In e-commerce and retail logistics, the “last mile” refers to the final geographical distance in the supply chain over which goods are delivered. Solving the last-mile challenge is an increasingly important task since e-commerce is the fastest growing customer channel in the retail industry. The cost associated with transporting goods over the last mile can range from 25 percent to more than 50 percent of overall transportation costs, so it can be a significant concern for shippers.

In today’s market, a company’s last-mile strategy also contributes to shaping the overall customer experience, which, in turn, affects the brand and influences repeat-purchase behavior. Developing effective and efficient last-mile delivery is absolutely essential to compete in today’s e-commerce environment.

Inspired by the rise of Uber in the past decade, a slew of startups emerged that sought to adapt the sharing economy business model from being centered on transporting people to transporting goods. There is now an entire class of business models that facilitates the movement of goods over the last mile as well as between upstream tiers of the supply chain.

In our paper, we used the term “crowdsourced logistics” to refer to those business models that recruit private individuals to serve as delivery agents in an e-commerce environment. Today, companies like GrubHub, Postmates, and UberEats have all become ubiquitous for on-demand delivery from restaurants. Big retailers like Walmart have experimented with crowdsourcing last-mile delivery, and Amazon is a clear leader in this space, having piloted AmazonFlex in 2015 in Seattle and New York City and now existing in over 40 U.S. cities.

If you’ve ever placed an order online and requested home delivery, you’ve been the recipient of a shipper’s last-mile strategy. If you’ve ordered food via an on-demand delivery service like GrubHub or Postmates, you’ve interacted with a crowdsourced driver. If you’ve bought something on Amazon recently and received notification of delivery along with a photo of where exactly the package was delivered at your home, then you’ve had a crowdsourced driver make your delivery. You can even go on websites like Instacart (and yes, Amazon Prime Now) to have groceries delivered to your home, when you want.

Companies are experimenting with crowdsourced logistics, especially for last-mile delivery, and as the true value continues to emerge, it’s likely to become more ubiquitous.

We wanted to understand how the prospect of crowdsourcing last-mile delivery would affect last-mile logistics. In other words, is it primarily a way of reducing cost, increasing responsiveness, or some combination of the two?

To answer this question, though, we needed to experiment with how crowdsourced logistics was being used and compare it to some baseline case. Unfortunately, not many companies that were experimenting with it in the real world were ready to share access to their driver networks with us, so we had to come up with a clever way of studying this phenomenon. We chose to do a stochastic simulation since we could have greater control over the system and ground it in empirical data as much as possible.

There’s a number of unique issues related to crowdsourcing logistics services that companies have to consider. For example, there is a degree of driver autonomy that doesn’t exist in dedicated (i.e. contracted) or privately owned delivery fleets. So, organizations seeking to crowdsource delivery should recognize that there is some uncertainty associated with those kinds of drivers — they decide how long they want to work and how often, which, in turn, impacts delivery capacity and could ultimately affect logistics performance and customer service. With this in mind, we chose to focus our attention in this study on the effect of the uncertainty in the supply of crowdsourced drivers emanating from their autonomy.

There were some interesting and somewhat surprising results that emerged. We expected to find that a dedicated fleet of delivery drivers would consistently be more effective than a crowdsourced fleet in terms of on-time delivery and total number of deliveries that could be made, which would have been a direct consequence of the gig-economy worker autonomy.

However, there were some scenarios in which the crowdsourced fleet outperformed the dedicated fleet. Namely, in scenarios when demand surged well above average levels, the crowdsourced fleet was actually making more deliveries than the dedicated fleet. This is because the dedicated fleet size is fixed, meaning that the delivery capacity is constrained. So, if demand for home deliveries exceeds that capacity, then those either become negative impacts on customer service or even lost sales. The crowdsourced fleet size is potentially unconstrained, so it appears that contracting gig-economy workers for delivery is a way of expanding delivery capacity, which can, in turn, increase responsiveness in the last mile.

The biggest takeaway is that crowdsourced logistics is not a replacement for dedicated or private last-mile delivery fleets. While there were a couple of scenarios in which dedicated was outperformed, it still had better effectiveness in the majority of situations we simulated. However, as consumers’ expectations continue to rise alongside the need for speedy, on-demand home delivery, sourcing extra delivery capacity from the crowd can be a means of increasing responsiveness and thus improving the customer experience.

The business model still needs refinement before we see widespread adoption by large retailers. It’s still not clear what the true value is, although it appears that the added capacity and responsiveness are the biggest benefits to be gained. However, the cost of crowdsourced delivery may be comparable or even higher than with dedicated fleets, but further research is needed on the underlying economics to answer this.

However, we’ve seen crowdsourced delivery become ubiquitous in the on-demand food delivery sector, so retailers are undoubtedly watching it closely. There’s also some public-policy uncertainty that needs to be worked out moving forward that can impact the long-term viability of the model. All in all, the “Uber for Logistics” business model has been steadily moving along the “technology hype cycle,” and it appears that crowdsourced delivery is moving up the “slope of enlightenment” and we’re starting to understand its role in logistics strategy.