FIX & VARIABLE DIMENSION VARIANCE
Fix dimension bridge analysis is the type of bridge analysis you are accustomed to and you have seen hundreds of times.
Variable dimension bridge analysis is something you probably rarely saw in action possibly also because it is hard to implement in Excel. However, it is nothing new. When you say "the top sales driver was the Chinese market (dimension➡️country), the second direct sales (dimension➡️channel) to large accounts" (dimension➡️segment) you are doing variable dimension variance analysis: you are changing the dimensions of each result.
Fix dimension bridge analysis.
The dimensions are the same in every row. These dimensions can be simply variance types as in this waterfall...
...or can be associated with variance type and with one or more columns.
The fix-dimension bridge analysis waterfalls should cover the full amount of the variance value, with no residuals. In this case the residual is tied to the fact that we limited the number of results. If we change the setting to 9 results
... and nine results is enough to cover all the values in the dimension, the balance goes to zero
Variable dimension bridge analysis.
The app computes all possible combinations, ranks them according to user-defined parameters, and returns a set of facts that explain the change.
The dimension columns of every row vary both in number and in type, and are automatically selected by rank of importance. The app looks for the combinations that better explain the variance. In this example the first row result has two dimensions, the second has seven, and the third has one.
Since the selected combinations are the most significant, this allows to explain significant positive and negative change with a limited number of results. In this example, the sum of the absolute values of the first chart above is 21k. The sum of the first five entries of the second second is 27k. The sum of the first five entries of the variable dimension waterfall is 39k.
It is possible to change the ranking parameters, for example ask the app to give weight to % change instead of just to absolute value change.
If the standard number of requested results - five - is not enough to cover all the variance, a "residual" entry will appear. This is completely normal and does not indicate an error.
As in fix-dimension bridge analysis, every row is net of the rows above. However, since the dimensions change in number and type, after a few rows it can be difficult to keep track of what has been filtered and what not. By default the app returns five row results.