What works for most people can exclude someone even when nobody intends exclusion. An application that is clear to a person with mobile data, adequate sight, spare time, and a private account can be a locked door to somebody else. Looking beyond the screen means noticing place, body, language, care networks, and the objects that make each action possible. Accessibility is not a final layer or a checklist; it is the discipline of refusing to turn average convenience into a compulsory condition of participation.
At this stage, the organisation must recognise real barriers with a named owner, a published timeframe, and an alternative that does not depend on having the newest device. It should test the journey with people using screen readers, unreliable connections, different languages, or third-party support; these conditions are not decorative exceptions but an ordinary part of service. Evidence should combine technical records, observation of real work, and confidential conversations. If harm appears, the rule should be paused, explained plainly, the loss remedied, and the change recorded. Measuring only clicks or rapid closures would reward the appearance of efficiency while concealing work shifted to households, frontline staff, and material resources.
Design for complete situations#
A responsible digital service begins by describing the whole task. Asking whether a button responds is not enough: a team needs to know where the person is, what happens when it fails, what another attempt costs, and who can help without invading privacy. Teams often design from a desk with reliable broadband and current screens. That viewpoint misses journeys, shift work, closed libraries, shared phones, and the exhaustion of repeating an explanation. The right decision may be digital, in person, or blended; what matters is an equivalent result.
A responsible digital service begins by describing the whole task. Asking whether a button responds is not enough: a team needs to know where the person is, what happens when it fails, what another attempt costs, and who can help without invading privacy.
At this stage, the organisation must describe the complete task with a named owner, a published timeframe, and an alternative that does not depend on having the newest device. It should test the journey with people using screen readers, unreliable connections, different languages, or third-party support; these conditions are not decorative exceptions but an ordinary part of service. Evidence should combine technical records, observation of real work, and confidential conversations. If harm appears, the rule should be paused, explained plainly, the loss remedied, and the change recorded. Measuring only clicks or rapid closures would reward the appearance of efficiency while concealing work shifted to households, frontline staff, and material resources.
Case 1: an inaccessible medical appointment#
A patient receives a link by text to confirm a specialist appointment. It opens a visual calendar that works poorly with a screen reader and expires after two hours. Her support person tries to help, but verification goes only to the patients number. The interface has turned healthcare into a test of dexterity and availability. The clinic does not solve this by adding a tutorial video: it retains the appointment, provides authorised telephone support, and asks why a short deadline decides a right whose importance is not short.
At this stage, the organisation must protect access to care with a named owner, a published timeframe, and an alternative that does not depend on having the newest device. It should test the journey with people using screen readers, unreliable connections, different languages, or third-party support; these conditions are not decorative exceptions but an ordinary part of service. Evidence should combine technical records, observation of real work, and confidential conversations. If harm appears, the rule should be paused, explained plainly, the loss remedied, and the change recorded. Measuring only clicks or rapid closures would reward the appearance of efficiency while concealing work shifted to households, frontline staff, and material resources.
Case 2: the employment portal that filters people out#
A job seeker uses a public computer. The portal requires a video introduction, identity confirmation on a recent phone, and an active session while a long form uploads. On the third attempt the browser deletes the progress. The employer may call this abandonment, although hours went into the application. A fair process accepts alternative formats, saves drafts without a burdensome account, and explains where human assessment happens. Recruitment should not mistake access to technology for professional capability.
At this stage, the organisation must accept alternative formats with a named owner, a published timeframe, and an alternative that does not depend on having the newest device. It should test the journey with people using screen readers, unreliable connections, different languages, or third-party support; these conditions are not decorative exceptions but an ordinary part of service. Evidence should combine technical records, observation of real work, and confidential conversations. If harm appears, the rule should be paused, explained plainly, the loss remedied, and the change recorded. Measuring only clicks or rapid closures would reward the appearance of efficiency while concealing work shifted to households, frontline staff, and material resources.
Case 3: an alert nobody can understand#
A municipality replaces emergency notices with app notifications. A resident with hearing loss receives an ambiguous vibration and cannot tell whether to evacuate, close windows, or wait. Other people have no battery, do not understand the message language, or live with irregular coverage. The failure is not that someone read a screen badly; it is treating one channel as the entire environment. Critical alerts need redundancy: sirens, radio, neighbourhood communication, plain text, and accessible channels that give consistent instructions.
At this stage, the organisation must coordinate redundant channels with a named owner, a published timeframe, and an alternative that does not depend on having the newest device. It should test the journey with people using screen readers, unreliable connections, different languages, or third-party support; these conditions are not decorative exceptions but an ordinary part of service. Evidence should combine technical records, observation of real work, and confidential conversations. If harm appears, the rule should be paused, explained plainly, the loss remedied, and the change recorded. Measuring only clicks or rapid closures would reward the appearance of efficiency while concealing work shifted to households, frontline staff, and material resources.
The material cost of false simplicity#
The promise of a simple experience can conceal frequent replacement. When a heavy application requires a still-useful phone to be replaced, or a connected sensor stops working after support ends, exclusion also becomes electronic waste. People with fewer resources lose access first, while buyers of new equipment bear costs rarely visible on a supplier dashboard. Inclusion must consider maintenance, energy demand, repairability, backward compatibility, and use of essential functions without permanent servers. Lightweight design can be social and environmental policy at once.
At this stage, the organisation must maintain the useful life of equipment with a named owner, a published timeframe, and an alternative that does not depend on having the newest device. It should test the journey with people using screen readers, unreliable connections, different languages, or third-party support; these conditions are not decorative exceptions but an ordinary part of service. Evidence should combine technical records, observation of real work, and confidential conversations. If harm appears, the rule should be paused, explained plainly, the loss remedied, and the change recorded. Measuring only clicks or rapid closures would reward the appearance of efficiency while concealing work shifted to households, frontline staff, and material resources.
Governance before personalisation#
Personalisation can adjust text, size, or recommendations, but it should not silently decide which services a person receives. Adaptation data can disclose health, finances, location, or family routines. When the logic is opaque, assistance becomes surveillance and a suggestion becomes a barrier. Governance sets purpose, minimisation, retention periods, and independent review. It also requires a usable explanation: what was decided, which information mattered, how an error can be corrected, and who can depart from an automated recommendation.
At this stage, the organisation must limit data use with a named owner, a published timeframe, and an alternative that does not depend on having the newest device. It should test the journey with people using screen readers, unreliable connections, different languages, or third-party support; these conditions are not decorative exceptions but an ordinary part of service. Evidence should combine technical records, observation of real work, and confidential conversations. If harm appears, the rule should be paused, explained plainly, the loss remedied, and the change recorded. Measuring only clicks or rapid closures would reward the appearance of efficiency while concealing work shifted to households, frontline staff, and material resources.
Testing with authority to change#
Accessibility testing loses value if the team cannot alter the product after finding a problem. Inviting participants to inspect a version already closed turns listening into ceremony. Teams should test early, pay for participants expertise, and reserve funding for content, compatibility, and human-support changes. A difficult case is especially revealing: a person balancing care and paid work, a household with limited connectivity, or somebody using assistive technology exposes dependencies an ideal test misses. Fixing before release costs less than demanding permanent adaptation later.
At this stage, the organisation must include participants experience with a named owner, a published timeframe, and an alternative that does not depend on having the newest device. It should test the journey with people using screen readers, unreliable connections, different languages, or third-party support; these conditions are not decorative exceptions but an ordinary part of service. Evidence should combine technical records, observation of real work, and confidential conversations. If harm appears, the rule should be paused, explained plainly, the loss remedied, and the change recorded. Measuring only clicks or rapid closures would reward the appearance of efficiency while concealing work shifted to households, frontline staff, and material resources.
Measures that reveal who is missing#
Measures should count outcomes rather than activity alone. An organisation can record how many people complete an application, yet it needs to learn why others do not, how long remedy takes, and whether an alternative channel actually works. Results should be disaggregated carefully, privacy protected, and stories reviewed when percentages do not explain them. It is also useful to track compatible-device life, data consumption, repeated calls, and cases resolved without compulsory travel. The aim is not a profile of every user; it is discovering rules that distribute effort unfairly.
At this stage, the organisation must examine unequal outcomes with a named owner, a published timeframe, and an alternative that does not depend on having the newest device. It should test the journey with people using screen readers, unreliable connections, different languages, or third-party support; these conditions are not decorative exceptions but an ordinary part of service. Evidence should combine technical records, observation of real work, and confidential conversations. If harm appears, the rule should be paused, explained plainly, the loss remedied, and the change recorded. Measuring only clicks or rapid closures would reward the appearance of efficiency while concealing work shifted to households, frontline staff, and material resources.
A commitment that can be checked#
A serious commitment appears in checkable decisions. An organisation can publish equivalent channels, maximum times, escalation owners, and compatibility criteria. It should announce changes before withdrawing a function, offer help without punishment, and account for problems repaired. Mature inclusion does not promise that everybody uses the same screen in the same way. It promises that nobody loses a right, opportunity, or essential warning because design treated the average as a boundary. That promise needs resources, training, and institutional humility to recognise that a good product can harm beyond its frame.
At this stage, the organisation must publish the remedy delivered with a named owner, a published timeframe, and an alternative that does not depend on having the newest device. It should test the journey with people using screen readers, unreliable connections, different languages, or third-party support; these conditions are not decorative exceptions but an ordinary part of service. Evidence should combine technical records, observation of real work, and confidential conversations. If harm appears, the rule should be paused, explained plainly, the loss remedied, and the change recorded. Measuring only clicks or rapid closures would reward the appearance of efficiency while concealing work shifted to households, frontline staff, and material resources.
Fuentes institucionales#
The European Commission on disability and rights, the WHO on disability, the OECD on digital government, and UNEPs Global E-waste Monitor provide institutional frameworks for connecting access, public accountability, and the material life of technology.
