How does Ratio of cases with certain process flow work?

Hi Tomas,

A quick and dirty idea, can it be that there cases where V->V happens more than once?

If this is the case, It could be that the process explorer count the number of cases where V-> V happens and the ratio counts the total number of v-> v (this ratio should be greater as V->V happens sometimes mre than once per case)

Hope it helps

Javier

Hi Javier,

thanks for your answer. I had the same idea, however when I checked path V->V->V there is no such result (Scrnshot 1).

My ultimate goal is to make table where I'd be able to see happy path ratios for different countries. When I try I still get strange

numbers but in this case, this one is lower than the one obtained from Algorithmic Happy Path ratio as you can see in screenshots below, I filtered out country (in this case SK) and from editor I get 20% ratio, however from process overview (with same selection) I get roughly 31% of cases.

When you leave the editor mode, does it still calculate as 2.1%? Sometimes the preview does not consider some other filters you might have applied.

If yes, can you do another check directly on the process explorer by applying a sheet filter as FILTER PROCESS EQUALS 'V' TO 'V' END and see if you also get 221 cases?

Hi Gabriel, no change after leaving edit mode Regarding the filter in process explorer, not really sure how to apply it with the "END" condition, if you could guide me please.

click on the 'wheel' and Load Script. Paste

FILTER PROCESS EQUALS 'V' TO 'V' END;

and check how many cases you have.

Thanks Gabriel, still 221 cases

Hi again Thomas,

When I see the ratio ('V' to 'V' to 'V') I think it only find the cases where thre stages V happen inmediately after each other.

Is possible that the process have V=> V then other activities and then V=> V?

When you change the KPIP of the process explorer from Case Frecuecny to activity Frecuency which result do you received 1.6% or 2.1%?

Hi Javier, I guess you are correct with your assumption, however there are no other connections from V than to "end" or V itself, 100% of connections is visible in the first slide.

After changing from case frequency to activity frequency, I'd get +221 on V

Hi,

If your ultimate goal is to make table where you'd be able to see happy path ratios for different countries, I think I have a good solution.

If you go to the conformance tab, you have the option to write the outcomes to a variable. Create a variable like 'conformance" and use this option. See image: 

Then you can use this formula to calculate the % of Conforming cases:

AVG(CASE WHEN PU_SUM("TABLENAME", ABS(<%=conformance%>)) = 0 THEN 1 ELSE 0 END)

Now all you have to do is put country as dimension in an OLAP Table and this formula as KPI. And then define the "to be flow" in the conformance checker.

This way, if you change your "to be flow" in the conformance checker, the percentages will automatically update as well.

Hi Joren,

thanks for your advice, this looks promising! However, I am still not able to get correct result.

If I omit average function, the distribution looks correct (img1), but when I add AVG function, result seems odd - I would expect result roughly 72% (1235/1735), not 59%. Have no idea, where the 59% came from. What is wrong in my formula?

Really sorry for probably basic questions, I am quite new in this, so trying to explore capabilities of Celonis.

Hi Tomas,

No worries, the formula seems fine. I think all we have to do is set up the conformance check and link it to the conformance variable. I have described it step by step below with images.

First: create a new variable called "Conformance" or however you would like to name it. In the screenshot you have provided you used "Conformance" with a capital so that will work. To do this: go to settings and click on variables

Then create a new variable and name it "Conformance"

The next step would be to go to, or to create a "Conformance check" page in your analysis.

Then mine the target process:

In there select the acceptable flows.

When done, click on this settings icon:

Then check the write conformance query to variable and select the variable you created earlier.

You can than test if the outcome is correct by comparing the % shown in the conformance check and the % shown in your newly created KPI. They should be identical.

I think I have this correct (see screenshots), but there is no positive impact on the result:

variable:

conformance tab (btw, here you can see, that the numbers are correct, just the average did not make sense)

result still 59%

Thanks @joren.kamer12 for sharing, awesome answer for learning!!

Can you go into this overview and see what % conforming cases it shows here? Ignore the red circle it is the same screenshot as earlier

Also can you double check that you don't have any active filters selected?

only that filter on the country

The percentage here should be equal to the outcome of the table/formula.

There are three levels of filters: Analysis wide / Analysis sheet / Table itself

Do you have any active filters on the Table or Analysis sheet level?

Perhaps one last test: Can you put the same formula in a "Single KPI - Number" and see whether it differs from the exact same formula in the OLAP table?

Hi Javier, thank you.

In case you are interested:

Another great addition to this is a Pie Chart with the main deviations. So that if you make a selection, of lets say a specific country, you also see the main deviations and its proportions.

Would look someting like this:

with single KPI, the value seems to be correct, but the value in OLAP table (upper left corner) is still 59% - even though the table filter is empty (second screenshot) 

Good to know, the formula works! The problem lies with the table. I mean the table should show the same result as the number...

I once had a similar issue and raised a question on this forum too. Somehow it just worked after 2 days when i changed nothing...

Based on your previous experience, is there anything you'd recommend to avoid such situation?

I mean, here I was able to check whether the number is OK or not quite without any problems, but with same problem for more complex calculations, it might be almost impossible to spot any inconsistency.

Best to do, I guess, is to do what we just did: input the exact same formula somewhere else like a Single KPI - Number and see if it matches

Thanks a lot Joren for your help.

I want to believe that someone from Celonis team could have more convenient solution.

I completely agree...