Planning parcel tours with deliveries and pickups in time windows
Which van delivers and picks up which parcels in which order, so that every customer is served in the booked time window, no van is ever over capacity and the day ends early?
Drag to rotate, right-drag or shift-drag to move the map, scroll or pinch to zoom. Each van shows how full it is: deliveries empty the gauge, pickups fill it. A pin turns green when its stop is served in the window and red when it is late. Hover a pin, a van or a bar for details. Plan computed with a Timefold model.
The planning question
A fictional delivery base in a rural area sends out six vans on a Tuesday: four diesel vans with 4,800 litres of cargo space for parcels and 900 kg payload, and two e-vans with 3,600 litres, 650 kg and a usable range of 110 km. The day has 378 stops. 270 deliveries go to homes, about one in five with a booked 2-hour window, the others any time. 44 deliveries go to businesses, which take parcels during their opening hours from 09:00 to 16:00. On the same tours the vans pick up 36 returns at homes and the outgoing parcels of 28 shops and small online sellers, 5 to 26 parcels each, in windows from 11:00 to 13:00, 13:00 to 15:00 or 15:00 to 17:00.
The drivers start at 07:30 and load until 08:00. A van leaves with every parcel of its deliveries on board. Each delivery makes room, and each pickup takes up space again. Cargo space (litres) and payload (kg) must hold at every point of the tour. Assumptions for this example: a stop takes 2.4 to 6 minutes (walking, ringing, scanning, a neighbour when nobody is at home), vans drive 12 % slower than cars, and the break of 45 minutes starts after the first stop reached after 11:30. Every van is back at the base by 17:30.
How we plan it
The Timefold model plans each van as one list of stops. For every change, the tour is simulated on the real road network: the van leaves the base at 08:00 fully loaded with its deliveries, drives to each stop, waits when a window has not opened yet, takes the parcels off or on, and the driver takes the break. The load is checked at the departure and after every stop. A stop after its window counts by minutes. Driver time from 07:30 until the van is back, kilometres, waiting at the curb and the end of the last tour are costs.
The comparison is a common way of working, computed with the same simulation. Each van has a fixed district, sized so that every van has the same work on an average day. The driver delivers in sorting order (the nearest next stop), leaves a stop with a window that has not opened for later, and collects the pickups at the end of the tour. Parcels that do not fit go to the van with the most room left. The Timefold plan starts from an empty plan and improves it for 5 minutes.
What the plan shows
| KPI | Optimized plan | Comparison: fixed districts, sorting order, pickups at the end |
|---|---|---|
| stops in their window | 100 % +4.8 | 95.2 % |
| driver hours | 35.4 h −18.8 | 54.2 h |
| last van back | 15:55 | 16:55 |
Decisions
- van for every delivery and pickup
- order of the stops on each tour, and with it the arrival times and the load at every point
Constraints
- time windows: booked 2-hour windows at homes, opening hours of businesses, pickup windows of shops and online sellers
- cargo space and payload at every point of the tour: deliveries make room, pickups fill the van
- driver shift from 07:30, back at the base by 17:30, with a 45-minute break
Objective
every stop in its window, few driver hours and kilometres, the last van back early
Questions
Where do the data come from?
The roads are real (OpenStreetMap). The delivery base, the 378 stops, the parcels, the windows and the vans are generated and plausible. In a project, the stops come from the sorting system and the shipment data, the windows from the booking or the customer master data.
Why are stops late in the comparison?
The pickups wait for the end of the tour. By then, 14 of the 28 business pickups have missed their window, most of them at midday. 4 booked deliveries are late as well. The vans also wait 17 hours in total at the curb for windows to open. The plan serves all 378 stops in their windows and waits 3 hours in total.
Why does one van stay at the base?
The district plan sends out all six vans. The plan needs five, and one of them only drives a short morning tour of 21 stops with early windows and is back at 09:35. Together the drivers work 35.4 hours instead of 54.2, and the vans drive 185 km instead of 297. Four vans leave the base full to the last litre, and the pickups fill them up again in the afternoon.
Can it plan again during the day?
Yes. Stops already served stay fixed. A new pickup, a van that breaks down or a road that is closed are planned in with the same rules.
More use cases
Your planning problem looks different?
Every operation has its own rules. Tell us about yours. We assess feasibility and expected impact in a free initial call.
Book a free call Write to us