HOW PREDICTIVE ANALYTICS IMPROVES LAST MILE DELIVERY

Thanks to analytics methods, historical data can be leveraged to reduce the delivery window and improve estimates about when a package is going to arrive at its destination. For example, end consumers receive a message in the morning telling them when they can expect to receive their package, and then another message 15 minutes before it arrives, based on continuously updated estimates.

So how does this work? And what are the benefits for companies, logistics providers and, ultimately, consumers?

We spoke with Matthias Vollmert, Managing Director and Partner at DHL Consulting, and three of his team members who were involved in the live tracking project—Franziska Gleiche (Senior Consultant), Christin Schrörs (Project Manager) and Jens Kaiser (Project Manager)—about how predictive analytics is improving last-mile logistics to take it to the next level of efficiency.

What Was the Reason for Kicking off Such a Project?

Matthias: Deutsche Post DHL Group (DPDHL) Post & Parcel in Germany realized they had a lot of great data that they weren’t utilizing to its full extent. Also, honestly speaking, we saw that competition was a bit ahead of us in regard to this service and other industries already had pretty accurate real-time track-and-trace solutions available, such as Uber. So, together with us from DHL Consulting, and a team of data scientists from DPDHL’s Center of Excellence for Data Analytics, Post & Parcel formed a team to find new ways on how to improve our delivery service along the last mile.

So Tell Us About Predictive Analytics and How It Can Improve Last-mile Delivery

Christin: It comes down to effective analysis of the historical data we have in our possession, because it’s a trove of useful information. The idea is that by studying reoccurring patterns in past delivery data, we can predict what will happen with future deliveries. Of course, we adhere to data protection legislation at all times when it comes to the usage of data.

Jens: The most important data sets are “scan events” from past deliveries (when a package is scanned by the courier prior to delivery or at the point of delivery). We always had this type of data in our system, but weren’t using it for optimization. Now, based on analysis of these events—combined with information we have on individual delivery districts—we can predict when a package’s last-mile delivery will occur.

What Benefit Does This Have for End Consumers and Companies?

Franziska: It makes the whole delivery process more transparent, so anyone expecting a package has a much clearer idea of when it will arrive.

Matthias: When the end consumer is happy, the company benefits from their satisfaction, because this directly translates in less service complaints and increased loyalty. Also, when knowing exactly when a parcel will arrive and when the courier will ring the doorbell, the end consumer can better plan to be at home at that time. This, in turn, increases delivery efficiency and reduces the number of notifications.

And What About Logistics Providers?

Matthias: It comes down to efficiency. We’re currently seeing a few macrotrends in the logistics industry. First, peak seasons are getting busier and also volumes outside of traditional peak seasons remain very high as more and more people are shopping online. At the same time, delivery companies like DHL continue having to hire new people more often because this keeps on being an industry with low job retention.

Franziska: So with ever-increasing volumes and newer delivery staff who don’t necessarily have experience of different delivery routes, it makes sense to support them with tour sequencing technology.

What Exactly Do You Mean by Tour Sequencing?

Franziska: It refers to the most logical order for packages to be delivered along a particular delivery route. When an experienced delivery driver loads packages into their van, they might map out an optimal delivery route in their head. But if they don’t, the technology now does it for them. A learning algorithm that was developed together with data scientists from DPDHL’s Center of Excellence for Data Analytics, generates a tour by analyzing historical driving sequences, taking into account the specific number of shipments and destinations to deliver to on that respective day.

Christin: Looking back at the 2020 Christmas season, which turned out to be the biggest peak season in history, this innovation especially helped our temporary or seasonal employees who support our permanent staff.

Still, Are There Challenges in Predicting When a Package Will Arrive?

Christin: Yes, there are a number of hard-to-predict factors, such as the amount of traffic on the roads, what the weather is like, routing to different delivery stops, opening times of shops, and if the recipient is home or not.

Matthias: Additionally, human behavior is another factor in this that shouldn’t be underestimated. Just one example is to take into account when our couriers will take their well-deserved breaks. Even if they personally like to stick to a routine, they may also depart from this in order to stay flexible and deliver the best service to our end consumers.

Franziska: There’s a lot of trial and error involved when it comes to tweaking the predictive model so that it can respond to these types of data outliers. You can’t say, “I’m facing hard-to-predict factor X, so I’ll pull lever Y and that will solve the problem.” There will never be the perfect predictive model for every single delivery.

Matthias: Yes, that’s right. It’s not about having the perfect predictive model, but about reaping the benefits of the data available to create a real impact for all stakeholders along the value chain. Leveraging the huge amounts of data is a challenge for most industry players, but this is exactly where the potential lies and needs to be unlocked. By tapping into predictive analysis and investing in its capabilities, companies can maintain a competitive edge in a landscape where the expectations of end consumers are higher than ever.

Were There Any Fears From Couriers About the Implementation of This Technology?

Jens: At first there was a concern that it could be a stealthy way to track a driver’s every move. You know: “Where are they now?”, “Why are they taking a break there?” and so on. And it did take a while to dismantle those fears.

Matthias: While we developed a transparent solution for our business customers, we also worked closely together with the workers’ council to ensure that Post & Parcel couriers feel comfortable with the solution. Training and upskilling certainly helped to get the buy-in needed.

Could Predictive Analytics Revolutionize the Logistics and Supply Chain Industry?

Jens: It’s more about evolution than revolution as far as we’re concerned. Having said that, I think it is a transformative service for the business customer.

Franziska: It’s also about keeping up with the times because the world has changed. Both our end consumers and our employees are used to the constant accessibility of information. For example, our delivery staff use navigation devices in their private lives, so it’s really hard to justify why their employer can’t give them a good navigation and routing system to help with their deliveries.

Matthias: The next step could be to combine predictive analytics with real-time technology, so we can respond to ad-hoc delivery changes. In the future, this could allow end consumers to interact with us and say, “I thought I’d be at home, but I am somewhere else now—can you please deliver my package there?” Of course, logistics providers will need to keep an eye on the implications of such changes for their operations and strike the balance between end consumer convenience and operational efficiency.

What About Companies in the Logistics and Supply Chain Industry and the Data They Have? Can Dhl Consulting Also Work With This Data for Predictive Analysis?

Christin: Many companies have this kind of data in their system, but don’t unlock its full potential. I think this project proves that it’s definitely worth exploring, because if it’s analyzed and used in the right way it can help increase consumer satisfaction. It’s like a hidden treasure.

Franziska: The trick is to collaborate with experts in the predictive analytics field, and use their knowledge and experience to provide a solution that staff find accessible and usable. It has to be acceptable to the people on the ground who are using it—because if it isn’t, it won’t get you anywhere.

Matthias: And DHL Consulting has the experts in the predictive analytics field ready to support business customers in boosting their own operations. In this solution, we’ve focused on last-mile delivery, but we are also heavily engaged in better prediction models across the entire value chain. This means predictive analysis has the potential to be a true game changer when soundly implemented. It will definitely play a leading role in boosting resilience to changing demands and building up agility for the logistics leaders of tomorrow.