Analyst tools fail in two opposite directions.
Consumer-grade dashboards are too sparse. They look clean in a screenshot and force endless navigation in practice, because the person using them needs six numbers side by side and can only see two.
Legacy terminals go the other way — genuinely dense, and unreadable, because nothing distinguishes the important from the incidental.
The people on these screens all day want density. What they cannot work with is undifferentiated density.
Hierarchy, not simplification
The answer was never fewer numbers. It was structure strong enough that the eye knows where to land first.
Density is not the problem. Undifferentiated density is.
Every decision followed from that. If a screen has to carry a hundred values, then the design work is entirely about rank — what is read first, second, and only on demand.
That is a typographic and spatial problem far more than a visual one.
Hierarchy as the primary tool rather than decoration.
A consistent charting language, a strict typographic scale, and colour reserved for meaning rather than variety. Comparison was designed in from the start.
One charting language
The same idea always looks the same way.
Chart types were defined against use cases rather than chosen per screen, so a trend always reads as a trend and a comparison always reads as a comparison.
This removes a cost users never articulate but always pay: re-learning the page every time they open a new report.
Colour carries meaning only
In a financial interface, colour is semantic before it is aesthetic.
Positive, negative, neutral and alert states own their hues. Nothing else gets to use them, which is what makes them readable at speed.
Everything decorative is carried by weight, spacing and rule, which leaves colour free to mean something.
Comparison on a shared baseline
Anything a user would want to compare sits at the same scale.
Where two figures are related, they share an axis, a scale and a visual grammar — because the most common analytical error is a comparison between things that only look comparable.
Dense surfaces that get scanned rather than read.
The reporting screens carry substantially more information per view than the pattern they replaced, and are faster to read because rank is explicit.
The charting language was documented well enough to be extended without a designer in the loop, which is the only way a reporting system survives contact with new requirements.