How efficient are Baku's bus routes? An open-data measurement and a fleet-neutral redesign

Baku's bus network of about 200 routes is being redesigned, but no peer-reviewed evaluation of it exists and the data needed for one (timetables, vehicle positions and origin–destination flows) is not published. This paper asks how much can be measured, and how far a redesign can be tested, from public data alone. From the transport agency's public map, a population grid, OpenStreetMap destinations, metro entries and one operator's daily boardings, I repair the route geometry, measure directness, duplication, detours, stop spacing and coverage, and build a door-to-door travel model whose gravity demand model ranks routes against observed boardings at a Spearman correlation of 0.58. Only 29 of 201 routes run within 1.4 times the straight line between their ends, and 66% of the 3.04 million residents live within 400 m of a stop. A greedy search that keeps today's fleet and applies the changes used by redesigns elsewhere shortens the average possible trip from 51.8 to 50.9 minutes (0.76 minutes with impossible trips counted at 120), or delivers today's modelled travel times with about 12% fewer buses than the estimated headways imply. On 10 independent trip samples the gain counted that way is 0.68 to 0.81 minutes, and under 14 alternative assumptions 0.46 to 0.88, the smallest without bus shifts; a stricter rule that strands nobody gains as much as the default. The pipeline, data and results are released to regenerate or, with a reader for another map format, apply to other cities that publish only a map.