the results are representative of all conditions.

uncontrollable external factors might influence the results.

the results are worse compared to experiments with internal validity.

 the experiments are done under real-world conditions.

Drill-down story

Scrollytelling interface

Whisky glass structure

Martini glass structure

Items can belong to multiple clusters.

Clusters can be represented as coherent blocks in a single heatmap without duplicating items.

Items are grouped hierarchically.

The result can be visualized using multiple, visually connected heatmaps.

Because there is often no tool to compare it with.

Because the analysis steps are often very time-consuming.

Because the tool cannot be tested in real-world environments.

Because measuring insights per minute is not feasible.

preserves angles.

distorts areas.

uses a plane to map a sphere to a plane. 

uses a cylinder to map a cone to a sphere.

change the transparency of the dots.

sample the data attributes.

subsample the data items.

remove outliers.

The channel position is not available for encoding data onto maps.

A scatterplot relies on the continuation grouping principle.

Chart junk increases the data-ink ratio.

Chart junk decreases the data-ink ratio.

One view can show a data subset of another view.

They are equivalent to small multiples.

They make use of the multiple-coordinated views principle.

The navigation can be shared between views.