Accessible charts: colour, contrast and screen readers

TL;DR

Difficulty: Intermediate

A chart that only works for someone with full colour vision on a bright laptop fails a good share of its readers.

  • About 1 in 12 men and 1 in 200 women have colour vision deficiency, so never let colour carry the meaning alone: add a dash, a marker or a direct label (WCAG 1.4.1).
  • Measure contrast: lines, bars and slices need 3:1, text 4.5:1, and check it again in dark mode.
  • Label directly instead of using a legend, and prefer sorted bars to a pie with many slices.
  • Give every chart a description with what it is, the point and the key numbers, plus the data as a table (WCAG 1.1.1).
  • Make it work on a phone, and call the result built to WCAG 2.1 AA, never guaranteed compliant.

You made a clean chart with one clear message. Now picture who reads it: a board member who cannot tell red from green, a buyer with a screen reader, a finance lead in dark mode, the owner on a phone. Here is how to make one chart work for all of them, step by step, with the maths behind contrast, a full audit of a report chart, and what the 2023 Accessibility Regulations mean for the charts on an Irish shop.

The question: can everyone read this chart?

You made a chart. It is clean, it has one message, and the title states the finding. Now picture who reads it. The owner on a phone in the stockroom. A board member who never told anyone that red and green look much the same to him. A buyer who uses a screen reader, so she hears your chart rather than sees it. A finance lead who reads everything in dark mode after 9pm.

If the chart only works for a person with full colour vision on a bright laptop screen, it fails a good share of the people it was made for. This article is about closing that gap, one rule at a time. Every rule comes with a worked example from Shannon Outfitters, the invented outdoor gear brand in Limerick that runs through this series, and its analyst, Sam.

None of this is decoration. Accessibility for charts comes down to a few measurable things: how colour is used, how much contrast there is, what a screen reader hears, and how the chart behaves on a small screen. The standard behind all of it is the Web Content Accessibility Guidelines (WCAG) 2.1, published by the W3C1. If you want the whole standard explained first, read WCAG 2.1 AA explained. This article stays with charts.

Colour vision deficiency is common, so plan for it

Colour vision deficiency (often called colour blindness) means a person sees some colours differently or cannot tell certain colours apart. Most often the trouble is with red and green. It is far more common than most people assume. Colour Blind Awareness, a charity based in Britain, puts it at about 1 in 12 men (8%) and 1 in 200 women (0.5%)2. In a group with as many women as men, that averages out at about 4.25%.

Bar chart of the share of people with colour vision deficiency, from Colour Blind Awareness. Men 8%, women 0.5%, and about 4.25% on average in a group with as many women as men. The point: in any group of men, one in twelve is likely to have trouble telling some colours apart. Tap or click to open full screen and move around

Those numbers matter most in small groups, because a small group feels safe. "Nobody on my board is colour blind" is a guess, not a fact. Probability turns that guess into a number.

If each person in a group has a chance pp of having colour vision deficiency, and people are independent of one another, the chance that at least one of them has it is:

P(at least one)=1−∏i=1n(1−pi)P(\text{at least one}) = 1 - \prod_{i=1}^{n} (1 - p_i)

where P(at least one)P(\text{at least one}) is the probability that one or more people in the group have colour vision deficiency, nn is the number of people in the group, pip_i is the chance for person ii (here 0.080.08 for a man and 0.0050.005 for a woman), and ∏\prod means "multiply all of these together". The logic is simple: first work out the chance that nobody has it, then take that away from 1. Foundations: probability explains the "at least one" trick step by step.

Example: Sam's board of eight

Shannon Outfitters has a board of eight: six men and two women. Sam wants to know how likely it is that at least one of them cannot tell a red line from a green one.

Step 1, the chance that a single man does not have it: 1−0.08=0.921 - 0.08 = 0.92.

Step 2, the chance that none of the six men has it: 0.926=0.60640.92^6 = 0.6064.

Step 3, the chance that a single woman does not have it: 1−0.005=0.9951 - 0.005 = 0.995, and for both women 0.9952=0.99000.995^2 = 0.9900.

Step 4, the chance that nobody on the board has it: 0.6064×0.9900=0.60030.6064 \times 0.9900 = 0.6003.

Step 5, the chance that at least one person has it:

1−0.6003=0.39971 - 0.6003 = 0.3997

where 0.60030.6003 is the chance from step 4 that nobody has it and 0.39970.3997, about 40%, is the chance that somebody does.

So there is roughly a two in five chance that someone at Sam's board table reads his charts with colour vision deficiency, and that person may never have said so. The newsletter tells the same story at a larger scale. Shannon's list has 6,000 subscribers. If about half of them are men, the average is (8+0.5)/2=4.25(8 + 0.5) / 2 = 4.25%, so roughly 255 of them may struggle with a chart that leans on red and green.

Never let colour carry the meaning on its own

This is the first rule and the one broken most often. WCAG success criterion 1.4.1, Use of Color, says colour must not be the only visual means of conveying information3. You may still use colour. You just need a second cue next to it: a pattern, a dash style, a marker shape, a direct text label. The W3C's own examples make the point with a sentence like "Mary's sales are in red, Tom's are in blue", which tells you nothing if you cannot see red and blue apart3. The W3C even lists a chart that relies on colour as a documented failure of the criterion4.

There is a useful escape hatch in the same guidance. A difference in colour also counts as a second cue if the two colours differ in lightness by a contrast ratio of at least 3:13. Lightness survives colour vision deficiency, greyscale printing and a dim projector. Hue often does not.

Example: Sam's channel chart, before and after

Sam's monthly report has a line chart of revenue from the web shop and from the shop on O'Connell Street in Limerick. In the original, online is a red line and the shop is a green line, with a legend in the corner. Nothing else tells the lines apart. Here is what that looks like, drawn in this site's own colours:

Line chart of Shannon Outfitters' invented monthly revenue in thousands of pounds, January to December, for online and the shop. The two lines differ only in colour, with no dash, no marker and no direct label, so a reader who cannot tell the two colours apart cannot tell which line is which, or that online overtook the shop in March. Tap or click to open full screen and move around

A reader who cannot separate the two colours sees two lines that cross in March and has to guess which is which. The legend does not help, because the legend uses the same two colours.

The fix takes three small changes. Draw the shop as a dashed line. Put the finding in the title. Say in the title which line is which, so the reader never needs to match a colour to a legend.

Line chart of Shannon Outfitters' invented monthly revenue in thousands of pounds. Online is a solid line, the shop is a dashed line. Online starts at €38k in January, below the shop at €44k, overtakes it in March at €45k against €44k, and ends December at €92k against €70k. The point: online has led every month since March and the gap widened in the Christmas quarter. Tap or click to open full screen and move around

Now the chart works in colour, in greyscale and for someone with red green colour vision deficiency. The series names carry "solid" and "dashed" too, so even the legend and the tooltip say which is which without colour.

Contrast you can measure: 3:1 for every line, bar and slice

Colour is a feeling, but contrast is a number. WCAG defines it precisely, and two success criteria matter for charts. Text, including axis labels and data labels, needs a contrast ratio of at least 4.5:1 against its background, or 3:1 for large text (criterion 1.4.3)5. The parts of a graphic you need in order to understand it, such as lines, bars and pie slices, need at least 3:1 against the colours next to them (criterion 1.4.11, Non text Contrast)4. The W3C is strict about the threshold: 2.999:1 fails4. Its own examples include a pie chart whose slices fail because neighbouring slices are too close in lightness, even though each slice carries a label4.

The ratio comes from a colour's relative luminance, how bright it looks on a scale from 0 for black to 1 for white. WCAG gives the formula for screen colours2:

L=0.2126 R+0.7152 G+0.0722 BL = 0.2126\,R + 0.7152\,G + 0.0722\,B

where LL is the relative luminance, and RR, GG and BB are the red, green and blue channels after two steps. First divide the 0 to 255 value by 255, giving cc. Then, if c≤0.04045c \le 0.04045, use c/12.92c / 12.92, otherwise use (c+0.0551.055)2.4\left(\frac{c + 0.055}{1.055}\right)^{2.4}. The weights 0.21260.2126, 0.71520.7152 and 0.07220.0722 reflect how much each channel adds to the brightness the eye sees: green counts most, blue least. The weighting is the same idea as a weighted average, covered in Foundations: mean.

The contrast ratio between two colours is then:

contrast ratio=L1+0.05L2+0.05\text{contrast ratio} = \frac{L_1 + 0.05}{L_2 + 0.05}

where L1L_1 is the relative luminance of the lighter colour, L2L_2 that of the darker one, and the 0.050.05 stands in for the glare and reflection on a real screen. The ratio runs from 1:1 (no difference) to 21:1 (black on white). It is a ratio of two brightness values, so Foundations: ratios and rates covers the idea behind it.

Example: is Sam's green line visible enough?

Sam's original green is #6aa84f, drawn on a white background. Here is the whole calculation.

Step 1, convert the hex code to its three channels: red 106, green 168, blue 79.

Step 2, divide each by 255: 106/255=0.4157106/255 = 0.4157, 168/255=0.6588168/255 = 0.6588, 79/255=0.309879/255 = 0.3098.

Step 3, linearise each value. All three are above 0.04045, so each uses the second formula: red becomes 0.14410.1441, green 0.39160.3916, blue 0.07820.0782.

Step 4, weight and add:

L=0.2126×0.1441+0.7152×0.3916+0.0722×0.0782=0.3163L = 0.2126 \times 0.1441 + 0.7152 \times 0.3916 + 0.0722 \times 0.0782 = 0.3163

where the three products are 0.03060.0306, 0.28010.2801 and 0.00560.0056, and their sum L=0.3163L = 0.3163 is the relative luminance of Sam's green.

Step 5, compare with white, whose relative luminance is exactly 1:

contrast ratio=1+0.050.3163+0.05=2.87\text{contrast ratio} = \frac{1 + 0.05}{0.3163 + 0.05} = 2.87

where 11 is the luminance of white, 0.31630.3163 that of the green, and 2.872.87 the result. That is below 3:1, so the green line fails criterion 1.4.11 on white. The red line, #e06666, has a relative luminance of 0.26310.2631 and reaches 3.35:1, a pass.

Step 6, compare the red and the green with each other: (0.3163+0.05)/(0.2631+0.05)=1.17(0.3163 + 0.05) / (0.2631 + 0.05) = 1.17. The two lines are almost the same lightness. For someone who cannot see the hue, they are close to the same line. That is the numerical reason the chart in the previous section failed, and why the dash style was the right fix rather than "pick a brighter red".

The analysis guidance published for the UK's government statisticians draws the same conclusions, and it travels across the border without any trouble: keep chart colours at 3:1 against the background and against neighbouring colours, check that the chart still works in greyscale, and avoid line charts with more than two series where possible, because keeping every line at 3:1 against every other becomes impossible6.

Dark mode needs its own check

A colour that passes on white can fail on a dark background, and the other way round. Contrast is always measured against the actual background, so a chart that switches to dark mode with the operating system (the CSS media feature prefers-color-scheme tells the page which mode the reader chose7) needs its colours checked twice.

Example: Sam's palette on white and on near black

Sam tries a corporate navy, #1f4e79, for the online line. On white it is superb. On the near black background of the dark mode report, #18181b, it nearly vanishes. Sam runs the same calculation as above for five colours on both backgrounds:

Grouped bar chart of contrast ratios, calculated with the WCAG formula, for five colours on white and on near black (#18181b). Red #e06666: 3.35 on white, 5.28 on near black. Green #6aa84f: 2.87 and 6.18. Orange #c55a11: 4.33 and 4.09. Dark green #2e7d32: 5.13 and 3.46. Navy #1f4e79: 8.66 and 2.05. The threshold for lines and bars is 3:1. Only orange and dark green clear it on both backgrounds. Tap or click to open full screen and move around

The pattern is plain once it is drawn. Dark colours shine on white and sink on black. Light colours do the opposite. Orange #c55a11 and dark green #2e7d32 are the only two that clear 3:1 on both backgrounds, so Sam takes them for the two channel lines, keeps the dash style, and the chart now passes in either mode without a second palette. When no single colour works in both modes, the answer is a second palette for dark mode, chosen and checked separately, never the light palette reused on a dark background.

Direct labels beat legends

A legend asks the reader to do a matching job: look at a colour in the chart, find the same colour in the legend, read the name, look back. That is hard work for everyone and impossible for someone who cannot tell the colours apart. A direct label puts the name right on the line, the bar or the slice. That same government analysis guidance recommends exactly that for lines and pie slices, and adds a related rule for the text around a chart: never refer to "the blue bar", because the reader may not see blue as you do6.

Example: Shannon's product mix as a pie and as bars

Sam's report shows the share of revenue by product category. The first draft is a seven slice pie with a legend. With seven slices, the renderer has to reach for lighter tints of the same few colours, and several slices end up close in lightness:

Pie chart of Shannon Outfitters' invented revenue share by product category: waterproof jackets 31%, boots 22%, rucksacks 15%, tents 12%, base layers 9%, socks 6%, accessories 5%. The slices are told apart only by colour and a legend, and several tints are close in lightness, so the smaller slices are hard to match to their names and hard to compare. Tap or click to open full screen and move around

Now the same numbers as sorted bars. One colour for all of them, because the categories have no meaningful grouping, and the name of each category sits directly under its bar:

Bar chart of Shannon Outfitters' invented revenue share by product category, sorted from largest to smallest, one colour, each bar labelled with its category and value: waterproof jackets 31%, boots 22%, rucksacks 15%, tents 12%, base layers 9%, socks 6%, accessories 5%. The point: jackets and boots together are 53% of revenue. Tap or click to open full screen and move around

The bar chart needs no colour matching at all. Every bar has its name and its value printed next to it, the order does the ranking for you, and the title says what to take away. Nothing here depends on telling one colour from another.

Patterns, dash styles and markers as the second cue

When you do need several series on one chart, give each one a cue that is not colour. On a line chart that means a dash style (solid, dashed, dotted) or a marker shape (circle, square, triangle) at each data point. On a bar chart it means a pattern fill such as stripes or dots, or simply a direct label. The W3C lists pairing colour with a pattern as a sufficient technique for exactly this8. Keep it to two or three styles. Six dash patterns on one chart are as hard to read as six colours.

Example: actual versus forecast

Sam's sales forecast chart has two series: actual sales up to October and the forecast from November. The convention he follows is solid for actual and dashed for forecast. That one choice does a lot of work. A reader with full colour vision, a reader with colour vision deficiency and a reader who printed the report in black and white all see where the facts end and the estimate begins. The chart title then says it in words: "Sales: actual to October (solid), forecast from November (dashed)". For how to draw the forecast itself honestly, with its uncertainty, see communicating forecasts.

A text alternative that says what, the point and the key numbers

A screen reader reads a page aloud, or sends it to a Braille display. It cannot look at a chart. What it reads is the text alternative, and WCAG criterion 1.1.1 requires one for all non text content, serving the equivalent purpose9. For a chart, "equivalent purpose" means the reader who hears it should come away with what the sighted reader gets from looking.

This matters in practice. In a study at the University of Washington, Sharif and colleagues compared 36 screen reader users with 36 people who do not use one. The screen reader users extracted information from online charts 61.48% less accurately and spent 210.96% more time doing it, and the authors found that many charts were not even discoverable to a screen reader10. The participants themselves suggested tables and text descriptions as the fix10.

Many web charts are drawn on an HTML canvas, which is a picture as far as a screen reader is concerned. The Chart.js documentation says this plainly: canvas content is not accessible by default, and you add accessibility yourself, with role="img" and an aria-label, or with meaningful fallback content inside the canvas element11. The W3C tutorial on complex images recommends a two part approach for charts: a short text alternative that names the chart, and a longer description that carries the trends and the values12.

What should the description say? Lundgard and Satyanarayan studied chart descriptions with 30 blind and 90 sighted readers and sorted what a description can contain into four levels: how the chart is built (chart type, axes, encodings), statistics (highest, lowest, averages), perceived patterns and trends, and context or meaning in the domain. Blind and sighted readers differed significantly in which levels they found most useful13. My practical reading of that is a simple order for every description:

  • What: the chart type and what it measures, with the time period.
  • The point: the one finding the chart exists to show.
  • The key numbers: the values that prove the point, start and end, highest and lowest.

Every chart in this article follows that order in its description, and so does every chart on my own site: each one carries a text description that a screen reader reads out, alongside the brand colours, light and dark mode, labels at 45 degrees and a full screen view.

Example: the description of Sam's channel chart

Sam's original alternative text was a single word: "Chart". A screen reader user hears "Chart, image" and nothing else.

His first rewrite was "Line chart of revenue by channel". That is grand as far as it goes, but it is only the first level: it describes the chart, not what it shows.

The final version follows the three steps:

Line chart of Shannon Outfitters' monthly revenue by channel, January to December. Online overtook the shop in March and led every month after. Online rose from €38k in January to €92k in December, the shop from €44k to €70k.

That is three sentences, under 50 words, and a listener now knows what the sighted reader knows: what the chart is, what it says, and the numbers behind it. Notice what is missing too. There is no "the red line" and no "as you can see". The description never depends on colour or on looking.

Offer the data as a table

A description carries the point. A table carries everything else. Some readers want to check a number for a particular month, and a screen reader can move through a proper HTML table row by row and column by column, announcing the headers as it goes. That is why the screen reader users in the Washington study asked for tables10, and why the W3C tutorial places a table of values inside the long description12. A table helps sighted readers too: anyone who wants the exact figure for October no longer has to estimate it from a line.

Example: the table under Sam's chart

Under the channel chart, Sam adds a toggle labelled "Show the data as a table". It opens this (the numbers are invented):

QuarterOnline (€k)Shop (€k)Online share
Q1 (Jan to Mar)12313148%
Q2 (Apr to Jun)14913053%
Q3 (Jul to Sep)17313556%
Q4 (Oct to Dec)23617557%
Year68157154%

He shows quarters rather than all 12 months, because the table's job is to support the chart's point, not to repeat every value. Each online share is online divided by the total for that row: for Q1, 123/(123+131)=0.484123 / (123 + 131) = 0.484, so 48%. The full monthly figures go in a downloadable file for anyone who wants them. The table has real column headers, so a screen reader announces "Online, €k, 149" rather than a bare "149".

Phones: orientation, reflow and a full screen view

A good share of your readers will open the report on a phone. Two WCAG criteria matter here. Criterion 1.3.4, Orientation, says content must not lock itself to portrait or landscape unless a particular orientation is essential14. Criterion 1.4.10, Reflow, says content must work at a width equivalent to 320 CSS pixels without scrolling in two directions15. A wide chart squeezed into 320 pixels is the classic failure: twelve month labels collide, the legend covers the lines, and the reader gives up.

Three habits solve most of it. Rotate long category labels to 45 degrees so they stop overlapping. Offer a full screen view, so a reader can turn the phone sideways and see the whole chart large. And draw process diagrams top to bottom rather than left to right, because a phone screen is tall, not wide. Those are the rules every chart and diagram on my site follows.

Example: Sam's twelve month chart on a phone

Sam opens his channel chart on a phone 360 pixels wide. The twelve month labels, written out as "January" to "December", collide into an unreadable smear. He makes three changes: short month names ("Jan" to "Dec"), labels at 45 degrees, and a full screen button under the chart. In portrait the chart is now readable, and in full screen view it fills the screen in landscape. The audit below also has a top to bottom flow chart, the shape that stays readable at phone width.

The law: the 2023 accessibility rules and the Equal Status Act

In Ireland, the European Accessibility Act is law through the European Union (Accessibility Requirements of Products and Services) Regulations 2023, S.I. No. 636 of 2023, which came into operation on 28 June 202516. The Act behind them is a directive that names eCommerce services17. For eCommerce services, the market surveillance authority is the Competition and Consumer Protection Commission (CCPC)18. So an online shop selling to consumers in Ireland is squarely in scope, and so are the charts on its pages.

Underneath that sits the older Equal Status Act 2000. A service provider has to do all that is reasonable to accommodate a disabled person, by providing special treatment or facilities where without them it would be impossible or unduly difficult to use the service, and a refusal is only reasonable if it would cost more than a nominal cost19. Readable charts are about the lowest cost accommodation there is.

The technical benchmark for the Regulations is the European standard EN 301 549, which builds on WCAG 2.1 at level AA for web content20. Charts on a shop's pages, a size guide, a product comparison or a delivery cost chart, are part of that content.

So the honest claim for a chart, a page or a whole site is "built to WCAG 2.1 AA, the standard the Accessibility Act relies on". Never "guaranteed compliant": compliance is a legal judgement about a whole service, not a property of one chart. For what the Act asks of a shop or a SaaS product in detail, read accessibility law for online businesses.

Example: which of Sam's charts are covered?

Sam lists his charts in three groups. The delivery cost chart on the public checkout page serves consumers in Ireland and across the EU, so it falls under the 2023 Regulations, with the CCPC as the body that can come knocking. The board report is internal, so the Regulations do not reach it, but if a board member or a colleague uses a screen reader, an unreadable report is simply a bad report. The marketing dashboard Sam uses alone has no legal exposure today, but the next marketing hire might use a screen reader, and building the habit now costs nothing extra. His conclusion: build every chart to the same standard, because the work is the same and the risk of sorting them wrong is not.

Worked example: auditing one of Sam's charts, step by step

Here is the whole process in one place, applied to the channel chart from the start of this article. The flow chart is the order Sam follows, and the steps under it are what he found.

Tap or click to open full screen and move around

Example: the channel chart audit

Step 1, the decision. The chart supports one decision: whether to shift part of the shop's window budget to online marketing. So the point it must make is "online overtook the shop in March and the gap is widening".

Step 2, colour alone. The two lines were red and green with a legend. Colour was the only cue. Fix: the shop line becomes dashed, the series names say "solid" and "dashed", and the title names both.

Step 3, contrast. Green #6aa84f on white was 2.87:1, a fail. Red against green was 1.17:1, nearly identical in lightness. Fix: orange #c55a11 for online (4.33:1 on white, 4.09:1 on near black) and dark green #2e7d32 for the shop (5.13:1 and 3.46:1). The axis labels were a pale grey; Sam checks them against 4.5:1 too and darkens them.

Step 4, the description. Before: "Chart". After: the three sentence description from the section above, with what, the point and the key numbers.

Step 5, the table. Sam adds the quarterly table with real column headers, plus a downloadable file of the monthly figures.

Step 6, the phone. The month names are shortened to three letters and rotated to 45 degrees, and the chart opens in a full screen view. At 360 pixels wide it is readable in portrait.

Step 7, the wording. Sam's report footer now says the charts are "built to WCAG 2.1 AA". It does not say "fully accessible" or "compliant", because those are claims about the whole service that one analyst cannot make on his own.

The before and after descriptions, side by side, show how much a screen reader user gained:

  • Before: "Chart."
  • After: "Line chart of Shannon Outfitters' monthly revenue by channel, January to December. Online overtook the shop in March and led every month after. Online rose from €38k in January to €92k in December, the shop from €44k to €70k."

None of the seven steps took Sam more than a few minutes. Together they turned a chart that failed a good share of its readers into one that works for all of them, and, as a side effect, a chart that is clearer for everyone else as well.

Conclusion: build the chart for every reader at once

Accessible charts are not a separate product for a small audience. They are the same good chart, finished properly. Never let colour carry the meaning alone. Measure contrast instead of trusting your eye, and measure it again for dark mode. Label directly, use dash styles and markers as a second cue, write a description that says what, the point and the key numbers, and put the data in a table beside it. Then check the whole thing on a phone.

Every one of those steps also helps the readers who need no help at all: the board member skimming on a train, the finance lead in dark mode, the owner who prints the report and reads it over a coffee. If you want the foundations of chart design first, start with why a chart, and when a table is better. And if you want to start from the decision a chart supports, as I do on every analytics project, read start with the decision, not the dashboard.

If your reports and dashboards should work for every reader, colour blind, screen reader or phone, book the free workshop and we go through your charts together, no bother. The first two weeks of work are on me.

Sources

  1. 1.

    W3C, Web Content Accessibility Guidelines (WCAG) 2.1, W3C Recommendation, 5 June 2018, updated 6 May 2025, definitions of relative luminance and contrast ratio, https://www.w3.org/TR/WCAG21/

  2. 2.

    Colour Blind Awareness, About colour blindness, colourblindawareness.org, https://www.colourblindawareness.org/colour-blindness/

  3. 3.

    W3C Web Accessibility Initiative, Understanding Success Criterion 1.4.1: Use of Color, W3C, https://www.w3.org/WAI/WCAG21/Understanding/use-of-color.html

  4. 4.

    W3C Web Accessibility Initiative, Understanding Success Criterion 1.4.11: Non-text Contrast, W3C, https://www.w3.org/WAI/WCAG21/Understanding/non-text-contrast.html ; and Failure F13, W3C, https://www.w3.org/WAI/WCAG21/Techniques/failures/F13

  5. 5.

    W3C Web Accessibility Initiative, Understanding Success Criterion 1.4.3: Contrast (Minimum), W3C, https://www.w3.org/WAI/WCAG21/Understanding/contrast-minimum.html

  6. 6.

    Government Analysis Function (Analysis Function Central Team, Office for National Statistics), Data visualisation: colours, first published 23 November 2021, updated 12 February 2026, https://analysisfunction.civilservice.gov.uk/policy-store/data-visualisation-colours-in-charts/

  7. 7.
  8. 8.

    W3C Web Accessibility Initiative, Technique G111: Using Color and Pattern, W3C, https://www.w3.org/WAI/WCAG21/Techniques/general/G111

  9. 9.

    W3C Web Accessibility Initiative, Understanding Success Criterion 1.1.1: Non-text Content, W3C, https://www.w3.org/WAI/WCAG21/Understanding/non-text-content.html

  10. 10.

    Ather Sharif, Sanjana S. Chintalapati, Jacob O. Wobbrock and Katharina Reinecke, Understanding Screen-Reader Users' Experiences with Online Data Visualizations, ASSETS '21, ACM, 2021, https://faculty.washington.edu/wobbrock/pubs/assets-21.01.pdf

  11. 11.

    Chart.js, Accessibility, Chart.js documentation, https://www.chartjs.org/docs/latest/general/accessibility.html

  12. 12.

    W3C Web Accessibility Initiative, Complex Images, Images Tutorial, W3C, https://www.w3.org/WAI/tutorials/images/complex/

  13. 13.

    Alan Lundgard and Arvind Satyanarayan, Accessible Visualization via Natural Language Descriptions: A Four-Level Model of Semantic Content, IEEE Transactions on Visualization and Computer Graphics (Proceedings of IEEE VIS), 2022, arXiv 2110.04406, https://arxiv.org/abs/2110.04406

  14. 14.

    W3C Web Accessibility Initiative, Understanding Success Criterion 1.3.4: Orientation, W3C, https://www.w3.org/WAI/WCAG21/Understanding/orientation.html

  15. 15.

    W3C Web Accessibility Initiative, Understanding Success Criterion 1.4.10: Reflow, W3C, https://www.w3.org/WAI/WCAG21/Understanding/reflow.html

  16. 16.

    Minister for Children, Equality, Disability, Integration and Youth, S.I. No. 636/2023, European Union (Accessibility Requirements of Products and Services) Regulations 2023, regulation 1, electronic Irish Statute Book, 2023, https://www.irishstatutebook.ie/eli/2023/si/636/made/en/print

  17. 17.

    European Parliament and Council, Directive (EU) 2019/882 of 17 April 2019 on the accessibility requirements for products and services (European Accessibility Act), articles 2 and 31, EUR Lex, 2019, https://eur-lex.europa.eu/legal-content/EN/TXT/HTML/?uri=CELEX:32019L0882

  18. 18.

    Competition and Consumer Protection Commission, CCPC role and responsibilities, Accessibility for businesses, ccpc.ie, https://www.ccpc.ie/enforcement-and-regulation/market-surveillance/accessibility/accessibility-for-businesses/ccpc-role-and-responsibilities

  19. 19.

    Oireachtas, Equal Status Act 2000, section 4, Discrimination on ground of disability, electronic Irish Statute Book, https://www.irishstatutebook.ie/eli/2000/act/8/section/4/enacted/en/html

  20. 20.

    CEN, CENELEC and ETSI, EN 301 549 V3.2.1 (2021 03), Harmonised European Standard, Accessibility requirements for ICT products and services, clause 9.0, ETSI, 2021, https://www.etsi.org/deliver/etsi_en/301500_301599/301549/03.02.01_60/en_301549v030201p.pdf