How Reliable Are Hong Kong Bus Arrival Times

Do not trust the countdown on a cross-harbour route. A published journey time that assumes free-flow traffic stretches by 15 to 40 percent during peak hours. The bus you are watching on the app can flicker from three minutes to twelve, then vanish. The real-time estimate is recalculating against a traffic jam you cannot see.

The core problem is not the tracking technology. It is the harbour tunnels. A single breakdown or queue at the Cross-Harbour Tunnel toll plaza sets off a chain of delays that the system cannot recover from during the same peak. The result is bus bunching: two or three buses arriving at your stop together, then nothing for 20 minutes. The schedule is an aspiration, not a promise.

Routes that never touch a tunnel portal are the most dependable. KMB services running entirely within the New Territories, or Citybus routes contained on Hong Kong Island, match their published journey times far more often. The variable is tunnel approach congestion. The Western Harbour Crossing approaches suffer worst. Toll plaza lanes feed into a merge that backs up onto the West Kowloon Highway during the morning rush. Eastern Harbour Crossing routes see similar bunching where the Kwun Tong Bypass meets the portal.

Overnight and N-prefix routes run closer to schedule. Traffic is light. The gap between a published journey time and an actual journey time shrinks to single-digit minutes after 23:00. For a commuter deciding whether to trust the arrival board, this distinction matters more than the operator name painted on the bus.

The government aggregator pulls telematics data from every franchised bus. The ETA algorithm blends scheduled headway with live GPS position and a historical traffic model. When a bus is moving steadily, the countdown is accurate to within one minute. The moment the bus stops moving, the ETA freezes or climbs. The system does not tell you why. A reading of "Arriving" that lasts four minutes usually means the bus is stationary at a set of lights before your stop.

The platform shows real-time arrival data for KMB, Citybus, Long Win, and most Cityflyer routes. NLB coverage is partial. Some NLB South Lantau routes show only scheduled departure times. The telematics installation is incomplete. Green minibus coverage is patchy. Red minibus routes have no real-time arrival data anywhere. If you are standing at a red minibus stop, you wait until you see the roof.

When The ETA Lies

A bus that shows "Due" and then disappears from the board has likely been redirected by a controller, pulled out of service, or the GPS unit has dropped offline. Cross-harbour routes are the worst offenders. A bus delayed in the Cross-Harbour Tunnel southbound approach can hold an ETA of five minutes for the next ten minutes, then suddenly appear as two minutes away when the tunnel exit clears. The system is reactive, not predictive.

Operator apps from KMB and Citybus sometimes update faster than the government platform. They poll their own fleet directly. The government platform lags by a few seconds. For a bus that is one minute away, that lag is the difference between seeing it and missing it. Check the operator app for the route you are about to board, especially during peak hours.

The telematics unit on each franchised bus reports position, speed, and door status to a central server. The accuracy of a Hong Kong bus real-time arrival estimate depends on how often that unit transmits. Most KMB and Citybus vehicles report every 10 to 30 seconds. NLB buses on remote routes may update less frequently. Mobile network gaps in South Lantau cause dropouts. A bus on Ngong Ping Road can disappear from the map for minutes at a time.

The system does not account for a wheelchair ramp deployment adding 90 seconds to a stop dwell. It does not know that the bus ahead has a driver change scheduled. It cannot see that a red minibus has stopped illegally in a bus lane, forcing every franchised bus behind it to merge right. These are the small errors that accumulate across a trip and turn a five-minute headway into a 15-minute wait.

Bus bunching is a systemic failure mode, not a driver error. It begins at the tunnel approach. A bus delayed by five minutes at the Cross-Harbour Tunnel portal picks up more passengers at every subsequent stop. The headway has stretched. Each extra passenger adds dwell time. The bus behind it, running closer to schedule, picks up fewer passengers and dwells for less time. It catches the delayed bus. A third bus, running on schedule, catches them both.

On a route with a scheduled frequency of six minutes at peak, the actual headway after bunching can be zero minutes for the platoon, then 18 minutes for the gap behind it. The controller at the depot can instruct a bus to hold at a timing point. On a cross-harbour route there are few safe places to hold. A bus stopped on a flyover approach creates a hazard. So the platoon rolls through, and the gap grows.

The worst corridors are the Western Harbour Crossing northbound in the evening peak and the Eastern Harbour Crossing southbound in the morning. Routes that use the Bus Only Lane on the approach roads fare slightly better. The lane ends before the tunnel portal. The merge point is where the queue reforms. A bus lane buys you 90 seconds on a 40-minute delay.

Traffic congestion is the headline cause, but it is only the trigger. The deeper delay causes are structural. A driver change at a remote terminus that runs three minutes late pushes the next trip into the peak shoulder. A bus that breaks down in the Cross-Harbour Tunnel southbound tube blocks a lane for 15 minutes. Every route sharing that tunnel loses its headway. A green minibus stopping in a bus lane to pick up passengers forces all following franchised buses to wait for a gap in the adjacent lane.

Service disruption from road works is the other predictable variable. The Transport Department publishes road occupation permits. The knock-on effect on bus schedules is not reflected in arrival-time displays until the delay is already in progress. A lane closure on Connaught Road Central that begins at 10:00 will have scrambled the arrival times of every Citybus route serving Exchange Square by 10:12.

The operator cannot fix these in real time. A controller can cancel a trip and reallocate the bus to the return direction to restore headway. That leaves a gap at the outbound end. The published journey time remains unchanged even when the actual journey time has blown out by a third. Until the next schedule revision, the timetable is a fiction that the ETA system tries and fails to correct.

Check the operator app for the specific route, not just the government platform. KMB and Citybus apps show the bus position on a map. You can see if the vehicle is stuck at a tunnel portal or moving. A bus that is stationary a long distance away with an ETA of three minutes is not arriving in three minutes.

If you are at an interchange, look at the next bus behind the one you planned to take. On a bunched route, the second or third bus of a platoon will have a much longer ETA. It is the same vehicle cluster. Walk to an earlier stop on the same route, one that the platoon has not yet passed, and you may board before it fills.

When the arrival board shows "Due" and no bus appears within two minutes, do not wait longer than five. Cross-harbour routes that miss their slot often get held at the next timing point, which can be two kilometres away. Take a parallel route or walk to an MTR station. The bus network is dense enough that a Plan B is rarely more than a five-minute walk.