A passenger reaches the interchange and finds a notice on the lift door. The trains are running. The platforms are open. Most people keep moving. But the step-free route has vanished, perhaps after twenty minutes already spent travelling.
The failure looks local: one machine, one station, one repair ticket. Its effect is not local. For someone who cannot use the stairs, a single lift can determine whether the whole journey exists. Public transport therefore needs a more honest unit of accessibility: not the uptime of each asset, but the availability of a usable route from origin to destination.
Accessibility is a chain
A step-free journey is a chain of dependencies. It may require a street-to-concourse lift at the first station, a platform lift at an interchange, a boarding ramp, staff coordination and another lift at the destination. The trip succeeds only if every necessary link works at the moment it is needed.
This produces a counter-intuitive result. Four critical links that are each available 99 per cent of the time would, in a simplified model where failures are independent, all be available together only about 96 per cent of the time: 0.99 multiplied by itself four times. Real failures are not neatly independent—power faults, staffing gaps and maintenance can affect several links at once—so the calculation is only an illustration. The important point is that excellent-looking asset statistics can still add up to a materially less reliable journey.
Averages hide geography too. A lift with modest usage may be the only step-free connection between two lines. Another may have a ramped alternative metres away. Giving both faults equal weight measures maintenance activity, not lost access. The question with public consequence is: how many journeys became impossible, substantially longer or dependent on unplanned assistance?
Rules already point beyond the machine
Policy is beginning from the right premise. The European Union’s rail-passenger rights regulation requires operators and station managers to provide information about the accessibility of stations, facilities and services, and to make reasonable efforts to propose acceptable alternative transport in relevant cases. It treats access as part of the transport service, not as an optional feature of the building.
Great Britain’s Office of Rail and Road is more operationally specific. Its October 2025 Accessible Travel Policy guidance says operators must keep station information current when lifts or other facilities make a station temporarily inaccessible. It also requires reasonable efforts to provide accessible replacement facilities when normal accessibility is reduced, and, in defined circumstances, alternative accessible transport without extra charge.
Transport for London’s public status pages show why this matters. A lift fault is reported not merely as a broken asset but as the loss of step-free access to particular platforms, often with a prompt to replan. Its planned station and lift closures page also describes alternative entrances or lines when they exist. That is closer to the passenger’s actual problem: what remains reachable now?
Yet information can still be technically correct and practically weak. “Lift unavailable” says what happened to equipment. It may not say whether a journey can continue, how long the detour takes, whether the alternative is itself step-free, who will arrange it, or when the information was last checked. As Alkemata has argued about public warnings, a message becomes useful only when someone can act on it.
Measure the promise people are trying to use
A route-level accessibility measure would start with journeys rather than inventories. For a defined set of origins, destinations and access requirements, an operator could ask at regular intervals whether a viable path exists. A useful public record would include at least five things:
- the critical lifts, ramps, gates, vehicles and staffed actions on the path;
- whether each dependency was known to be available at the time;
- the accessible alternative if one link failed;
- the added time, distance and transfers imposed by that alternative; and
- the age and confidence of the underlying status information.
This need not collapse into one league-table score. A single number would again hide important differences. Operators could publish a small family of measures: the share of sampled journeys with a complete step-free path; the number of routes with a single point of failure; median extra time when replanning is required; and the proportion of faults for which a confirmed alternative was offered before the passenger arrived.
Different people also need different routes. “Step-free” does not describe doorway widths, gradients, boarding gaps, turning space, fatigue, sensory conditions or the availability of assistance. A route that works for a parent with a pushchair may not work for a power-wheelchair user. Route-level reporting should therefore state its access profile and allow several profiles, developed with disabled travellers, rather than claim universal accessibility.
The serious objection
The strongest argument against this approach is not that routes do not matter. It is that a live network contains too many routes, needs and changing conditions to certify them reliably. A stale green tick may be worse than an honest equipment list. Modelling could consume money better spent on repairs, duplicate journey planners and expose operators to promises they cannot keep.
That objection should shape the method. Start with a bounded sample: common interchanges, routes requested through assistance services and corridors with known single points of failure. Publish timestamps and uncertainty. Distinguish a confirmed observation from an inferred path. Keep raw asset status visible. Most importantly, do not use better information as a substitute for maintenance, staff or accessible alternatives.
Nor should passengers become unpaid monitoring infrastructure. Voluntary reports can reveal gaps, but the duty to maintain assets, verify status and provide assistance remains with the organisations operating the service. The value of route-level measurement is precisely that it gives those organisations a way to see where their separate responsibilities fail to join up.
A one-route accessibility audit
A transport authority, access group, university or workplace can test the idea without building a new platform. Choose one recurring journey that is advertised as step-free. Map every indispensable dependency from pavement to destination. Check the published status before departure and again at the expected time of use. Then model the failure of each lift in turn—using information and observation, never by obstructing equipment—and record whether a safe alternative exists, who must act, and the extra time it adds.
Run the audit weekly for a month and publish a one-page route card. It should show the weakest link, the oldest piece of status data, the best alternative and the contact responsible for arranging it. Invite disabled travellers to define what counts as a viable alternative and pay them for that expertise.
The result will not be a universal accessibility index. It will be something more useful: a test of whether the system’s operational data can answer a human question. If this lift fails, can I still get where I am going?
A city that can answer that question promptly has begun to treat access as a journey-wide public promise. A city that cannot may have impressive uptime figures, but it does not yet know when part of its network has closed.
