Plans that make sense in the physical world

We’re a team of machine learning specialists, software engineers, and experts in human-centric design.

From this experience, we’ve identified three key areas which need to be fixed.

We created this framework working with urban logistics, but the principles are broadly applicable to any operations-heavy industry.

Systems that understand the operation

Each plan is based on a combination of goals and constraints, ranging from cost and reliability to employee well-being and customer needs, which the planner has to evaluate on a case-by-case basis.

We chart the operation, then formalise it.

Working with domain experts, we map the rules that make up a real operation; what they are, how they interact, where the trade-offs sit.

We then define the mathematical representation that captures them, allowing state-of-the-art solvers to find the best plans for an operation as it actually runs.

Interfaces that empower the user

A domain expert will intuitively know a plan is wrong, but often struggle to explain their reasoning in a way their system can understand.

We capture knowledge through interaction, and design the interfaces that carry it.

We believe in empowering, not automating, experts, working with them to understand how they read problems.

Our interfaces let them explore many possible scenarios and the system tracks which ones are corrected, capturing tacit knowledge.

Plans that make sense in the real world

These systems can't be trained offline. The constraints which break plans are situational, flexible, and most have never been written down.

We learn inside real operations, alongside the people who run them.

The only way to reach that knowledge is to deploy and learn through interaction with a domain expert.

The system proposes and the expert decides; each correction sharpens its model and the expert acts as a safety layer while it learns.

Our findings are the product of our work in urban logistics, specifically in the shift away from vans towards light electric vehicles underway across European cities.

Moving things through a city is one of the hardest planning problems there is; the increase in mixed vehicle fleets and demand for fast delivery only makes it harder.

If our systems work here, they’ll work anywhere.

It also allows us to be at the forefront of building technology for greener, more humane cities.

Backed by evidence

We’ve analysed logistics data from and built representations of more than ten cities, shadowed dispatchers, and designed urban logistics models around new vehicle types, publishing our findings.

This research is the backbone of our development process.

It has informed OECD work and briefings to transport ministers.

Tested on the street

Parcelle puts our research and theory into practice.

It’s a last-mile planning app for mixed fleets of vans, cargo bikes, and LEVs, developed alongside the dispatchers that use it. Our goal with Parcelle is to enhance planning, while easing operators’ transition to more sustainable fleets.

The app is currently being rolled out in partnership with CARGOBIKE-SCALE, our EIT consortium, to operators in the UK, France, Belgium, and Spain.