‘It is more sustainable to physically go to the store instead of shopping online.’ According to an insight from the organisation TNO, this statement is a myth [1]. Although it is used a lot, even in parliament, the question of which option is the most sustainable is not that straightforward. Which is the better option, is very dependent on the situation. For example, it depends on where you live, the means of transportation you use to get to a store, and whether you are home to receive the parcel or not. In many cases, online shopping might be the more sustainable alternative. This notion is also supported by another joined research between the Hogeschool van Amsterdam and TNO [2].

### More flexible deliveries

So, instead of completely writing off one of the options because it is supposedly more polluting, maybe we should focus on reducing their carbon footprints. That is why I have researched the effect of a more flexible delivery policy. More specifically, it compares a multi-day delivery approach with the one-day approach. In the one-day approach, all parcels have a one-day window in which they are allowed to be delivered. This is often the default approach for retailers, even though delivering one day later does not always matter much from the customer’s perspective. For example, when ordering a present for a birthday you have in a week, it does not matter whether it is delivered tomorrow or the day after.

This flexibility is captured in the multi-day approach, in which parcels may be delivered on any day within a period of a specified number of days. For the purpose of this research, the multi-day approach was defined such that all parcels must be delivered within a period of two days, though the same model can be applied for varying configurations. One might expect that allowing for this flexibility gives rise to more efficient routing of the delivery vehicles. A reduced total travel distance of the vehicles would result in a decrease in cost and carbon footprint of deliveries.

### Multi-day approach

The intuitive explanation for this twofold effect is that, for example, a delivery truck does not need to visit the same street on two consecutive days, therefore reducing the total distance traveled. Let’s say that you order something online for delivery on Monday but your neighbour orders something for delivery on Tuesday. Under the one-day approach, that would mean that a delivery truck has to visit your street twice in two days. However, by allowing for multi-day deliveries, both parcels can be delivered on Tuesday in one visit, which is evidently more efficient.

This idea is also illustrated in the figure below. Under the one-day approach (a), parcels 1 to 4 must be delivered on day 0, even though parcel 4 is to be delivered to a location much closer to the parcels delivered on day 1. With a multi-day approach (b), parcel 4 may be delivered on day 1 instead, together with parcels 5 to 8, which are to be delivered in the same area. This leads to a clear increase in efficiency.

### Parcel Selection Procedure

The routing challenge delivery companies face each day consists of two key components. First, they must decide which parcels to deliver today and which to postpone. Secondly, the actual planning of the routes must be done to deliver all selected parcels. The construction of efficient routes, once a selection of the parcels has been made, is a much-studied problem. Thus, this research focusses on the parcel selection procedure. The question to be answered is how to decide on which parcels to postpone.

The solution strategy proposed in this research is inspired by algorithms that are proven to perform well in similar problems. It uses several measures of efficiency to decide which parcels to delay. A parcel is said to be inefficient if it does not combine well with other parcels on the corresponding day. If a parcel scores higher on efficiency, it is given a higher priority to be delivered on that day. This way, if a parcel does not combine well with other parcels on a particular day, it is more likely to be postponed to a later day in hopes of a better fit with the parcels on a new day.

### Decreased travel distance

One evident complexity is the uncertainty of the future because it is unknown how many parcels will need to be delivered tomorrow and, thus, how many parcels can be delayed till tomorrow without exceeding the capacity. This is accounted for by making an approximation of the set of parcels that will have to be delivered on future days. This is done by dividing the total service area into neighbourhoods. Each neighbourhood is then represented by one order, of which the size is based on historic data from that neighbourhood.

Without knowledge of the future, the algorithm attained an average reduction of 3.40% of the total distance traveled by delivery vehicles over a period of five days. This effect goes hand in hand with a decrease in CO_{2} emissions and costs. The uncertainty does not have a significant impact on the outcome, as giving the model full prior knowledge on all days in the planning period only increases the reduction of total distance to an average of 4.47%. Therefore, the model is still viable under uncertainty.

To conclude: by adopting a more flexible delivery policy, such as the multi-day approach, online retailers can decrease both total cost and their carbon footprint. Besides following potential sustainability regulations, the reduced emissions also make the retailer more attractive to the increasingly more sustainability conscious consumers, therefore, appealing to a larger market share. Thus, as stated in a research report by Descartes [3], “sustainability is not a challenge, it’s an opportunity.”

Lars van der Meer

Doing the Math

[1] 5 mythes over de thuislevereconomie | TNO

[2] Retail Insiders | De mobiliteitseffecten en carbon footprint van online winkelen

[3] Retailers: Sustainability is Not a Challenge, It’s an Opportunity | descartes.com