#4 Research Meet-up: Cutting up to 100,000 taxi kms a day while improving passenger satisfaction

In this research meet-up, we looked at how analytics was used to improve long-distance assisted travel for elderly and disabled people in the Netherlands. A new nationwide taxi-planning algorithm reduced driving by up to 100,000 kilometres per day compared with planning rides separately within subcontracted regions, while also improving passenger satisfaction.

Why is planning assisted travel difficult?

Hundreds of thousands of Dutch citizens are eligible for subsidised long-distance taxi travel through the Valys mobility system.

On a regular day, the system handles around 5,000 rides. On Christmas Day, demand can rise to 15,000 rides. Trips are relatively long, averaging about 50 kilometres from origin to destination.

Planning these journeys is complex because operations are divided across dozens of subcontractors. Coordinating rides only within individual subcontracted regions can leave opportunities for combining journeys across regional boundaries unused.

Why is this a prescriptive analytics problem?

The challenge is not simply to predict how many people will travel. The system needs to decide which passengers can share rides, which vehicles should serve them and how those journeys should be combined across the country.

Transvision, the government’s main contractor and transport coordinator for Valys, organised a competitive challenge to develop better operational plans.

In response, a new taxi combination algorithm was created to plan rides nationally rather than restricting combinations to individual subcontracted regions.

What changed after the algorithm was implemented?

The new approach was implemented in practice and resulted in improved passenger satisfaction.

It also reduced the distance travelled by taxis by as much as 100,000 kilometres per day compared with plans that combine rides only within subcontracted regions.

For a service handling thousands of long-distance journeys each day, reducing unnecessary kilometres can mean less vehicle time on the road while making better use of the available transport capacity.

What does it take to use analytics in a large-scale mobility system?

In the webinar, we discussed both the impact of the algorithm and the practical considerations involved in designing analytics for a real-world service operating at national scale.

The model has also had to respond to changing travel conditions. The session explained how the algorithm was adapted when new travel requirements were introduced following the COVID-19 epidemic.

The case shows how prescriptive analytics can move beyond producing an efficient plan on paper: the solution also needs to remain usable as passenger needs, operating conditions and service requirements change.