How Intelcom's package volume estimation project is improving delivery efficiency

URL copied!

Intelcom's tech team is constantly evolving and innovating; one of the latest optimization project is no exception.

Seeing an opportunity to further refine data gaps associated with certain package size, they turned their minds to ensuring drivers have the right amount of packages, sized correctly, to further reduce the friction in day-to-day operations.

They've called it the MS Package Estimation Project, and it aims to ensure no package gets left behind.

Share this article

At a Glance

The Opportunity: Inaccurate package volume data results in overloading or underusing delivery trucks.

The Approach: Account for missing data by utilizing alternate data points to estimate package volume.

The Result: Further optimizing our navigation algorithm (VRS) to get more accurate truck fills, which saves independent delivery contractors and clients time and money.

The Opportunity

Packages' cubic volume sometimes exceeds the capacity of the vehicles that independently contracted drivers use to deliver orders. This sizing disparity affects route optimization and, ultimately, delivery speed.

The reason these packages don’t fit is due to missing information about package volume. Until now, when no size information was provided, the operations team couldn’t account for the change and reassign the parcels to other vehicles.

Drivers have to leave these oversized packages at the station and go about their deliveries. If the packages don't get reassigned to another truck right away, the delivery gets pushed either later in the afternoon or the next day.

Emmanuel Motte, Data Scientist on the project, confirmed, "In most cases, 80% of the cases, we know the volume [of the packages]. In about 20% of the cases, we don't know the volume, and that's where we want to estimate."

"We know that there are certain pain points for the drivers at the station regarding the optimization." Michael Vu, Senior Product Manager, explains, "The way it works right now is a bit simplified. We can optimize the amount of packages they have, but we're missing a couple data points to help enhance it even further. One of those data points is understanding the volume inside the truck... We can assume that for one route there are a certain amount of packages that can be delivered, but we don't know how this fits in conjunction with the driver's trucks. It is possible that there are way bigger packages than the truck size itself."

Another variable in the mix is that not all drivers have the same sized vehicles: some are driving smaller 120 cubic-foot trucks, while others use larger vans. The idea here is to continue to improve the algorithm so that when the optimization team and the planners run it, it's smart enough to take the vehicle capacity into account as well.

"If they're driving and buying these larger trucks, and we're filling them up to half capacity, it's kind of sunk costs for them, right? We definitely want to capture that aspect and use case as well. The best runs are the ones with highest densities. Drivers end up making a higher profit for filling up their trucks; so we need to be able to do that for them, and that's one thing the project will also solve", said Michael.

The Approach

The plan was to produce a reliable estimate of package volume based on various inputs. The volume estimation for the 20% of missing data would then be integrated with Intelcom's proprietary algorithm for implementation in daily route optimization.

Michael said, "We've known this had some pain points from driver feedback on the operations side of the business, but also knowing that in the future this project could be leveraged for a lot of applications."

“

This is a key piece that would not only solve a problem that we have today, but in terms of where we want the optimization to grow in the future, it's creating that operational efficiency.

Michael Vu

Senior Product Manager

First, the tech team needed to streamline. They standardized the unit of measure used for all packages.

Step two was to use the client-provided information and convert it to the standard unit of measure: cubic feet. Oftentimes dimensions are provided for the packages when volume is missing, so the tech team needed to use those dimensions to calculate the volume in-house.

Next, they used the micro service. Jennie Jianying Liang, Business Analyst on the project, explained, "We make the micro service able to leverage and populate proper events. Currently this is where we are [in the project]. We are updating the micro service to fetch the information from Emmanuel's model."

Once the volume data is ready, we'll be able to integrate it with our systems to optimize routes accounting for volume.

The Result

Integrated into the algorithm, MS Volume will add more depth to the vehicle routing system, allow less opportunity for mistakes (e.g. lost packages or delivery delays), make for more efficient loading and create a better overall driver experience.

Michael talked about the impacts of this project, "The KPI is to reduce the amount of friction in situations where the packages the drivers are assigned don’t fit in the truck.”

Jennie shared, "For users, like drivers, they will easily witness a change because nowadays if they have small trucks, maybe we overload their small trucks, but in the future when we launch this project, the users can witness, they will no longer have packages overload their small trucks. Or if they have big trucks, maybe sometimes their big truck isn't fully utilized. It's a waste if a big truck goes out and drives with an empty truck."

By evaluating package volume, we can reduce stops and allow our Independent Delivery Contractors to deliver more in less time, increasing their efficiency. .