Data visualization
How to turn BCMS data into charts that answer a question at a glance. This is the reasoning behind chart choice and styling; for the components themselves, see the Charts section.
A chart answers one question
Before choosing a chart type, write the question it answers — “are we recovery-ready?”, “which activities carry the most risk?”. The question picks the chart. A chart that doesn't answer a clear question is decoration, and decoration costs the reader time.
| The question | Reach for |
|---|---|
| How does one value track over time? | Sparkline / line |
| How do a few categories compare? | Bar chart |
| How is a whole split into parts? | Stacked / segmented bar |
| How ready are we against a target? | Gauge / KPI progress |
| Where do risks fall by likelihood × impact? | Risk matrix |
| How does activity spread across two dimensions? | Heatmap |
| How does flow move between stages? | Sankey |
Lead with the answer, not the geometry
The headline number or trend should be readable before the chart is. Put the takeaway in text next to the visualization, and let the chart provide the supporting shape. This is also what makes a chart accessible — the answer is in words, not just pixels.
Show the key figure as text (“72% ready”) alongside the gauge that depicts it.
Don't make the reader infer the number by eyeballing an arc or bar.
Give every meaningful chart a short text/alt summary of its takeaway.
Don't rely on a legend's colors alone to carry the meaning — label directly where you can.
Be honest with scale
A chart is an argument about the data; it must be a fair one. The fastest way to lose a careful audience — and ours includes auditors and the C-suite — is a misleading axis.
Start bar-chart value axes at zero so bar lengths are proportional to values.
Don't truncate a bar axis to exaggerate small differences.
Keep axes, units, and time ranges consistent when charts sit side by side.
Don't mix a 7-day and a 30-day window in adjacent sparklines without labels.
Color in charts
Charts amplify the “color is the loudest signal” rule, because they often use many colors at once. Keep palettes small, reuse status meanings, and never make color the only encoding.
Reuse the status intents — success/warning/error — so a red segment means the same thing it does everywhere.
Don't use a rainbow of hues for categories that have no inherent order or meaning.
Add a second encoding (label, pattern, position) so the chart survives greyscale and color-blindness.
Don't rely on subtle hue differences a color-blind reader can't distinguish.
Density and restraint
BCMS dashboards pack many charts together. Keep each one quiet so the set stays scannable: minimal gridlines, no 3-D, no drop shadows, no chartjunk. The data is the ink.