Unmet demand, the indicator almost nobody measures

Open the monthly report of any public transport service and you will find passengers, kilometres, average occupancy, punctuality and cost. Everything on the page describes what the system did. Nothing describes what the system could not do.

And yet that is exactly where the information that allows improvement lives.

Why a fixed route cannot measure it

A scheduled bus has no way of knowing who did not board it. The person who checks the timetable, sees that the only workable departure is two hours before they need it and decides to drive, leaves no trace. There is no record of a "failed attempt".

The result is a classic survivorship bias: the system optimises around the demand it already serves, and gets better and better at serving whoever it already served, while potential demand becomes invisible and eventually disappears.

What changes with a flexible service

In a demand-responsive service every request is recorded, whether it is served or not. That produces data that did not exist before:

• Requests refused for lack of capacity. More fleet is needed in that time band.
• Requests refused for being outside the zone. The service area is drawn wrong.
• Requests refused for being outside operating hours. The service window does not match how people actually live.
• Requests accepted but cancelled by the user. Almost always a sign that the time offered was not really usable.

Each of those four categories points to a different decision. And none of them looks like "the service needs reinforcing", which is the generic conclusion you reach when you only measure occupancy.

An example of reading the data

Imagine a district service with 92% of requests served. That sounds good. But breaking down the remaining 8% shows that two thirds of the refusals cluster between 06:30 and 07:15, and almost all of them are headed for the same industrial estate.

That is not a general capacity problem: it is a very specific need for reinforcement, probably a single vehicle, across a forty-five minute window. The difference between those two readings can be an entire contract.

The 8% the system refuses usually contains more planning information than the 92% it serves without incident.

How to start measuring it

You do not need to rebuild the system. Start with the cheapest thing available:

1. Log every call to the customer service line, including the ones that do not end in a trip, with the reason.
2. Record in the phone channel when a user asks for one time and accepts a different one, and by how much.
3. Publish that figure in the quarterly report alongside the classic indicators, even if it is imperfect at first.

The moment that number appears in a report, the conversation changes. People stop arguing about whether the service "works well" and start asking who is still being left out.