Tile tool not behaving as I'd expect

Question

edit: included 2nd half of question as got truncated

Hello community, I'm hoping someone can explain what's going on here to me.

I've got some options which I'm trying to split into 2 equally sized groups (within the Group field)

This sounds like a perfect use case for the tile tool, but I don't get the output I'd expect.

When I use Equal Sum setup like this:

It outputs unequally split tiles:

If I generate rows and use equal records setup like this:

It outputs unequally split tiles, and also only creates one tile for Group 4:

I can get the correct answer if I setup the tool in equal records to sort by item like this:

Which gives me this (correct) output:

I've run this workflow in versions 2025.1 and 2025.2 with AMP on and off, but results are consistent.

Can someone explain what's going on here please

I've attached the workflow and it's here too:

splitting groups.yxmd

CoG · Accepted Answer

The Tile Tool can be very useful in certain circumstances although I do not think it is the right tool for this job. The Tile Tool appears to take a greedy approach in grouping the data (for the sake of speed). This is what causes the problem in all of your cases (presently):

1. Equal Sum - effectively performs a running total until the sum is greater than the target average expected tile sum of all records being tiled (or all records within a particular group), then splits the tile there or at the prior row depending on which is closer to the target sum. It does not analyze permutations to find the optimal tiling (which is what it seems like you are looking for).

2. Equal Records (both): Does exactly what it says, the tricky bit is that you set "Do not split tile on column": [Option]. This does not mean that [Option] will always be kept together in a tile, only that if Alteryx planned to split a tile and [Option] was the same for the records immediately before and after the split, then the split would be delayed until the next immediately available splitting point.

Tiling (via Tile Tool) is generally, in practice, more of an approximation. If you need an optimal tiling, you will need to build a macro that considers the different permutations and selects an option with the lowest total deviation from the target.

Hope this helps and Happy Solving!

CoG · Accepted Answer

I will illustrate the difference with two example tables

No sorting ("Doesn't work" example):

GroupOptionItemNote4A1 4A2 4B1 4C1Can't 4C2Split Here4C3 4C4 4C5

The table above is "unsorted" and the tile tool using "Equal Records" method would try to break into two groups of 4, which would place the split between items 1 & 2 of Option 'C' (green-blue divide above), but because "Do not split on [Option]" is selected, this is not allowed, forcing the Tile tool to keep adding rows to tile 1 until the next suitable break point is found, which, in this case, does not exist (and thus all of group 5 will be labeled as tile 1)

Sorting on Item ("Can work" example):

GroupOptionItem4A14B14C14A24C24C34C44C5

We follow the same logic as above, except now, the split happens between Option 'A' and 'C', which does not violate the "Do not split on column" rule, thus the expected behavior is achieved (as far as equal records is concerned), the difference now though is that there is an option 'C' listed in both tile 1 and tile 2 (which it seems is not desired behavior).

To solidify the concept. In the first case if we change the last row to be Option 'D', then the tile tool would fail to break at the desired 4 record position (because of the "Do not split" rule), and continue on until the next available split point between 'C' and 'D' like so:

GroupOptionItem4A14A24B14C14C24C34C44D5