There is a moment familiar to anyone who has planned a day out around a body that does not cooperate. You find the place. The listing says accessible. You go. And the ramp is at 1:8 with no handrail, or the accessible restroom is a store cupboard with a grab rail screwed to the wall, or the room is so loud that forty minutes in you have to leave anyway.
The word was not a lie, exactly. Somebody ticked a box honestly. The problem is that the box was asking one question, and you needed the answers to five.
The five questions
When we rebuilt how Wayfind stores access data, we split that single flag into five separate dimensions, because they fail independently:
Mobility — entrances, steps, gradients, door widths, surfaces underfoot. Vision — signage contrast, tactile guidance, audio description, lighting. Hearing — induction loops, captioning, whether announcements are visual. Sensory — crowding, noise levels and when they peak, quiet spaces, lighting. Stamina — distances, seating frequency, wait times, rest opportunities.
A concert hall can be exemplary on mobility and unusable on sensory. A gallery can be flawless on stamina — benches in every room — and hostile on vision. The single-flag model cannot express either of those, so it tells you neither.
The fields we leave blank
Here is the part that matters more, and that most sites would quietly skip.
Most guides do not yet have all five dimensions filled in. Open the Rome guide and you will see rows marked not yet rated. That is deliberate, and we are not going to fix it by generating the missing values.
An access rating is not a recommendation. If a restaurant recommendation is wrong, you have a mediocre dinner. If an access rating is wrong, someone arrives at a building they cannot enter, in a city they do not know, possibly having spent money and energy they did not have to spare. Those two failure modes are not comparable, so they do not get the same standard of evidence.
So the rule, written into the schema itself: an access claim is never generated. It comes from a person who has been there, or the field stays empty and says so. An empty field is honest. A plausible guess is not.
What this costs us
It costs us the ability to say we have complete data, which we do not. It makes the summary panel look sparse on plenty of guides. It means the work of filling those fields is slow, human, and dependent on people telling us what they found.
That seems like the right trade. A sparse table that you can trust is worth more than a full one you cannot — and if you have been somewhere and know what the entrance is actually like, the forum is where that knowledge does the most good.