Research Highlight: Further investigation of last-mile grocery delivery in a suburban environment

The objective of this investigation was to map the process that packaged goods go through, as well as any potential hazards that may occur from the moment they are picked up from the shelf to the moment they arrive at a consumer’s doorstep. Looking at the existing studies into last mile grocery delivery services, a gap was identified in research regarding the potential effects of delivery on consumer-packaged goods.

Grocery delivery has been showing increasing prevalence and interest in recent years. Trends from a 2014 study showed that 25% of respondents in a global survey had already used online grocery delivery services and that 55% of respondents were willing to try it in the future. A more-recent 2019 survey showed 31% of respondents engaged regularly in online grocery shopping. An estimated one-third of the national grocery market now utilizes online delivery services, so investigating last mile grocery delivery will reflect the probable outcome for a large share of grocery packaging.

The increased handling experienced by packages being delivered presents a variety of potential hazards that need to be considered because these hazards may yield damaged, spoilt, or otherwise compromised products. Most grocery delivery services today are carried out in consumer vehicles which are built to provide a smooth and steady ride compared to delivery trucks. Since the vibration level of a consumer vehicle is significantly lower, for purposes of this investigation, shock hazards were the primary focus. This shock data will aid in simulating high-risk situations such as sudden stops, driving over potholes, sharp turns and maneuvers, as well as any drop scenarios that could compromise the packaged goods. Proper accounting of these factors will help to evaluate whether current packaging standards are sufficient to guarantee acceptable conditions during grocery delivery.

The mapping process was discovered through a series of interviews with grocery retail staff as well as independent drivers for the third-party services in order to develop a vision of the process both in theory and practice. Based on interviews conducted, we found that there are several different ways a delivery may be conducted based on the drivers’ preferences. Half of the drivers interviewed stated that they always put the groceries in the backseat of the car, whereas the other half said they always put them in the trunk. Most drivers indicated that they use self-checkout most of the time. Most delivery drivers said that when in transit they drive normally or more carefully to prevent products from falling out of bags.

After gathering sufficient information regarding the handling processes from different grocery delivery services, we were able create an outline for the general procedure for a delivery route. This outline was then used to create grocery delivery simulations.

The shock data collected for this experiment was taken using a decoy package containing a Lansmont Saver 3D15 sensor housed in a corrugated container. To account for an average sized product that is found in the grocery store, a 3.43” x 6.43” x 5.25” corrugated box was used as a simulated grocery item. The box held an aluminum and wooden frame (Image 1 & Image 2) with the Saver 3D15 attached in it. The Saver 3D15 can record the duration of shock impacts as they occur either as timer-triggered or signal-triggered events. Dummy packages were also used to ensure the volunteers would not handle the decoy packages in a way that may skew the data.

Using these packages, 18 simulated events, and 25 repetitions, were conducted using volunteers. The events simulated the grocery delivery process through four segments: picking, checkout, loading, and delivery. Picking was simulated with the volunteers choosing packages from various shelf heights and dropping them into shopping carts. Both self-checkout and regular belt-driven checkout systems were used. Then, using their personal vehicles, volunteers were instructed to put the groceries, with the sensor and dummy boxes, into either the cabin or trunk of their vehicle at random. Then they “delivered” the packages to various surfaces, such as concrete, wood, and welcome mats. The volunteers were instructed to conduct all of these trials at two different paces: normal and expedited. The purpose of this was to simulate a grocery delivery driver that would either be rushing to complete the order to taking their time to complete the order.

During picking, we saw that most of the drops were in the flat orientation because the volunteers would drop the packages holding the box by its sides, straight down into the cart, not allowing for the orientation to be manipulated. During checkout, most of the drop orientations was either on the edge or flat orientations because the volunteers would attempt to place the box in the grocery bag on the flat orientation, but because of the creases in the bags, the box would reorient itself onto an edge.  During loading, there was an equal distribution between the orientation types because as the volunteers would put the bag into their vehicle, they would often let go before the bag landed. This allowed the boxes to reorient while in free-fall, resulting in random drop orientations (Image 3).

Through the data, it was determined that the loading process has the highest potential to damage the products due to the increased drop heights. Most drops occurred between 0.89 in. and 4.09 in. with an average drop height of 3.59 in. and a COV of 70.25%. Compared to the ISTA 6a standard, last-mile grocery delivery has fewer overall hazards than e-commerce shipping.

Due to the lack of secondary packaging, we can assume that consumers and drivers handle the groceries more gently. The basic finding of this project was that primary packaging which was designed for e-commerce or commercial shipping, is overdesigned for last-mile grocery delivery services. A redesign of primary packaging may open the opportunity for cost savings and more sustainable designs.

In the future, we hope to increase the representativeness of the trials by refining the simulations and increasing the data samples. The approach moving forward consists of four primary goals: developing a decoy package with reduced volume, conducting simulations with the help of store employees, sending mass surveys to delivery drivers, and conducting simulations of grocery handling with the help of student volunteers.

Corporate approval is required to advance the investigation into the last-mile distribution environment for major grocery retailers. Until then, we are unable to work with local stores to run simulations or investigate their pickers’ behavior. As things are currently, future simulations will either mimic store conditions as closely as possible, or the scope will be brought in to focus solely on third party delivery services.