The Analyst's Path

Data-ink ratio

The governing idea is Tufte's data-ink ratio. Of all the ink (pixels) in your chart, what fraction actually encodes data? Every non-data pixel is noise, competing with the signal for the reader's limited attention: a heavy gridline, a chart border, a redundant legend, a drop shadow, a 3-D bevel. So the work is subtractive. Draw the chart, then remove everything that isn't carrying the message, and stop just before you delete something the reader needs.

3 min readPart of Charts & Dashboard CraftOpen in course mode

Advertisement

The governing idea is Tufte's data-ink ratio. Of all the ink (pixels) in your chart, what fraction actually encodes data? Every non-data pixel is noise, competing with the signal for the reader's limited attention: a heavy gridline, a chart border, a redundant legend, a drop shadow, a 3-D bevel. So the work is subtractive. Draw the chart, then remove everything that isn't carrying the message, and stop just before you delete something the reader needs. A clean exhibit is minimalist because attention is scarce and it's the reader's attention you're spending. Fashion has nothing to do with it.

What are the working rules for a clean chart?

The working rules, each tied to why:

  • One message per chart. A chart that says three things says nothing, because the reader can't tell which of the three you brought them here to see. Split a busy exhibit into two clean ones, or pick the one comparison that matters. It's the Minto Pyramid Principle applied to pixels. Lead with the one governing idea, support it, and don't bury it under co-equal rivals. The analyst communication toolkit teaches the same doctrine for prose. A slide's chart and its headline should make the same single claim.
  • The title states the message, not the metric. Excel's default title is the column name, "Revenue." A professional title is the sentence you want the reader to leave with, such as "Revenue grew 38% over the year, led by the festive quarter." Let the metric name the axis and the title name the conclusion. That one change does more for comprehension than any formatting.
  • Direct-label instead of legending. A legend forces the reader's eye to ping-pong between the colour key and the lines, holding a mapping from colour to name in working memory. Put the label at the end of each line or on each bar, and the mapping vanishes, because the reader reads the name where the data is. Use the legend only when direct labels would collide.
  • Delete non-data ink. Turn off the chart border and most gridlines. If the reader must read approximate values, keep a few faint horizontal ones at most. Drop the background fill. Thin out the tick marks or remove them. When a chart or dashboard is on a worksheet, set View ▸ Gridlines off for the worksheet itself.
  • Colour with intent, not by default. Use one calm accent to highlight the series or bar that carries the message, and grey for the rest, so the eye goes where you point it. Excel's default rainbow palette gives every series equal, competing emphasis, which is the opposite of a message. Your colours also have to survive colour-blindness and greyscale printing, so never encode meaning with red against green alone.

How do you check a finished chart?

So make it a habit. Build the chart and then interrogate every mark. Does it help the reader see your one message? If it doesn't, delete it. The finished exhibit should look almost empty and read almost instantly.

Advertisement

Part of the course

This page is part of Charts & Dashboard Craft (EX3.02), in Phase 3: Corporate finance and valuation. The module's flashcards and its quiz are in the app.