Innovative Celonis Usage - Let's share & brainstorm ideas!



Hi Community, do you know of innovative/creative/new ways of using Celonis?

Feature of the week series #2 - Variant Clustering

Hi @a.boukaboub,
I don’t know if it is innovative/creative/new, but here are some ways we are already applying it:

  • using Celonis in Software Development to have a de facto additional Quality Manager in place. As we are developing new tools (e.g. a Quotation tool), we are monitoring our development by creating event logs in a way how we want to measure / be measured. Are the sprints really adding value towards our overall goals? The process explorer can easily show us based on the data trail we are creating whether the tool matches our expectations of our to-be process or not.

  • for a specific customer, we download our O2C Data with KPIs (Digital Fit Rate, Automation Rate, Rework Rate) and talk to them in our regular exchanges how the correlation is to their P2P process.



Dear Boukaboub,

We have setup an innovative “Target Fulfillment Reporting Cockpit”.
In that Cockpit the User has for each Process (Direct Shipping, Sales from Stock etc.) the possibility to put in the target value for PPI (Touchrate, Rework Rate, Automation Rate, EDI Rate, First Time Right etc.) He then can see from all kind of angles how far the different company are. For example 70% of target Touchrate is reached, 60% of target Automation Rate etc. etc. Due to this Target fulfillment procedure we normalize all data into % rteached. In this way you have an easy way to look at all the PPI across all companies / Product Hierarchy / Customer / etc.
Trends, How many hours saving are still to be done and of course going down unto activity or Objects.
This is really great because you do not need to setup any reporting outside Celonis and always have the latest data.

We also have a “Process Cost & Capacity Planning” Cockpit.
In here the same idea as above but input fields for alls important activities (based on process type); costs per hour for all our companies and yearly hours.
A user can calculate the cost per line item.
He can see where the costs are hidden as well how many FTE are needed for each activity.
Of course a complete drill down on dimension like company, product hierarchy etc. etc.
We also have nice trend graphs to understand the development of the fulfillment rates; costs per line item; capacity needed.
It is quite amazing!
Again with trend graphs.
It gives a great overview where for example automation investments should done first. It was for us an absolute eye opener. Many manager have no clue what the personnel is doing. This cockpit makes it clear that when processes are not running right due to all kind of reason you get according results.

I think the special part of Celonis is, that although a “BI-tool”, you have the possibility to have Input Fields. This is extremely useful.

Best Regards,


Hi Amina,
on my side I have one of my customer who came up with mining Celonis with Celonis.
In order to determine how much analyses were used and how they mined Celonis logs and now monitor how every user navigates between dashboards by visualizing it as a process.
In this way they can keep an eye on which analyses are the most useful, when users use them the most, how many active users there are, how adoption grows over time, and so on and so forth.


I forgot probably a couple others we are already using, but one which has stirred a lot of interest is this use case:

  • We are using Process Mining to also identify potential RPA/RDA use cases using a so called “RPA Scout”.

Background: We do have super RPA enthusiasts who have used our O2C Monitor to proclaim that a gazillion RPA use cases are floating around. Unfortunately, very many times this does not materialize because mathematical statistics are not correctly applied to the reality of the business to the ERP settings.

Example: In a particular country you see lots of Delivery block rework activities which are taken care of manually, thus, the RPA enthusiast says there is lots of gold to discover here.

Reality: There is not, because the Delivery block is due to various reasons which are not always following a specific rule set (or different rule sets), etc.

Solution: The RPA scout can easily visualize across different dimensions along the main x/y of absolute/% - this way, context and content are challenged and the user can easily see whether one is barking up the wrong tree. Furthermore, you can see all existing robot use cases by country, unit, activity, etc. to get an idea.


Not sure if these are innovative, but our different types of application are as follows:

  1. Central process info: we currently have 1 central data warehouse (“data core”) with curated data collected and combined from different source systems into 1 canonical view (i.e. customer info from all systems are represented as 1 table with clear definitions). Recently tables such as ‘case’, ‘process’, ‘activity’ etc. became available so we are currently investigating how to connect Celonis. This would allow us to build a generic dashboard with a ‘process selector’ to select which process you want to analyse and make this available for business users.
  2. Self-service training: we train the internal business users to use, and possibly extend, dashboards themselves. This allows them to be less dependent on our time. Celonis is user friendly enough to allow for this usage.
  3. Customer journey: we collected data from processes, website visits, phone calls, letters send/received, etc. and analysed full customer journeys. We for instance realized that people contacted us with questions before we send them a triggered information letter. We now send this letter earlier resulting in less incoming calls.
  4. JIRA analysis: we are building a dashboard for ourselves to confront us with our JIRA/agile habits (skipping workflows, exceeding deadlines, no updates, etc.)

About us I’m working at APG, the largest Dutch pension fund executor. We manage ~450 billion Euro’s in investments, but also execute the pension administration, payments, call center, etc. etc. for about 8 pension funds, resulting in 4.5 ‘customers’.

@a.petri : I would really like to know more about your ‘analyse Celonis usage with Celonis’ as this is very relevant for us (which dashboards are not used and can be cleaned up?).


Hi A.Boukaboub,

In many webshops we use heat maps to understand how the webshop is being used. This of course says nothing how a person came to result=buying an article.
With process mining we really could see all routes people are taking until they buy something…
You actually think of optimization of the webshop.
Of course you could think also about optimization of any kind of software by tracking how certain tasks are done…

Have fun!



Already long on our list is a top management cockpit across all datamodels. They want to see in an easy way how (for example) O2C , P2P etc. are running in one sheet/cockpit. Of course with drill down possibilities to the specialized cockpits.

Have fun!



Dear a. Boukaboub,

At the moment we have a huge topic and also many misinterpretations concerning the topic " amount of Variants". In O2C considering Direct Delivery and Sales from Stock we have more then 55K Variants.
Many managers think this is a disaster as a starting point. Bus as you know any new connection are new small activity (we have 120 unique activities in O2C) ends up in again a new variant.
To get an much better understanding of this topic and an what is driving the “complexity” we are setting up a variant cockpit. A typical analysis the see the trees in the wood again is to do a 2-k means clustering.
This would be a great starting point to get an better insight of the many variants and what it really means and how we can increase the right cluster.
This kind of analysis are of course also very usefull for RPA initiatives. There you want to automise “stable” variants…

Met vriendelijke groet / Best Regards,


Hello @Hans.van.der.Zandt ,
can you elaborate a little bit more on the variant cockpit you are building? Variant clustering is a hot topic and we would love to hear about your exact use case for that, e.g. on what distance metric and data you want to cluster?

We already implemented a CLUSTER_VARIANTS operator that will be part of the next IBC release and Celonis 4.5.

You can try it out here.

Let me know what you think!


Dear Nicloas,

My user story is exactly the topic to get a better understanding "what do theses many variants really mean and can we clusters them? Typically you could go for different dimensions. I first would like to see how close al these variants are to each other. One dimension would be the activities and the other could any other attribute.
It would be nice to understand which attribute is driving it the amount of variance . (in our case it will be certainly the product hierachy).
I read a nice article about using K.means: .
This one shows also the weaknesses of this 2 k means algorithm and favors the Anscombe’s Quartet. Maybe it would be smart to enable the possibilty to go through different algorythm automatically. Or have a choice of these side by side.
What do you think?



Could you please describe what analysis have you build for HR?
What analysis are mostly used for HR by majority or you think HR can be analysed by?



Dear Dina,

I am not sure to whom you are referring. As a principle you analyse a process to get out all the bottlenecks and gain efficiency or you want to get rid of compliance issues.
In my eyes: In HR probably the most important process is the hiring of new employees.
Some HR organisations still do 4-6 weeks hiring process and after that they first give all the candidates to the departments. They take another 1-2 weeks to analyse the candidates and then the invitation process starts. And they are very suprised that the top candidates have already accepted another job… Anyway a nice process to analyse if you ask me (when you have the data of course)
Maybe somebody has other example of a nice HR process.

Have fun!


Hi, Dina
I would like to add some extra ideas to response from Hans:

  • Variant analysis for hiring process such as job offer rejection on the late stages (i.e. 1 day before estimated start date) so you investigate reasons behind it
  • Concertion rate at hiring. I.e. high number of candidates reject at 3rd interview means that job position requirements could be misunderstood by hr
  • analyze employee life cycle using his ID as case id, so you can see how your motivation programm works (you can anonimize employees as you are looking for patterns, not specifics)