number of entries, so every sync cycle means manually identifying and stripping out all the styled rows before the upload goes through. Is there any way to allow semicolons in the

Many of you already know the Use Case Club - and some of you have even taken the stage yourselves.Join us again or drop in for the first time so you don't miss out on these unfiltered, practical insights!Forget polished corporate slide decks: At the Use Case Club, we take an honest look under the hood. Champions share raw, 10-minute snapshots of how they solved real operational bottlenecks and scaled results in practice.📅 When: Wednesday, September 23 | 15:30 – 16:30 CEST📍 Where: Virtual (Customer-exclusive format)What to expect: Real Insights: Raw, unpolished snapshots of actual live implementations. Direct Q&A: Answers and exchange directly from fellow Celonis practitioners. Maximum Focus: Zero marketing fluff, pure peer-to-peer learning. 🔒 Note: This format is strictly "from customers, for customers" (excluding academic/consultancy partners).Are you against grabbing your spot for Global Edition #4 today?👉 [Register here for September 23rd]

Came in this morning to find our Studio environment basically broken. All data model variables had disappeared and were no longer connected to the knowledge models, so none of our views were showing any data. Nobody on the team had touched anything.Manually recreating the variables fixed it for most knowledge models, but two of them kept throwing "Failed to change Data Model" whenever I tried to select the variable — so those are still broken.On top of that, a lot of KPIs are now displaying oddly formatted names with a prefix and the data model name included, something like km-master.kpis.some_kpi.displayname instead of just the plain KPI name they had before. Not sure if that's related or a separate issue.Anyone seen something like this, or have an idea what could cause variables to disappear like that without anyone actively deleting them?

Hello dear community members, A little while ago the process management topic received its own section at the Celonis community page and I would like to take this opportunity to introduce myself and provide a little bit of context around this little part of our larger community website.My name is Caspar Jans and I head up the global go-to-market team for the Celonis Process Management solution (hence the name of this section: process management). I've been in this space (BPM) for about 15-20 years now and I am looking forward to interact with all of you on this community page, at events or in customer calls.One important aspect to highlight here, is that when we speak about process management, we usually refer to process modeling and process governance and not about the process mining aspect (which is covered in the rest of the community website). We have a number of experienced CPM Specialists, Services Consultants and Support Engineers at the ready to answer your questions. I will al

Having a frustrating recurring issue and not sure what the actual fix is at this point.We have several Action Flows that send emails via an Outlook account. Every few days, one or more of them deactivate automatically with this error in the Send Email module:Invalid login: 535 5.7.139 Authentication unsuccessful, basic authentication is disabled.Reconnecting the email connection seems to help temporarily but the error comes back after a few days. Has anyone dealt with this, or found a more permanent fix?

Hitting a wall with an extraction of tens of millions of rows from Amazon Athena via the SQL editor. The pull itself seems to finish without complaint, but the pipeline then breaks at the Data Push Job step with an InternalException.A couple of things I've noticed:A similar extraction through the visual editor (same columns, but a much smaller row count) runs without any issue as part of the daily load, so I don't think it's the source system itself causing this. I only moved to the SQL editor because I needed JOIN and CAST logic that the visual editor doesn't support. Not sure yet whether it's the volume or the JOIN/CAST operations that's actually the trigger.Has anyone hit Data Push failures at this scale, or specifically with JOIN/CAST in SQL editor extractions? Any insight into what usually causes this would help.

Hi! I tried to add a new table to my data model (EORD) linked to EINE vía WERKS, MATNR, SYS_ID, MANDT, EKORG, ESOKZ, LIFNR.If I check this unique key in EINE ( Using: select WERKS,MATNR,SYS_ID,MANDT,EKORG,ESOKZ,LIFNR, count(*) AS Amount FROM ALL_NPM_EINE group by WERKS,MATNR,SYS_ID,MANDT,EKORG,ESOKZ,LIFNR order by Amount DESC ) and there is only one case per combination. So in a 1:N relationship it should work or? (EINE 1 → N EORD)Error Message: Case insensitive duplicated joins detected between tables: [EORD] and [EINE].

I'm working with a REST API that returns doubly-nested JSON (think: a top-level array of objects, each containing a nested array, which itself contains another nested array). Celonis splits this correctly into three separate tables, and the parent key from the top level propagates fine into the first nested table. The problem is that it stops there.The second-level nested table has no reference back to the top-level parent. The intermediate ID that's available at that level isn't unique (it's a type identifier that's the same across all top-level records), so I have no way to join the second-level table back to a specific top-level record without getting a cartesian product.The root issue as far as I can tell: in the Connector configurator, you can only choose from fields that exist in the API response as Primary Key. The propagated parent key that Celonis adds to the first nested table isn't available as a selectable column when configuring the second nested table, so there's no way t

Working on an OCPM data model and running into an issue where transformations based on SAP tables won't preview or publish anymore, regardless of which specific transformation it is, even standard Celonis ones fail as long as an SAP table is involved.A few observations:Transformations on other source systems (e.g. Dynamics AX) seem unaffected, so it looks specific to SAP tables somehow. Saving with validation fails outright, saving without validation works, but publish still isn't possible afterward. Existing, unchanged transformations still seem to run fine in the backend, so this feels isolated to the Objects & Events interface specifically.Anyone run into something similar with SAP-table-based transformations?

Been pulling my hair out on this one. I'm the admin of our space and have full permissions, but the gear icon in the ML Workbench that's supposed to let you adjust memory allocation is just... not there. Checked on multiple browsers, different devices, had a colleague confirm it's not showing up for them either.The script I'm trying to run keeps failing or hitting memory errors, and it's something our team relies on daily, so this is becoming a real problem. I've gone through the usual troubleshooting (refreshes, browser switches, confirmed admin rights) and I'm stuck.Is there something I'm missing on the permissions side, or is there a way to enable this?