How to calculate correct parent-level percentages in a Tableau Sankey chart with Path bins and duplicated rows?

Hi everyone, 

I’m building a Sankey chart in Tableau and I’m struggling with the correct percentage labels for each level. 

My data structure is a pre-aggregated funnel table with these fields: 

• ds
• Step1
• Step2_Filter
• Step2
• Step3
• Step4
• users
• Table Name

The Sankey is built using duplicated rows via Table Name: 

• Orders
• Orders1

Then I use a Path calculation to create the left/right side of the polygon: 

// Path
IIF([Table Name] = {MIN([Table Name])}, 0, 97)

Path is then converted to Path (bin) with bin size 1. 

The funnel logic is roughly like this: 

• Step1: No Order / With Order
• Step2_Filter: Card / Pip
• Step2: Pip Enabled / Pip Disabled / Card Enabled / Card Disabled
• Step3: ... → Pip Confirm, ... → Card Confirm, ... → No Confirm
• Step4: ... → Purchased, ... → Lost after Confirm, or NULL for No Confirm

Example filtered view: 

Step1 = No Order
Step2_Filter = Pip

Then the first visible split is: 

Pip Enabled
Pip Disabled

I want the labels to show users and percentages correctly at each level: 

1. For Step2, I need each branch as a percentage of the selected base:

Pip Enabled / (Pip Enabled + Pip Disabled)
Pip Disabled / (Pip Enabled + Pip Disabled)

1. For Step3, I need each branch as a percentage of its parent Step2 node:

Pip Enabled → Pip Confirm / Pip Enabled
Pip Enabled → No Confirm / Pip Enabled
Pip Enabled → Card Confirm / Pip Enabled

1. For Step4, I need each branch as a percentage of its parent Step3 node:

Pip Enabled → Pip Confirm → Purchased / Pip Enabled → Pip Confirm
Pip Enabled → Pip Confirm → Lost after Confirm / Pip Enabled → Pip Confirm

The Sankey polygon itself works, but the percentage labels are incorrect because the view has multiple technical dimensions: 

• Path (bin)
• Table Name
• Step2
• Step3
• Step4
• table calculations like INDEX(), RUNNING_SUM(), WINDOW_MAX()

Because of this, when I try to calculate percentages in Tableau, the denominators are either duplicated or calculated at the wrong partition level. 

Current core Sankey calculations: 

// Flow Size
SUM([users]) * (1 - [Whitespace]) / TOTAL(SUM([users]))
// Position-L
RUNNING_SUM([Flow Size]) - [Flow Size] + [Whitespace L]
// Position-R
RUNNING_SUM([Flow Size]) - [Flow Size] + [Whitespace R]
// Curve Max Win
[Position Max Win L] + ([Position Max Win R] - [Position Max Win L]) * [Sigmoid]
// Curve Min Win
[Position Min Win L] + ([Position Min Win R] - [Position Min Win L]) * [Sigmoid]
// Curve Polygon Win
IIF(INDEX() >= 50, [Curve Min Win], [Curve Max Win])

For label positioning I currently use something like: 

// Label
IF [t] = 6 THEN WINDOW_MAX([Bar Position]) END

What I need help with: 

How should I calculate the correct denominators for each Sankey level so that: 

• Step2 % = Step2 users / selected base users
• Step3 % = Step3 users / parent Step2 users
• Step4 % = Step4 users / parent Step3 users

Should this be done in Tableau using table calculations, or is it better to pre-calculate parent denominators in SQL and only use Tableau for rendering? 

If SQL pre-calculation is the better approach, what grain should the SQL output have to work correctly with Tableau Sankey densification?