Short answer
Something that cannot be mistaken for any value on the scale — usually a pattern or hatch, or a neutral grey clearly outside the scale's lightness range — explicitly labelled 'no data' in the legend. White and pale grey are the commonest choices and the commonest failures, because they look like the lowest class of a light-to-dark scale. Uncertainty is harder still: fading or desaturating a colour for low confidence shifts its apparent value, so pair it with a separate encoding.
A region with no data and a region with a value of zero are different claims, and a map that draws them in the same colour tells readers there is nothing where in fact nobody measured. The problem is sharpest on light-to-dark sequential maps, where the lowest class is near-white: a white or pale-grey no-data fill is visually a member of the scale. The table measures four neutral fills against a six-class blue sequential ramp. White and light grey sit close to the ramp's pale end in colour and closer still in greyscale; mid grey is further in colour but lands in the middle of the ramp's lightness range, so in greyscale it can pass for a mid class. A dark grey can work on a light-to-dark ramp but competes with the high end.
Because every flat neutral fill has some lightness, and any sequential scale covers most of the lightness range, there is often no grey that is guaranteed to read as 'outside' the scale. A texture solves this: diagonal hatching, dots or a crossed pattern is visibly a different kind of mark rather than another shade. Patterns survive greyscale printing and colour-vision deficiency, and they carry the right connotation — 'something is different here' — without implying a value. Keep the pattern low-contrast so it does not dominate, give it its own legend entry, and make sure it is described in the chart's text alternative so the gap is not lost for screen-reader users.
Uncertainty — a confidence interval, a small sample, a modelled rather than measured value — is often shown by fading the colour, lowering its saturation or increasing transparency. The difficulty is that on a sequential scale lightness is the value: a faded dark blue looks like a paler, lower class. Desaturation is gentler but reduces the hue difference that multi-hue scales depend on. Safer approaches separate the two dimensions: keep the value colour intact and add a second mark for uncertainty (a hatch overlay, a dot, an outline style, error bars or interval bands in charts). If colour must carry both, design a two-dimensional legend that shows value along one axis and certainty along the other, so readers can see exactly what fading means.
Colour and texture tell readers that something is missing or uncertain; only text tells them why. A legend entry 'No data (not collected in 2021)' or 'Estimate — fewer than 30 responses' changes interpretation far more than any shade. For interactive charts, the tooltip for a missing or uncertain value should state it explicitly rather than showing a blank or zero. And in data tables that accompany charts, use an explicit marker such as an em dash or 'n/a', never an empty cell or 0. These are also what WCAG's use-of-colour criterion expects: a distinction conveyed visually needs a non-colour carrier.
| Fill | ΔE00 to closest class, colour | ΔE00 to closest class, greyscale |
|---|---|---|
| White #FFFFFF | 8.2 | 3.5 |
| Light grey #E0E0E0 | 8.1 | 3.0 |
| Mid grey #9E9E9E | 19.3 | 1.2 |
| Dark grey #4D4D4D | 19.5 | 4.9 |
Why: No-data is drawn in the same pale colour as the lowest class.
Fix: Use a hatch pattern with its own legend entry.
Why: Uncertainty is shown by fading the value colour.
Fix: Keep the value colour and add hatching, outline style or interval marks for uncertainty.
Why: Missing values were coerced to zero in the data pipeline.
Fix: Keep missing as null through to the display and label it explicitly.
Each statement is labelled by kind — established fact, a standard’s requirement, observed market data, a convention, or Colourwise’s own interpretation or analysis — with the strength of the evidence behind it.
Colourwise analysisModerate evidence
White and light-grey no-data fills lie close to the pale end of a light-to-dark sequential ramp in CIEDE2000 terms, and closer still in greyscale, so they can be mistaken for the lowest class.
Based on: Computed by Colourwise against a six-class OKLCH blue ramp using CIEDE2000 and luminance greyscale; see the table.
Caveat: Results depend on the ramp; a ramp that stops well short of white leaves more room for a pale no-data fill.
Source: CIE (International Commission on Illumination) publications; Digital video colourmaps for checking the legibility of displays by dichromats
StandardStrong evidence
WCAG 2.2 criterion 1.4.1 requires information conveyed by colour, such as a no-data or uncertain state, also to be available by another visual means.
Source: Web Content Accessibility Guidelines (WCAG) 2.2; Understanding Success Criterion 1.4.1: Use of Color (WCAG 2.2)
Colourwise interpretationLimited evidence
On a sequential scale, fading or lightening a colour to show uncertainty changes the value a reader perceives, so uncertainty is better encoded in a separate channel such as texture or outline.
Based on: Follows from lightness being the value channel in sequential scales, as described by ColorBrewer and Crameri et al.; Colourwise's inference rather than a cited experiment.
Caveat: Carefully designed bivariate value–uncertainty legends can make fading interpretable.
Reviewed 29 September 2026. Colourwise summarises its sources in its own words and does not reproduce standards text or proprietary colour data. Spotted an error? Tell us.