Skip to content
Preview – data under validation

Consectetur adipiscing elit

à portée colours every census dissemination block in Canada by how long it takes to walk to things. That choice of shape — the block — was inherited, not chosen: it’s the geography the census hands you, and it’s what our travel times attach to. This spring we asked whether it’s the right shape, by routing a hexagonal grid over Hamilton and comparing.

The short answer: blocks are already fine in cities, and badly wrong in rural villages.

The experiment

We laid 71,242 H3 resolution-10 hexagons over the Hamilton census subdivision and routed every one of them the same way the production map routes blocks — real road network, slope-aware walking speed, 60-minute cap, all 18 amenity types. Same colours, same ceiling, same sentinels, rendered in the real app, so any visible difference is geometry and nothing else.

Hamilton, coloured by walk time to grocery and café — the production block surface.open in à portée →

Cities: blocks are already fine

Hamilton’s median block is 0.022 km² — about 149 m across, which is finer than a res-10 hex (132 m across). Downtown, hexes add nothing: the blocks are already smaller than the grid you’d replace them with.

Villages: the block lies

Rural blocks are enormous, and one number has to stand for everyone inside. The village of Lynden sits in a 5.45 km² block with 347 people. The block says the grocery store is a 30-minute walk. The 327 hexes inside that block say 30 to 60 minutes, with a median of 51:

The block understates the median resident’s walk by 21 minutes — not because the routing is wrong, but because one origin point has to speak for an area forty times the size of a downtown block.

Where Median block vs a 132 m hex
Downtown Hamilton ~149 m across already finer
Lynden (rural) 5.45 km² ~313 hexes per block

The number you’d publish if you didn’t check

One methodological trap worth recording: the shipped Hamilton surface caps walk times at 30 minutes. If you compare hexes against blocks without dropping the capped values, you get a mean error of −11 minutes and 66% of blocks “off” by more than 5 — dramatic, and fake, because the censored values manufacture disagreement. Drop them and the real figures are −2.7 minutes and 23%. The dramatic number is the one you’d publish if you didn’t check.

What this means for the map

The value of hexes is concentrated where blocks are big — which is exactly where Canada’s map is emptiest and least trustworthy today. The plan that came out of this spike is hybrid: keep blocks where they’re already finer than the grid, subdivide only the blocks that are too big. That work is now a real pipeline — plan, route, tile, verify — and it gets its own note when it ships.

← all posts