topic Re: Sankey Diagram Count Distinct > 100% in App Development

Sankey Diagram Count Distinct > 100%

newuser — Thu, 02 Jun 2022 14:28:36 GMT

I've seen a few posts mentioning Sankey diagrams resulting in more than 100%. Has anyone encountered this, and if so, what is the reason it happens? We've found depending on the dimensions used, the %'s change, e.g., using two vs. three dimensions in the Sankey can vary the parent %'s shown. We are counting distinct and then dividing by the "total".

Re: Sankey Diagram Count Distinct > 100%

stevedark — Thu, 02 Jun 2022 19:51:49 GMT

Hi @newuser

This would happen if the values that you are getting a distinct count of appear in more than one of the dimension values. If you picture having the following table:

Dim1	Dim2
A	Z
B	Z

Then if you count distinct Dim2 within Dim1 you would have:

Dim1	Count
A	1
B	1
TOTAL	1

And if you then view the dimension values as percentages of the total:

Dim1	Count
A	100%
B	100%
SUM	200%

That's obviously an extreme example, but if you have many Dim2 values across Dim1 and a few of them appear in more than one Dim1 values you could end up with a sankey total over 100%.

If you count distinct a composite value of Dim1 & '-' & Dim2 then that would remove that risk of going over 100%, as a duplicate value would double count in the total, as well as against each dimension.

Hope that makes sense.

Cheers,
Steve

Re: Sankey Diagram Count Distinct > 100%

newuser — Fri, 03 Jun 2022 00:03:03 GMT

Thank you @stevedark ! When using the composite field, would that preclude me from using multiple separate dimensions in the Sankey? In my case, I do have instances where multiple Dim2's reside in a Dim1.

I'd still like to use Dim2, Dim1 as separate dimensions in the Sankey, so I am assuming your composite key suggestion would still allow for that?

Would the composite field be in the highest-level parent field only (left-most field) or any parent field needing child fields grouped (if that makes sense)?

Re: Sankey Diagram Count Distinct > 100%

stevedark — Fri, 03 Jun 2022 07:36:58 GMT

Hi @newuser

You don't need to create it as a field, or add it to a dimension, just reference the composite value in the Measure of the sankey chart:

count(DISTINCT Dim1 & '-' & Dim2)

If both Dim1 and Dim2 exist in the same source table then you can make your app more efficient by concatenating on load to a new field that you can reference in the count.