Insyt
All charts in the library

Part of a whole

Decomposition tree

Breaks an aggregate into successive dimension values to explore contributors.

  • Diagram
  • Contribution
  • Drill-down

What it answers

Which dimensions explain a large aggregate?

Shape
Branching aggregate-value nodes
What it shows
Large contributors and conditional breakdowns.

Example

The question

Why is revenue below plan: which region, then which product?

Revenue gap to plan broken down step by step (EUR million)

Revenue gap to plan broken down step by step (EUR million)The −3.0 million gap splits by region (North −2.2, South −0.5, East −0.3), and North splits by product (laptops −1.6, phones −0.4, audio −0.2). Most of the gap is North laptops.Total gap-3RegionNorth -2.2South -0.5East -0.3North by productLaptops -1.6Phones -0.4Audio -0.2
Revenue gap to plan broken down step by step (EUR million)The −3.0 million gap splits by region (North −2.2, South −0.5, East −0.3), and North splits by product (laptops −1.6, phones −0.4, audio −0.2). Most of the gap is North laptops.Total gap -3North -2.2South -0.5East -0.3Laptops -1.6Phones -0.4Audio -0.2Pink path: total, North, Laptops
The −3.0 million gap splits by region (North −2.2, South −0.5, East −0.3), and North splits by product (laptops −1.6, phones −0.4, audio −0.2). Most of the gap is North laptops.

Illustrative data. It belongs to no company.

When to use it

Use it when

Interactive contribution drill-down.

Avoid it when

Nonadditive measures decomposed with an invalid aggregation.

How to read it

Read each split as a partition of the current node.

Note

Dimension order may be selected interactively.

The data you need

Layout
Wide fact table
Fields
RecordID: text; Dimension1...DimensionN: categorical columns; Value: number
One row is
One measured record with its descriptive dimensions. What is the grain?
Rules
Define aggregation and hierarchy rules. Avoid double-counting already aggregated records.
Optional
Weight,Date,additional drill-down dimensions.
Preparation
Validate the input grain and missing values. Define aggregation and hierarchy rules. Avoid double-counting already aggregated records.
Units
Relationships and positions have no numeric unit by default. Label any counts, durations, probabilities or weights separately.
Example of the input layout: Wide fact tableA wide table: one row per item, with a separate column for each measure or period.RecordIDV1V2V3V4V5ABC
A wide table: one row per item, with a separate column for each measure or period.

Worked example

Rows North-A30, North-B20, South-A40 total 90. Split by Region: North 50, South 40; splitNorth by Product:30 and 20.

RecordIDRegionProductValue
R1NorthA30
R2NorthB20
R3SouthA40
All charts in the library