The new Decomposition Tree visual

This has been in preview for yonks, but now sees the cold hard light of day:

A warm welcome, please, for the new Decomposition Tree visual.

This is one of the more complex visuals, and as a consequence this is going to be a longer than usual blog page!

An outline of what the DT visual does

Suppose you're looking at a value - here it's the total quantity of goods sold in our portfolio of Construct-a-Creature stores:

You can choose to break down your chosen statistic or measure by any of the dimensions you've attached to the visual, in any order.

Here I've chosen to break sales down firstly by region, then by the environment in which each animal lives and finally by the type of retail centre:

Viewing by region, environment and centre type.

Creating a DT visual

The field well for the DT visual is deceptively simple:

You can choose to analyse any single statistic (here I've gone for the quantity of goods sold) by one or more dimensions.

Note that it doesn't really matter in which order you put the explanation fields, as you can choose in which order to apply them within the DT visual.  The above choices would give you this visual initially:

I've changed the default formatting, but otherwise this is pretty much what you'll see initially.

Hover over the tiny + symbol and you'll see this:

You can choose how to explain this single figure.

Here I've chosen to expand firstly by region:

I can remove any level by clicking on the cross shown, or click on any + symbol to choose another explanation field.

I can keep expanding levels to get to the visual shown at the top of this page (although initially the formatting will be simpler).

Formatting your title

The top of my visual looks like this:

My visual is showing the titles and subtitles for each level.

Here are the relevant formatting settings:

You can give different colours to your titles and subtitles (and also hide your subtitles, if you don't want to show the currently selected item at each level).

Formatting your tree

There are a host of options for this:

The Density can be Dense, Default or (as here) Sparse.  As you might imagine, the Sparse setting requires most space on screen. 

Instead of using the default connector shape, you could choose Round instead (the only other option):

Spot the difference? There isn't much.

Changing the default action to collapse

When you click on a node of the tree, what should happen?  One option would be to tell Power BI to collapse the node instead of just selecting it:

If you choose this option, when you click on a node it will collapse or expand.

So clicking here will collapse the Air environment node:

Click to hide/display the centre type dimension breakdown.

Formatting the data bars

You can set what colours Power BI uses to display the data bars:

These are the settings I've used. I've made my bars a bit wider than they would be normally - 123% of normal, in fact - and set them to scale to the parent node.

Here are the 3 scaling options:

The options are explained below.

To illustrate how these work, consider the Water environment node below:

The percentage of the Water node which is coloured pink is 1331/12731 (that is, the proportion of the total figure for the parent node). 

Here are the 3 options:

Conditional formatting data bar colours

You can even tell Power BI to colour data bars proportional to the number they show (or to some other measure):

Turn conditional formatting on for the data bar colours, then click on the link shown.

This is what you'd see if you set the colour scale to go from a minimum of 0 to a maximum of 20,000 for our example, and didn't choose the colours wisely:

The bigger the number, the darker the bar.

Locking levels

Sometimes you'll want to freeze the top levels, but let people play about with ones further down the tree.  Here's how to do this:

Hover over the title of a level and the padlock symbol will appear. Click on this to lock the level.

You can't remove a locked level without first unlocking it:

The Region level is locked, so we can't remove it.

Standard features supported by DT visuals

A decomposition tree visual supports sorting:

You can sort by any level.

It also supports drill-through, if you've set it up:

If you've set up drill-through, this will work when you right-click on a DT visual node.

AI Splits

As if all of the above weren't enough, you can choose to apply Artificial Intelligence (well, ish) when choosing how to break down a node.  This is a feature that you can turn on or off:

I've turned the feature to Enable AI splits back on.

When you click to expand a node, you can choose one of the options with a light bulb next to it:

The extra two options you'll see for AI splits.

Here I've chosen Low value, so Power BI has automatically selected the environment or centre type with the lowest sales:

Power BI decides which dimension to show next based on your choice, and selects the node within it.

Using absolute or relative analysis

Suppose that I expand this:

I want to show the breakdown of sales for the Shopping Centre centre type, and choose Low value.

Here's what this would normally give:

Out of all of the environment and region dimension values, the one which gives the lowest sales is East Anglia, at 2,135 - so that's what Power BI selects.

However, you could make a case that this isn't actually that low a value.  There are lots of regions, and the sales for East Anglia aren't actually that much lower than those for the next lowest region by sales (East Midlands, at 3,745).  Let's change the Analysis type to Relative:

I've changed the analysis type from absolute to relative.

If I now choose to show the lowest value for the Shopping Centre node, I see this:

The sales for the Air environment (at 4411) are higher than those for the East Anglia region (at 2135), so shouldn't really appear.  However, the Air environment's contribution to total sales at this level for all environments is far lower than the proportion which East Anglia contributes to the total sales for all regions.

Conclusion

I've been waiting to blog on the Decomposition Tree visual until it emerged from preview.  I'm really impressed with it - what do you think?