# DEALING WITH DIMENSIONALITY

What if my dataset has many dimensions, say more than three or four? Does multidimensional variance analysis still provide useful insights once the number of dimensions starts becoming a challenge for our - human - capacity of understanding?

This is the core challenge of 'Mparanza. 'Mparanza is designed to manage multidimensional datasets, which in itself poses non-trivial technical challenges. However, once these challenges are met, we are left with a critical question: will the - inevitably complex - result of the computation be meaningful for the human user?

This is especially true when the dimensions are independent. When the columns are tied with hierarchical relationships (for instance city-region-country, or product-brand-category) the processing algorithm is relatively straightforward: the number of possible combinations between dimensions remains limited. More importantly, human intuition has a much easier time finding grips that lead to understanding.

If dimensions are independent, the number of combinations explodes. The app takes more time to run. It also becomes a more difficult for the human user to understand the interrelation between dimensions.

An example of a "tough" dataset is "30 Rewards". This dataset is not especially big, only 32,000 rows, but it has ten independent columns and returns 240k combinations. It takes more time to run.

Leave the default options, hit submit, wait a little while.

The tool returns 5 out 240k combinations that "explain" the change. You need to understand that this particular set of results is but one of hundreds, if not thousands, of other combinations of results that "explain" the change in revenues. 'Mparanza allows you to quickly search for and analyze the most insightful sets of combinations. Something that would be practically impossible to do with Excel.

This particular combinations enlightens us to the fact that most of the growth is due to Online sales, that have growth 22% year over year.

To understand more, we can drilldown on the first result row. We find out that it is mostly "non-active", "cristal" customers, that are driving revenues and growing at over 30% per year. In this second report, we are looking at the interplay of 6 different dimensions (Market, Active, Category, Payment, Channel, Lever) that explain a change of 310M.

This second report might or might not deliver the actionable insights needed the a specific situation.

The tool offers a variety of tricks and functionalities to help pin down the "right" set of combinations.