Why not analyzing a dataset as if it were an image (with brushes, erasers,)?
I know this might sound like a stupid question, but while I was developing Datastripes (https://datastripes.com/), which is a visual data exploration web engine, I wanted to try and draw a dataset as an image, or, in other words, as a heatmap.
I tried using a brush, an eraser, and a color picker as tools. For example, the color picker is the filter, and the brush does averages and sums.
The experience was an out-of-body experience :D