Visualization Islands: Mapping the Reachability of Data Online During the Covid-19 Pandemic

In Progress

Overview

We introduce "visualization islands," a term describing how technical barriers — slow-loading webpages and accessibility errors — can isolate data visualizations from their audiences. Even a carefully designed chart can fail to reach people if the page hosting it is slow, paywalled, or inaccessible to screen readers. We analyzed more than 3,000 webpages hosting COVID-19 visualizations to map where these barriers show up most, and who they leave out.

Approach

We based our analysis on the COVIC (COVID-19 Online Visualization Collection) archive, drawing on over 3,000 visualizations published during the pandemic. For each webpage, we measured:

Accessibility, using WebAIM's WAVE tool to check for WCAG violations like missing alt text, low contrast, and improper ARIA attributes
Performance, using WebPageTest to simulate load times under three connectivity conditions — 3G, DSL, and fiber

This let us look at reachability holistically, rather than evaluating a visualization's design in isolation from the environment hosting it.

[Picture]

Findings

Reachability turned out to be highly uneven across the web:

News media sites were the slowest to load overall, often due to paywalls. In one case, a NYTimes COVID dashboard took over a minute to load on 3G, only for a paywall to block access before it finished.
Social media posts loaded fastest across all connection speeds, but were riddled with accessibility errors, making them difficult or impossible for many blind and visually impaired users to use.
Government, NGO, and peer-reviewed sources were the most consistently reachable, balancing reasonable load times with better accessibility, though even these weren't error-free.
Overall, more than 72% of webpages failed on at least one front, either load time or accessibility, meaning "fully reachable" visualizations were the exception, not the norm.

Collaborators

This work was done in collaboration with Prof Jonathan Zong (University of Colorado) and Prof Evan Peck (University of Colorado).

[Gallery picture]