Here is what you should know about Datama to understand what is behind the solution

1. Waterfall Analysis

Identify which step of the funnel is driving the change. Better understand how to interpret the waterfall.

See details here

2. Dimension Analysis

Identify which dimension in a specific step is driving the change with some details on Mix and Performance effect.

See details here for Mix Effects and for performance effects

2.1 Mix Effect

For each step, analyze how much mix effects on each dimension contributes to the observed gap

See details here

2.2 Segment Performance

For each step and each dimension, identify which specific segment performance is driving the observed gap

See details here

3. Other Modeling Components

3.1 Clustering

Clustering is necessary for dimension analysis:

For discrete dimensions, anything below X% (X=2) of primary numerator is aggregated into “other”
For continuous dimensions, cuts are made using weighted decision tree methodology, in order to create coherent buckets.

Read docs related to continuous dimension

3.2 Interdependence

In ‘Safe Mode’, most correlated dimensions are flagged. Interdependencies between dimensions are tested using Chi-Square and simple business calculation.

3.3 Combined Dimension

Combined dimension is created by concatenating all clustered dimensions into one “Combined_Dimension”. It is then considered as all other dimensions and its contribution in the variation performance is assessed as it is for the other dimensions.

3.4 Significance

In ‘Safe Mode’, simple check of minimal volume (manually inputted) for given metric in Start and End. You can also use Datama Impact to assess properly significance of variations.

3.5 Scope

‘Out’ segment defined in column ‘Scope’ is excluded from analysis, and simply stacked on Start and End column in waterfall chart.

3.6 Covariance

A Covariance ratio appears on the top left of the waterfall.

For waterfall analysis, covariance is distributed on each step. User should check that it remains reasonable (typically, <30%).

For Dimension analysis, covariance is not distributed on neither mix nor performance sizing. Hence user should be careful when looking at dimension impact.