Average road speed, city centre by city centre: comparing them fairly
Raw km/h flatters cities with wide roads. Fix the points, the local hour and the measure, then compare the share of free-flow speed each centre keeps.
Part of the guide Location intelligence automation
To work out the average road speed city centre by city centre and compare them fairly, measure the same kind of points in each centre, at the same local hour, on the same kind of day, and compare the share of free-flow speed the roads keep rather than the raw km/h. A centre where traffic moves at 22 km/h on streets built for 30 is in better shape than one moving at 30 on roads built for 60. Raw speed mostly tells you how the streets were designed; the ratio tells you how jammed they are.
This is for analysts at retailers, delivery firms, consultancies and city bodies who are asked questions like "is the centre of Sao Paulo slower than the centre of Mumbai at rush hour?" or "should our courier promise in Toronto match the one in Sydney?" It explains what a centre should mean in practice, the measures worth comparing, a worked example with invented numbers, the traps, and a flow that collects the readings for you. It sits under the wider guide to location intelligence automation.
Decide what the centre is before you measure it
Every city draws its centre differently, and the answer moves with the line you draw. A useful rule is to replace the boundary with a fixed set of points: for example the main railway station, the central business district's busiest junction, the main shopping street, the city hall, and four or five points about a kilometre out in each direction. Use the same recipe in every city, write the points down, and never change them mid-study.
Points matter because live speed is read around a place, not over a polygon. The Traffic Flow block measures the roads next to each point (about 150 m), within 500 m, or within 1 km, and averages them by the length of each road stretch. For a dense centre, 500 m around each point gives a fair slice of the grid without reaching the motorway on the edge.
Which number to compare
For each point, Traffic Flow returns the current speed and the free-flow speed in km/h, the jam factor from 0 to 10, a congestion percentage (how much slower than free flow) and a level in words. Of these, three suit a comparison between cities:
- Speed kept: current speed divided by free-flow speed. 0.8 means traffic moves at four fifths of its usual free pace. It cancels out road design.
- Jam factor: on a fixed 0 to 10 scale, so it reads the same in Nairobi as in London.
- Raw speed: still worth keeping, because a courier's arrival time depends on km/h, not on a ratio. Report it, but do not rank cities on it alone.
Report the median of the points in a centre, not the mean. One point beside a broken traffic light can drag a mean down; the middle value resists it.
Same local hour, same kind of day
08:30 in Toronto is not 08:30 in Mumbai. Schedule each city in its own time zone so every reading is taken at the local hour you chose. Then keep the kind of day equal: compare weekdays with weekdays, and drop public holidays, school holidays and big match days, which change a centre's traffic beyond recognition. The Schedule block can skip public holidays of a chosen country, which keeps those days out without anyone remembering to.
One day proves little. Collect at least three or four weeks of readings at the chosen hours before you write anything down, and keep weekends in a separate column.
A worked example with invented numbers
None of these figures is real; they show the arithmetic. Five centres, nine points each, read at 08:30 local time on weekdays for four weeks, medians shown.
| Centre | Median speed | Median free flow | Speed kept | Median jam factor |
|---|---|---|---|---|
| City A | 19 km/h | 28 km/h | 0.68 | 4.1 |
| City B | 27 km/h | 52 km/h | 0.52 | 5.6 |
| City C | 24 km/h | 31 km/h | 0.77 | 3.0 |
| City D | 14 km/h | 26 km/h | 0.54 | 5.4 |
| City E | 31 km/h | 44 km/h | 0.70 | 3.8 |
Invented readings, to show why raw speed and congestion rank cities differently.
Ranked on raw speed, City B looks second fastest. Ranked on speed kept, it is the most congested of the five: its wide roads should allow 52 km/h and deliver barely half. City D is slow on both counts. City C has modest raw speeds but keeps three quarters of its free pace, which suggests narrow streets that flow rather than a jam.
Traps that make comparisons unfair
- Different point recipes. Nine points around a station in one city and three around a park in another are not the same centre.
- Mixing radii. 150 m in one city and 1 km in another measure different roads.
- Treating closures as congestion. A closed road counts 10 in the jam factor. Note the closed stretches the block reports, and check the Traffic incidents block for what was shut and why.
- Low-confidence readings. Each reading carries a confidence from 0 to 1. Early morning streets with few vehicles give thin figures; drop the weakest before taking medians.
- Calling it traffic volume. Live speed does not count vehicles. A quiet centre and a busy one can both move at the limit. Hourly car counts exist only for roads in the United Arab Emirates, through Regional data.
Collect the readings with an AI employee
One flow per city keeps the time zones honest. Build it once, then make linked copies for the other cities so each copy keeps its own points and time zone.
- Schedule: weekdays at 08:30 and 17:30 in the city's own time zone, skipping its public holidays.
- My data: the table of the centre's points, each with a name and position.
- Traffic Flow: every road within 500 m of each point.
- Calculated field: speed_kept = speed_kmh / free_flow_kmh.
- Google Sheets: one row per point per run, with the city, the hour, the speeds, the jam factor and the confidence.
- Once a month, a second flow with the Data analyst block reads the sheet and writes the medians by city and hour, emailed as a short note.
What lands in the inbox is a monthly table you can defend, built from hundreds of readings rather than one afternoon. For a ready-made start, the templates Road speeds around the airport and Road speeds around the stadiums show the same measuring pattern around fixed places. City teams will find more on the government and cities page.
Questions
Why not just compare average speeds in km/h?
Because km/h mostly reflects how wide and fast the streets were built. The share of free-flow speed kept, or the jam factor, shows how congested they are, which is usually the real question.
How many points does a city centre need?
Eight to twelve fixed points, chosen by the same recipe in every city, give a stable median. Fewer points let one bad junction dominate.
Can I measure speeds over a whole district at once?
Traffic Flow reads roads around places, so a district is measured through a set of points spread across it. Places within 300 m of each other share one search.
Does the speed data include historical averages?
Readings are live at the moment the flow runs. History comes from running it on a timetable and keeping each reading in a sheet or database.