Two weeks before the bid deadline, blocked by one badge
An anonymised case: a medical device company bid on a city government contract whose specification required the site to pass accessibility testing at level AA. Engineering got the ticket two weeks before the deadline. The first Freego scan returned more than three hundred findings: product images with no alternative text, an entire quote form with no labels, a floating support button trapping keyboard focus, and a carousel whose controls were div elements pretending to be buttons. The emergency remediation took 38 working days and NT$210,000, the bid was missed anyway, and the company only qualified the following year. The problem was never technical difficulty. It was treating accessibility as a pre-launch checklist item instead of a requirement.
Who should start now, who can wait
Start now
- You bid on Taiwanese government or public school contracts that require the official accessibility certification badge.
- You sell into the EU: the European Accessibility Act has applied to e-commerce services since 28 June 2025, and non-EU sellers serving EU consumers are in scope.
- Your audience includes older or disabled users: banking, insurance, healthcare, elder care, utilities.
- Your conversions depend on forms: registration, appointments, quote requests, applications.
- You need auditable digital inclusion evidence for an ESG or sustainability report.
Wait
- A full redesign ships within three months: fold accessibility into the new design system instead of paying twice.
- Internal tools with a fixed user base: make them fully keyboard operable and stop there for now.
- An MVP whose features still change weekly: lock down the critical task paths first.
- Your budget only covers one year of an overlay widget: an audit plus fixing the four highest-frequency defect types buys more.
- You cannot name your three critical task paths: map the flows first.
Four routes, real coverage and cost
| Approach | What it actually covers | Time | First-year cost | Limits |
|---|---|---|---|---|
| Overlay widget (accessWidget and similar) | Cosmetic layer, source code untouched | 1 day | NT$16,000–130,000/yr | Not accepted as conformance evidence; screen reader users often turn it off |
| Automated scan, fix in-house | Roughly 30–40% of WCAG issues | 2–4 weeks | NT$0–30,000 in internal hours | Misses keyboard traps, focus order, whether alt text is actually correct |
| Professional manual remediation | Can reach WCAG 2.2 AA | 4–8 weeks | NT$80,000–250,000 | Needs engineering capacity; decade-old templates may resist change |
| Front-end rebuild with an accessible component library | AA that survives future releases | 8–16 weeks | From NT$300,000 | Highest cost, requires design system changes in parallel |
For most SMEs the sweet spot is route three. Route four pays back only when you ship redesigns more than twice a year and reuse components heavily.
Four-phase rollout
- Phase 0: scope and baseline audit (3–5 days). Pick 15–25 representative pages plus three critical task paths, for example find product, add to cart, checkout, payment confirmed. Tools: axe DevTools, Lighthouse, Freego. Deliverable: a defect list mapped to WCAG 2.2 success criteria plus baseline scores.
- Phase 1: manual verification and prioritisation (5–7 days). Walk all three paths using only Tab, Enter and Esc, sample-test with NVDA in Chrome, and measure every text and background pair with the WebAIM Contrast Checker. Deliverable: P0/P1/P2 remediation list with hour estimates.
- Phase 2: remediation and componentisation (3–6 weeks). Fix contrast, missing alternative text, unlabelled form inputs, and empty links and buttons first: those account for 96% of detectable errors across the web. Then rebuild modals, date pickers and carousels as shared components with real focus management. Deliverable: PR list and an accessible component library.
- Phase 3: regression gate and submission (1–2 weeks). Wire axe-core or pa11y-ci into CI so any new violation blocks the pull request. For public tenders, run the Freego self-assessment and file the badge application. Deliverable: CI config, published accessibility statement, submission pack.
What it really costs
Project fees at 2026 Taiwan market rates: audit and defect report NT$25,000–60,000, manual remediation NT$80,000–250,000, rebuild with a component library from NT$300,000. ScriptWalker's Web Accessibility Audit and Remediation package starts at NT$35,000 for the audit, NT$90,000 for remediation, and NT$2,500 per month for regression monitoring.
The line items nobody budgets for:
- Writing alternative text: 500 product images at two minutes each is about 17 hours, NT$12,000–25,000, and it has to be written by someone who knows the products.
- Captions and transcripts: NT$40–120 per minute, so thirty five-minute videos is NT$6,000–18,000.
- Accessible PDFs: NT$150–400 per page for catalogues and investor decks, frequently more expensive than the website itself.
- Replacing third-party components: support widgets, carousel plugins, an accessible alternative to reCAPTCHA, NT$20,000–60,000.
- Overlay subscriptions: accessWidget is priced by monthly traffic from about USD 59 per month, and the tier jumps as traffic grows.
- Annual re-testing and badge resubmission, 8–16 hours, repeated after every significant redesign.
What clients expect vs what happens
- Expectation: one script tag makes us compliant. Reality: overlays never touch the source. The Overlay Fact Sheet is signed by hundreds of accessibility practitioners opposing them, and 456 US web accessibility lawsuits in the first half of 2025 targeted sites that already had a widget installed.
- Expectation: a Lighthouse score of 100 means we passed. Reality: automated tooling covers roughly a third to 40% of the criteria. Keyboard traps, focus order and whether alt text is meaningful are invisible to it.
- Expectation: this is a visual design problem. Reality: the top defects are low contrast text at 83.9% of pages, missing alt text at 53.1% and unlabelled form inputs at 51%, most of which live in front-end code and content workflow.
- Expectation: we fix it once. Reality: average home page element count rose 22.5% in a single year. Without a CI gate you regress within a quarter.
- Expectation: more ARIA means more accessible. Reality: WebAIM Million 2026 found pages using ARIA averaged 59.1 errors against 42 for pages without it.
Six traps and how to avoid them
- Fixing the home page but not checkout. Fix: accept work by critical task path, and treat any broken step as a failed path.
- Buttons built from div plus onclick. Fix: use native button, a, label and fieldset before reaching for ARIA. ARIA is a patch, not semantics.
- Alt text that is a filename or the word image. Fix: describe purpose for functional images, use alt="" for decorative ones, and make the field mandatory in the CMS.
- Third-party components blocking the keyboard. Fix: inventory support widgets, payment iframes and captchas at contract stage and require an accessibility statement from each vendor.
- Chinese text baked into images. Fix: use real text with CSS, and confirm the html element carries lang="zh-Hant". Globally 12.7% of home pages still declare no language at all.
- No regression gate after remediation. Fix: run axe-core in CI, publish a monthly scan report, and put accessibility into design review and the PR template.
Success metrics and a 90-day roadmap
- Day 30: baseline audit complete, all P0 items cleared (paths the keyboard cannot finish, unlabelled forms, functional images without alt text), zero critical axe issues on home and checkout.
- Day 60: site-wide text contrast at AA, all three critical paths completable by keyboard alone, NVDA can complete checkout, CI blocking new violations, accessibility statement published.
- Day 90: 15–25 representative pages pass a WCAG 2.2 AA self-assessment, Freego testing passed and the badge application filed, monthly regression reporting running, and before-and-after comparison of form completion rate and support call volume.
Decision checklist
- ☐ I know whether we are targeting level A or AA
- ☐ I can name our three critical task paths
- ☐ Someone has completed checkout using only a keyboard
- ☐ Someone has tested at least one path with a screen reader
- ☐ Our html element carries a correct lang attribute
- ☐ Every form field has a programmatically associated label
- ☐ Functional images have alt text and decorative ones use empty alt
- ☐ All text meets 4.5:1 contrast, or 3:1 for large text
- ☐ Keyboard focus is visible on every interactive component
- ☐ Support widget, captcha and payment pages were keyboard tested
- ☐ Videos have captions and PDF catalogues have a real text layer
- ☐ CI or a monthly job scans automatically and notifies someone
- ☐ A public accessibility statement with a contact channel exists
Fewer than eight ticks means a submission today would most likely be rejected.
FAQ
Does installing an accessibility widget make us compliant?
No. An overlay paints over the page in the browser while the underlying code stays broken, and screen reader users frequently have to switch it off to get anything done. Taiwan's badge testing inspects the page source and structure, so an overlay will not get you through Freego. The defensible use is temporary mitigation while real remediation is scheduled, never as the destination.
We scored 100 on Lighthouse accessibility. Are we done?
No. Automated tooling covers roughly a third to 40% of the success criteria and the rest needs a human. The minimum viable manual pass is three things: complete all three critical paths with the keyboard only, sample-test forms and checkout with NVDA or VoiceOver, and measure contrast pair by pair. Together that is usually 4–8 hours and it surfaces what automation structurally cannot see.
Do we have to apply for the accessibility badge?
It depends who your customers are. Government agencies and public schools are obliged to apply under Ministry of Digital Affairs rules, and tender specifications usually name the required level outright. Private companies face no blanket mandate, but if you sell to government or into the EU under the EAA, the badge and a published accessibility statement are the cheapest proof to hand an auditor.
Will remediation ruin our design?
In practice the two things that change most are contrast and focus styling, and both can be solved inside the existing brand palette by darkening text one or two steps and using a brand-coloured focus ring with an offset. What genuinely needs redrawing is usually banners with text baked into the image and icon-only buttons with no accessible name. Both already hurt SEO, so fixing them together is a net gain.
Next step
If you do not know how far you are from AA, skip the quote conversation for now. We run a 60-minute accessibility health check covering an automated scan, a keyboard walkthrough and one screen reader path, and you get back a defect list mapped to WCAG 2.2 criteria, an hour estimate and tiered pricing. Then you decide which route to take.
- Email: [email protected]
- Phone: 0916-224-047
- LINE: @ufv9089p