ab/Venture Labs← all ten stories
CASE 01 · AMAZON · LAST MILE, GEOSPATIAL & ML · 2020–2022

Tokyo Doesn't Have Parking

Density is not a penalty to absorb. It is a modelling problem with a specific answer — parking, transport mode, and the last hundred feet on foot.

10–15%
route volume lift
Bike + foot
modelled transport modes
Curb access
the binding constraint
Tokyo · NYC
where it was hardest
dense-metro route view
three park points · walking legs · drive between
THE FRAME
H1 · in the densest metros, route volume is limited by parking and the last hundred feet — not by the driver
test tokyo, nyc and comparable metros · alternative transport modes · 2020–2022
target absorb density instead of routing around it
result shipped · 10–15% route volume lift in the densest metros
THE BUSINESS CASE

A double-digit lift where a lift is impossible

Routing models are built on assumptions that hold almost everywhere: the van can stop near the door, the driver walks a short distance, the next stop is a drive away. Tokyo and New York violate all three. There is nowhere to park, the last hundred feet are on foot or by bike, and one badly sequenced stop costs twenty minutes.

The industry response is to treat those markets as structurally expensive and move on. That is a defensible read of the constraint and a bad read of the opportunity: these are also the highest-volume, highest-density postal codes in the network, so a small per-route gain compounds faster here than anywhere else.

The bar was not a percentage. It was whether operations would run the routes the model produced in cities where drivers already knew better than the software.

WHAT WE BUILT
01

Parking as a first-class input

Where a vehicle can actually stop — legally, at that hour — became part of route construction rather than something the driver resolves alone at the curb.

02

Mixed transport modes in one route

Bikes and on-foot legs stopped being exceptions handled offline and became modelled segments the optimizer could choose, with their own speed and capacity behaviour.

03

Walking sequence, not just driving order

In a tower or a dense block the expensive decision is the order of doors, not the order of streets. The model sequences the last hundred feet explicitly.

04

Explainability, because drivers knew better

Operations will not run a route they cannot interrogate. We shipped the reasoning alongside the recommendation and treated override rate as a product metric rather than a nuisance.

packages-per-route curve
bronze = post-launch series
walking sequence view
tower + dense block
WHAT TRANSFERS

Your worst-performing segment is usually a modelling gap

The mechanism was never the model. It was refusing to accept that a hard segment is structurally hard, naming the specific constraint the existing system ignores, and shipping the reasoning alongside the recommendation so the people doing the work would trust it.

That is the same sequence we run in a three-week thesis sprint and a ninety-day build pod — at a company with a fraction of the scale and none of the internal AI bench.

Which of your segments has everyone agreed to stop trying to fix?

Book 30 minutes →← back to all ten stories
ab/Venture Labs
est. 2026 · seattle, wa · hello@abventurelabs.com
Hypotheses, shipped.