Discover more from data + feelings
An Opinionated Guide To Alt-text
Personal preferences on one aspect of data visualization accessibility, plus updates from the past few months.
This March I had the amazing opportunity to talk about data visualization accessibility at NICAR alongside some industry heavyweights: Frank Elavsky, Patrick Garvin, Thomas Wilburn and Joe Murphy. We all gave short presentations on different topics, and I spoke about my particular passion: data visualization alt-text.
I first became invested in proper alt-text when I joined a slack group for disabled people in tech six or seven years ago. Many members expressed frustration with a lack of or woefully inadequate alt-text. As someone in graphic communication, that was something I could take action on!
There has been a lot of excellent work towards accessible data visualization over the past few years. For alt-text specifically, Amy Cesal published a thorough introductory guide and Patrick Garvin has an article geared towards journalists on Source. The Urban Institute published a Do No Harm guide (pdf) centered on accessibility in data visualization, and the section on alt-text is particularly useful. In addition to people already mentioned, Doug Schepers and Sarah Fossheim consistently share in-depth information about accessible visualization.
I have been working with alt-text long enough to have gathered my own opinions on the craft. It’s a constant evolution based on the most current research, gathering insights from industry colleagues and my own personal style.
The basics of alt-text are covered elsewhere pretty extensively. Below are the tips I shared in my NICAR presentation. In other words, here are my feelings about data.
Have a “data experiences” mindset. The alt-text is not meant to encompass everything in the chart. I find Frank Elavsky’s language of “data experiences” very helpful here: people will have different experiences of the data based on a variety of factors. Alt-text creates one experience. I focus on creating an excellent alt-text experience, not doing a one-to-one translation of my chart into text.
Be verbose. Do not be afraid to include specific numbers in the alt-text. Sometimes I’ll write 3-5 sentences of alt-text for a chart. There isn’t a character limit, so use as many words as you need in order to communicate the important parts of the graph.
That being said: Be judicious. When writing alt-text, cognitive load is top of mind. I can write out every number on a bar chart, but how much information can someone hold in their head before they check out? (There is a lot of very technical research on this, but I adhere more to a principle than a hard limit on information quantity.) I explain the data point relevant to a story first. For example, when in a story about Black Women’s Equal Pay Day, I always write out the data points for Black women in the charts. I also include their placement among other data points (example detailed below) and give maybe one relevant min or max data point to give a better sense of where the relevant data point fits within a set. I do not attempt to include all the data.
Don’t be afraid to use words that are not present in the visualization. With data visualization, we are trained to be succinct. So many editorial choices are made in the choice of hed, dek, labeling, colors and chart type. It can feel weird to be verbose when so often a good chart will speak for itself. One thing I had to get used to expanding my visualization vocabulary in service of a better description.
Think (and write) in patterns. Note any significant information only illustrated visually: huge drops, large disparities, a lack of differentiation. Calling out a pattern isn’t necessarily editorialization (something which should be avoided), but rather a plain-language explanation of trends. I employ this most often when it comes to line charts or scatter plots. Is there an S-curve? A sudden drop at a certain date? Is there minimal change over time? Are data points clustered around certain areas? All of that can be explained in the alt-text. I try to stick to one big shape-based takeaway with line charts, sometimes followed by an additional point of interest. In that case, I include the relevant Y-axis information (e.g. “There is a large drop between January and February”).
Know how screenreaders work, preferably by using them. I use Datawrapper a lot to create charts, and it’s useful for me to know that the hed, dek and note are read as separate elements from the actual graph. When I construct alt-text, I do not repeat text that I know will already be read by the screen reader. First off, that’s annoying. Second, it can create confusion — I don’t want someone to be focused on parsing what is actually new information when hearing a chart description.
Utilize shortcuts. Use voice-to-text if you are intimidated by the number of digits you need to type out.
“Basically, what this chart is doing…” If you’re struggling to be more verbose in your alt-text descriptions, try this framing. Go ahead and explain why this chart is in the article, and what function it is serving for readers. That can be a quick way to determine the most important information to include in the alt-text.
Here is an example of alt-text I wrote for a recent story.
Bar chart showing that Black men and women have the largest amount of federal loans 12 months after completing a bachelor’s degree. Black women have the most debt, at $38,800 on average; Black men have $35,997 on average. The next highest is Native American women, with $32,619. Asian women have the least debt on average, at $20,417.
Let’s break it down.
I started by identifying the chart type and what data is being shown. I highlighted Black women, the focus of the story, and who were the group with the highest amount of debt. I also shared the average debt for Black men, to show the difference between gender in the racial category. I shared the group with the next-highest amount of debt, Native American women, to better qualify how much more debt Black women had on average. I rounded it out with information by calling out the group with the least amount of debt, in order to showcase the entire range of the chart.
I could have stopped after the first sentence: Bar chart showing that Black men and women have the largest amount of federal loans 12 months after completing a bachelor’s degree. But I wanted to emphasize the dollar amounts involved, so a reader had a better sense of how burdensome the loans could be.
I did include four numbers in the alt-text, which is just about the limit of what I’m willing to include. I tried to include the numbers which would provide the most context to the chart.
Here is the chart.
What do you think? Comment or reply with your critique of my alt-text, and any other tips you have to share.
For Disability Pride Month, I wrote a story about disability doulas: disabled people who help shepherd newly disabled people into a new identity. This piece is near and dear to my heart, as explained in a column I wrote for The 19th’s newsletter.
After hearing an interview with one of the authors on NPR, I checked out “Easy Money: Cryptocurrency, Casino Capitalism, and the Golden Age of Fraud” from my library. It has the most detailed and specific explanations of cryptocurrency, bitcoin and blockchain I have ever encountered. The audiobook is lots of fun, if a bit self-aggrandizing. (It's narrated by actor Ben McKensie from “The O.C.” who also wrote it.) Admittedly, my loan is going to expire before I finish it.
I have read 41 books since my last newsletter in mid-April, bringing me to a total of 73 for the year. I generally only have a single goal for the year (total number of books read), but I’m toying with setting a mini-goal related to format or genre for the rest of 2023.
With Kae Petrin of Chalkbeat I will be presenting on how to do good work using bad LGBTQ+ data at SRCCON. I love this little conference; if you have never been please consider attending in October. Let me know if you will be there!
I am on Bluesky. 😔 I’m still trying to figure out my relationship to social media In These Times.
Thanks for reading, friends! Make sure to reply or comment with your thoughts about alt-text. I’m looking forward to reading them.
Thanks for reading data + feelings! Subscribe for free to receive new posts and support my work.