Diagnosing Mobility and Accessibility · Lesson 5

Audit service regularity and missing service periods

6 min read

Lesson objectives

  • Turn timetables and observed departures into an interpretable regularity measure and explain the effect of uneven headways on waiting and access.
Reading tools

A map may show a route between a home and a job, but actual journey time also depends on when vehicles arrive. The time between consecutive departures at one stop is a headway, distinct from in-vehicle time and boarding count. If a timetable promises ten-minute intervals but a fifteen-minute gap occurs, a worker can miss the start time even while bus running speed is unchanged. Identify a stop, direction, period, day type and observation source before calculating regularity. Do not mix morning peak and afternoon gaps in one average when the service decision concerns the peak. Separate scheduled departure, observed departure and a hypothetical rider's arrival at the stop. Automated feeds can omit a vehicle or record a passing vehicle that did not board passengers; inspect completeness. A posted timetable describes planned supply, while effective resident access still depends on walking, connections, return service and fare.

Construct the headway sequence by subtracting each observed departure time from the next, correctly handling hour and date changes. Six departures yield five within-window gaps, not six, unless a preceding or following departure is also observed. Calculate the mean gap, median, share exceeding a locally declared operating threshold and maximum gap. The mean can hide bunching: a two-minute gap followed by eighteen minutes averages ten but gives riders very different waits. If rider arrivals are random and uniform over time, expected wait across observed gaps is the sum of squared gap lengths divided by twice their sum. This is conditional on the arrival assumption and stable observations; it is not every individual's waiting time. On a low-frequency line, riders may plan around a known timetable, changing the model. Where service is uncertain, they may arrive early and spend contingency time the formula omits. State the behavioural assumption whenever the formula is used.

Compare planned and observed service for each pair of departures, not only a daily total. If a late bus is compensated near the end of the hour, the number of completed trips may equal the published count even though a critical work-start window had no service. Record cancellations, diversions and skipped stops as well as endpoint times. Separate lateness from regularity. Vehicles that are all five minutes late but remain evenly spaced can still be predictable to a frequent user; vehicles bunched around their schedule can make access worse. Connect operational indicators to user outcomes: the share who reach a job or clinic within an acceptable time, the periods lacking a return journey, and wait at an accessible stop for people with mobility constraints. Without a representative week of observation, avoid generalizing from one morning. Record weather, holidays, events and construction if they affect comparison; do not automatically delete a poor day because it weakens the result.

Data quality checks are concrete. Match route, direction and stop identifiers, use a consistent time zone, and remove duplicate messages while retaining an exclusion log. A device with the wrong clock may appear to depart before it arrives; do not silently rewrite the record. Verify gaps in an automated feed with operator or field records. Dividing incomplete intervals by a complete-interval count can understate waiting if a long service gap was lost. A regularity dashboard should show observation days and feed completeness alongside the indicator. Agree on a local operating threshold with operator and users; 'over twelve minutes' is not a universal standard. Respect its mathematical boundary: exactly twelve does not breach a rule stated as greater than twelve, but rider experience does not jump at 12.1 minutes. Display the distribution or longest gap as well, so the threshold does not hide how far a breach exceeds the limit.

Lesson summary

Headway regularity changes waiting and access even when vehicle speed is unchanged; calculations require a stated arrival model and known data completeness.