Progress, not perfection, across a portfolio of 1,618 sites.
An interactive look at where LSA's web accessibility program started, where it stands four days before the Title II deadline, and the work still ahead.
Five phases, one continuous practice.
Web accessibility at LSA is not a project with a finish line. It is an operational discipline. This is how the current phase took shape.
Foundation
Siteimprove configured across AEM and WordPress. Baseline established. First cross-site reports generated. Governance conversations begin with Maria and Mikhail.
Setup and onboarding
Canvas courses for AEM and WordPress editors launched. Open Lab sessions opened. Custom Siteimprove dashboards shared with Communications Administrators. Per-site remediation plans drafted.
- Platform-specific training published
- Stakeholder engagement templates in place
- TDX response workflows established with Caihui
The pivot to sustained practice
Title II compliance deadline four days away. AEM template-level ARIA fix staged for production. Only 12 of 1,373 active sites meet their accessibility target, though 268 more sit within 1.5 points. Portfolio target score averages 95.96. AEM sites trail at 91.17. Active engagement with stakeholders replaces setup work.
AEM template push, intern onboarded
AEM-only ARIA landmark fix ships. Roughly 51% of automated occurrences resolve. AEM average target score projected to rise toward 95. Stacy Shang joins for the summer. WordPress and Omeka template work enters the next-wave backlog.
Sustained tiered practice and culture-building
First-pass manual audits begin on Tier 1 sites. Manual testing courses launch for editors. Accessibility Champions network forms. Governance documentation formalized. Approximately 2,090 hours of first-pass effort phased across 18 to 24 months.
A portfolio, a deadline, a philosophy.
The program began with a question: how does a college with 1,600+ registered sites, four content platforms, and a federal compliance deadline actually move forward?
The answer was Progress Over Perfection. Rather than attempt a perfect audit of every page (an impossible task with available resources), the team committed to a continuous practice. Find the highest-leverage issues first. Fix them at the template level where they cascade. Train the editors who create new content every day. Build the systems that sustain the work after the deadline.
1,618 registered sites
Across Adobe Experience Manager, WordPress, Omeka, and legacy platforms. 185 had no scanned pages. 60 had only a placeholder page. Another 36 are unreachable. The real active surface: 1,373 sites, 463,926 pages.
April 24, 2026
ADA Title II compliance with WCAG 2.1 Level AA, established through U-M SPG 601.20. The college serves as a support partner in this effort, not a compliance enforcer, which shaped every choice about how we engage with site owners.
Automated scans can flag what is broken. They cannot tell you what your users are actually experiencing. That part is ours. — Program framing, LSA-TS Accessibility
The portfolio in April 2026.
Four lenses on the data: how sites are distributed, where they sit on the compliance scale, how platforms differ, and where the highest-leverage work is concentrated.
An "active site" in this report is a Siteimprove record with at least one crawled page. Zero-page records include empty, authenticated, or unreachable sites — they are excluded from every aggregate. This definition gives us 1,373 active sites carrying 463,926 pages. Of those, 12 currently hold "Target met" status; the remaining 1,361 are working toward it, with 268 more within 1.5 points of target.
Portfolio composition
Figure 3.1Of the 1,618 registered sites, 1,373 carry real content (at least one crawled page). 245 are empty, placeholder, or unreachable records. Of the active sites, only 12 have reached "target met" status in Siteimprove; the remaining 1,361 are working toward it.
Beyond WCAG: Siteimprove's Best Practices
A separate axisSiteimprove flags a "Best Practices" category alongside the WCAG conformance levels — guidance like consistent heading hierarchy, predictable focus order, descriptive link text. These don't directly affect Title II compliance (we score conformance on A + AA + ARIA), but they shape how usable the experience is for assistive-tech users and how durable the remediation work will be.
Spelling quality across the portfolio
Editorial signalSpelling errors aren't an accessibility violation, but they shape reader trust and degrade screen-reader output (mispronunciations, misparsed sentences). Siteimprove's QA module flags confirmed misspellings across the same crawl pass, so we get a free portfolio-wide quality read.
How close are sites to their target?
Figure 3.2Most active sites need only a handful of points to reach target. The distribution shows that 268 sites are within 1.5 points of target, which is a reasonable scope for editor-level remediation. Another 837 sites need 1.5 to 5 points, where template-level fixes come into play.
Score distribution across 1,372 active sites
Figure 3.3The target score and accessibility score tell different stories about the same portfolio. Target score reflects compliance conformance; accessibility score is more sensitive to best-practice deductions. Toggle to compare.
Platforms at a glance
Figure 3.4AEM carries roughly 1.3 times the pages of WordPress on 1/8 the number of sites, and has a much larger accessibility-to-target gap. AEM sites need an average of 7.27 points to reach target; WordPress sites average 2.85. The gap is where the template-level work lives.
Pareto of pages
Figure 3.5The top 50 sites contain 81% of all pages. Tiered audit planning follows this curve.
Top 10 sites by remediation debt
Figure 3.6Debt equals pages multiplied by the gap from a perfect target score. The top 10 account for roughly 65% of portfolio-wide debt. Three of the largest contributors deserve a closer look: sites.lsa.umich.edu is the WordPress parent aggregator, AEM Staging is a non-production environment, and the International Institute sub-sites appear to duplicate each other's page counts and should be reviewed for scan configuration.
Sites flagged in amber (AEM Staging, sites.lsa.umich.edu parent) likely do not represent real user-facing content and should be excluded from compliance reporting after confirmation.
What the data is telling us
One. The compliance picture has real room to grow. Only 12 of 1,373 active sites fully meet target. But 268 more sit within 1.5 points of target, which is realistic editor-level remediation scope.
Two. The operational picture is concentrated on AEM and a handful of large sites. AEM (91.17) and Omeka (91.41) trail WordPress (96.60) substantially. The 29 largest sites (each 2000+ pages) hold 75% of all pages.
Three. The inventory needs cleanup before compliance reporting. AEM Staging should be excluded from user-facing metrics. The International Institute sub-sites show duplicate page counts suggesting scan configuration needs review. sites.lsa.umich.edu is a parent aggregator that may double-count.
Four. The sustainability picture depends on systems, not heroics. A single AEM template change cascades across 184,077 pages. Editor culture, developer tooling, and governance are the long-term levers.
Half the automated volume, one platform at a time.
The most consequential remediation staged for deployment is an ARIA landmark template fix. It applies only to AEM sites. Here is what shifts when it lands, and what stays the same.
This fix is AEM-only. WordPress and Omeka sites also carry ARIA landmark issues in their themes, but those require separate template-level work that has not yet been scoped. The projections below assume the AEM fix ships as planned and non-AEM platforms remain unchanged.
Total automated occurrences across the active portfolio. 30.5M of these are ARIA-related, of which an estimated 58% sit on AEM pages (40% of portfolio page share, higher per-page issue density).
Roughly 51% reduction, driven by eliminating ARIA landmark occurrences on AEM's 184,077 pages. WordPress (144K pages) and Omeka (49K pages) persist until their templates receive similar treatment.
Projected target score uplift by platform
Figure 4.1AEM sees the gain. Other platforms are unchanged until their own template work is done. Portfolio average rises modestly because AEM represents 136 of 1,373 active sites.
Projections are estimates based on AEM's current accessibility-to-target gap and typical Siteimprove weighting. They should be validated against the first post-deployment scan.
Ownership of remaining work
Figure 4.2Development remains the dominant owner of what's left, because non-AEM ARIA issues still exist. Visual design's share rises but does not dominate the way it would have with a portfolio-wide fix.
Next-wave AEM priorities
Figure 4.3Once the ARIA fix ships, the biggest remaining AEM issues are target size and color contrast. Both have template components and would benefit from the same centralized fix approach.
Manual testing becomes the larger share of what's left
Before the push, the ratio of manual-only findings to automated findings is roughly 2 to 1. After the AEM push, that ratio rises to roughly 4 to 1. The total universe of accessibility issues on the sites does not change; the share of work that can only be done by human testers grows. Capacity planning should reflect this shift, even though the AEM-only scope means the reframe is less dramatic than a portfolio-wide fix would have been.
The 60 to 70 percent that needs a person.
Automated tools catch roughly a third of WCAG failures. The rest requires human judgment, assistive technology, and context. This is how that work breaks down.
Tester time by WCAG area
Figure 5.1Rough allocation of manual audit time across major WCAG categories. Keyboard, focus, and screen reader review together account for more than half of any session.
Expected findings per site
Figure 5.2Empirical ranges drawn from published audit literature. These estimate distinct remediation tasks, not raw occurrences. Most findings are systemic and repeat across templates.
The manual-only WCAG criteria
Figure 5.3Twenty success criteria that Siteimprove cannot reliably evaluate. Each requires human testers, assistive technology, or contextual judgment. This is a representative set, not a complete list.
| Area | WCAG | Criterion | Level |
|---|---|---|---|
| Keyboard | 2.1.1 | All functionality operable without a mouse | A |
| Keyboard | 2.1.2 | No keyboard traps in widgets or modals | A |
| Focus | 2.4.3 | Logical tab order | A |
| Focus | 2.4.7 | Focus indicator visible | AA |
| Focus | 2.4.11 | Focus not obscured by sticky elements | AA |
| Screen reader | 1.1.1 | Alt text describes the image meaningfully | A |
| Screen reader | 1.3.1 | Semantic structure conveys intended relationships | A |
| Screen reader | 2.4.6 | Headings and labels are descriptive | AA |
| Screen reader | 4.1.2 | Custom components expose name, role, value | A |
| Visual | 1.3.3 | Instructions not reliant on color or position | A |
| Visual | 1.4.1 | Color is not the sole indicator of meaning | A |
| Visual | 1.4.10 | Content reflows at 320px and 200% zoom | AA |
| Visual | 1.4.13 | Hover and focus content is dismissible | AA |
| Forms | 3.3.1 | Error identification explains what failed | A |
| Forms | 3.3.3 | Error suggestions help users fix problems | AA |
| Forms | 3.3.8 | Authentication accessible (no memory puzzles) | AA |
| Media | 1.2.2 | Captions accurate and synchronized | A |
| Media | 1.2.5 | Audio description provided where needed | AA |
| Cognitive | 3.2.3 | Navigation consistent across pages | AA |
| Cognitive | 3.2.4 | Same function has consistent identification | AA |
No dashboard can tell you what a screen reader user heard
Siteimprove is essential, but it is an early-warning system, not a conformance test. A heading labeled "More" passes every automated check. A form field with a label of "Field 1" passes too. A button with an accessible name of "submit-btn-1" is technically present. Manual testing is where compliance becomes lived experience.
A tiered plan sized for the real portfolio.
Approximately 2,090 hours of first-pass manual testing over 18 to 24 months, concentrated where user exposure actually lives.
Inventory cleanup
Empty, placeholder, or unreachable records. Decommission or rescan before compliance reporting. Includes 36 unreachable sites.
Full WCAG 2.1 AA audit
Top sites by pages. 81% of user exposure concentrated here. ~20 hours per site, representative page sample.
Abbreviated audit
Next band by size (7% additional page coverage). 4 hours per site focused on keyboard, screen reader, and any forms present.
Automated + spot check
Long tail, mostly small WordPress sites. Siteimprove monitoring plus a 20-minute automated-review check per site.
Top-5 debt sites
Residential College, Home, LSA Strategic Vision, chahrour, UMS 1580. Large-page sites that warrant dedicated remediation projects with site-owner collaboration.
Coverage achieved per tier
Figure 6.1The tiered approach earns its keep from Pareto distribution: Tier 1 alone covers 81% of pages. Adding Tier 2 covers 89%. The remaining 11% lives in the long tail of 1,172 smaller sites, which is where automation plus editor training carries the work.
Roadmap for the next 18 months
Figure 6.2Phased execution matched to capacity. AEM push completes in May. Tier 1 audits begin in June. Intern accelerates Tier 3 through August.
Accessibility is a practice, not a project. A compliance deadline is a beginning, not a finish line. — Closing note
Toward a culture where accessibility is the default.
Compliance deadlines are a beginning. What sustains the work is culture: shared language, shared tools, shared ownership.
LSA's accessibility program over the next three years is not measured by dashboards alone. It is measured by the number of editors who know how to test with a screen reader. The number of designers who think in reflow before they think in pixels. The number of site owners who ask "who will be excluded?" before launch. The number of conversations where accessibility is a design input, not a post-hoc audit. The compliance scan is the thermometer; the culture is the body temperature.
Where we are on the maturity curve
Figure 7.1A five-level capability maturity framework. LSA's accessibility program sits between "Documented" and "Practiced" today. The 18-month target is "Measured" — data-driven, integrated into everyday workflows. "Embedded" is the longer arc.
Five pillars of a sustainable accessibility system
Each pillar has a current state that we can build on, and a near-term set of investments that take the program from "program with processes" to "culture with practice."
Learn
Capability building for every role.
- Canvas courses for AEM editors
- Canvas courses for WordPress editors
- Open Lab drop-in sessions
- Quick-reference guides on the accessibility site
- Manual testing course for editors
- Screen reader basics workshop series
- Developer accessibility certification path
- Role-specific learning paths (editor, developer, designer, comms admin)
- LSA Web Accessibility lunchtime seminars
Practice
Daily habits and tools that make the right thing easy.
- Siteimprove continuous monitoring
- Custom dashboards shared with CAs
- TDX intake and triage workflows
- AEM component accessibility tracker
- Accessibility browser toolkit rollout (axe DevTools, WAVE)
- Pre-publish accessibility checklists in CMS
- Developer pre-commit accessibility linting
- Template-first approach for new content types
- Alt text gem and supporting tooling for editors
Connect
Community, recognition, shared ownership.
- Open Lab monthly sessions
- Stakeholder engagement templates
- Connection to U-M Digital Accessibility Strategic Initiative
- Accessibility Champions network across LSA units
- Quarterly "Progress stories" showcase
- Dedicated Slack channel for accessibility questions
- Annual recognition for editors making measurable improvements
- Cross-unit sharing of audit findings and fixes
Govern
Policies, gates, and the shift-left principle.
- U-M SPG 601.20 framework
- LSA-level accessibility implementation guide (draft)
- Technical Service Expectations document
- Accessibility review as a gate in design and dev workflows
- Procurement accessibility requirements for third-party tools and vendors
- Template approval process with accessibility sign-off
- Annual accessibility review cycles per unit
- Publication standards for editorial content (alt text, headings, links)
Measure
Transparency and feedback loops.
- Siteimprove scores and dashboards
- This report (and future iterations)
- TDX ticket metrics
- Editor-specific progress reports sent quarterly
- Site-owner quarterly review meetings for top 50 sites
- Culture pulse surveys (twice yearly)
- Public accessibility status page
- Manual audit finding tracker (JIRA) with pattern analysis
From "LSA-TS does it" to "the college owns it, we support."
Every accessibility ticket that becomes a self-service correction. Every editor who catches a contrast issue during review. Every developer who ships a component keyboard-tested from the start. Every designer who reviews a mockup with reflow in mind. These are the markers of a sustainable system.
Our team does not shrink in this model. We become the infrastructure that lets the rest of the college own inclusion. Training, tooling, templates, office hours, escalation paths. Editors do the work of accessibility; we make sure they have everything they need to do it well.
Manual testing course for editors · proposed outline
Planned launch · Jul 2026A self-paced Canvas course to build confidence in basic manual accessibility testing. Target audience: Communications Administrators and content editors on AEM and WordPress. Expected completion time: 3 to 4 hours across four modules, with hands-on exercises in each.
Course development scoped for Stacy Shang's summer project. First cohort opens to Communications Administrators; expanded access follows iteration on feedback.
Seven priorities shaping the work ahead.
Resolve the inventory and data quality questions
Reconcile the 245 inactive records (185 zero-page, 60 single-page, 36 unreachable). Investigate AEM Staging showing up in production dashboards. Validate why ten International Institute sub-sites share identical 9,677 page counts. Clean data is a prerequisite to any compliance reporting.
Ship the AEM ARIA template fix
Confirm production timeline with Ananta. Validate score uplift against the first post-deployment scan. Identify the second wave of AEM template work (target size, contrast) and scope tickets for it.
Scope the WordPress and Omeka template work
The AEM fix only addresses one platform. WordPress (1,148 active sites) and Omeka (12 sites with 48,579 pages) still carry ARIA landmark issues in their themes. Separate scoping conversation with Jessica and the WP team should follow the AEM push.
Launch the manual testing course with Stacy
Four-module Canvas course for editors, scoped for Stacy Shang's summer project. Pilot with Communications Administrators in August. Expand access across the college after two rounds of feedback iteration.
Start the Accessibility Champions network
Identify one editor or developer per major LSA unit to serve as a point of contact and advocate for accessibility work in their department. Monthly gatherings, shared Slack channel, recognition for contributions. The network becomes the culture's connective tissue.
Launch Tier 1 manual audits
Build page-sampling frames for the top 50 sites. Document one complete audit as reference methodology. Begin weekly cadence in June. Use early findings to refine the audit template and Canvas course content.
Open conversations on the top-5 debt sites
Scoped conversations with owners of Residential College, Home (lsa.umich.edu), LSA Strategic Vision, chahrour, and UMS 1580. These five sites together represent more than 2 million units of remediation debt. Each deserves a dedicated plan rather than a generic tier assignment.