Optimus Web: Selective Delivery of Desktop or Mobile Web Pages
Nádia Fernandes, Tiago Guerreiro, Diogo Marques, Luís Carriço · 2015 · Proceedings of the 12th International Web for All Conference (W4A) · doi:10.1145/2745555.2746668
Summary
This paper presents Optimus Web, a client-side web proxy system that automatically selects and delivers the most accessible version of a website — either its desktop or mobile representation — to blind users. The system uses QualWeb, an automated accessibility evaluator that assesses pages against WCAG 2.0 after browser interpretation (including JavaScript execution), to determine which version has fewer accessibility barriers. The researchers conducted a user study with 19 blind participants using JAWS screen reader, who performed navigation, identification, selection, and search tasks across three websites (a newspaper, a recipe catalogue, and an e-commerce site) in both desktop and mobile versions. The study used a within-subjects design with counterbalanced ordering. The key premise challenges the assumption that mobile web pages are inherently more accessible — while mobile versions tend to be simpler and more structured, the researchers' prior work showed this is not universally true, motivating a selective approach rather than always forcing mobile delivery.
Key findings
Of the 19 webpages evaluated, only 63% were more accessible in their mobile version according to QualWeb, confirming that mobile is not always more accessible. The e-commerce site showed statistically significant performance improvements with the mobile version (Wilcoxon test, Z=-2.54, p=0.011), while the recipes site actually performed better in its desktop version due to search functionality issues in the mobile version. Expert screen reader users were 20-65% more efficient than non-experts and showed no correlation between the number of accessibility barriers and their performance — they effectively circumvented problems using strategies like element lists, virtual search, and keyboard shortcuts. Medium-level users performed worse than both experts and basic users in some cases, because they knew advanced strategies but applied them incorrectly, causing extra navigation overhead. Basic users encountered approximately 70% of accessibility barriers by navigating through nearly all page elements sequentially. Common barriers included irrelevant content before task content, unclear link destinations, violated heading structures, and missing skip links — with mobile versions reducing most of these but sometimes introducing content comprehension issues.
Relevance
This research offers a practical, non-invasive approach to improving web accessibility without requiring website owners to change their code. The selective delivery concept — choosing the best existing representation rather than transforming content — avoids the deep structural knowledge that transcoding approaches require. For accessibility practitioners, the finding that expert users perform independently of barrier counts highlights both the value of screen reader proficiency training and the risk of using only expert evaluators in user testing, since their coping strategies mask real barriers that novice users face. The observation that medium-level users sometimes perform worse than beginners when misapplying advanced navigation strategies has implications for assistive technology training programs. While responsive design has largely replaced separate mobile sites since 2015, the core insight — that automated evaluation can inform content delivery decisions to improve accessibility — remains relevant for personalized accessibility solutions.
Tags: web accessibility · blind users · screen readers · mobile accessibility · automated testing · content adaptation · web proxy · WCAG evaluation
Standards referenced: WCAG 2.0