Amazon’s New Waste Reduction Strategy: Batching

Amazon is trying something new in an effort to reduce packaging materials as well as the amount of fuel burned by making fewer deliveries. According to the article Amazon’s new waste reduction strategy: deliver only once a week, when Prime members enroll in the “Amazon Day” service, “Amazon will hold everything they order throughout the week and deliver the items together on the customer’s selected day.” In other words, Amazon is reducing waste by giving the customer the option to batch their weekly orders to arrive only once per week, instead of one at a time. This is fascinating.

On the one hand it is clear that combining a customer’s multiple orders from a week into a single weekly delivery to will reduce packaging waste. But in terms of reducing energy waste and improving carbon footprint, I am skeptical that this new Amazon Day service will be a positive contributor.

Batching is OK when it buffers an unavoidable transaction cost, such as capacity lost to changeovers or transportation, provided that said capacity is the most important factor. The total system cost of batching is often higher than the local cost of increased setups or deliveries. Lean aims for the best of both worlds by keeping batch sizes as small as possible while reducing said transaction costs.

Here are some scenarios on how Amazon Day may play out.

Success by design. In the best-case scenario, a combination of big data, massive computing power and smart people well-versed in how flow works in supply chains on continental scales have built Amazon Day around tested algorithms. If there was low-hanging fruit in the form of frequent high-energy-consumption deliveries to remote locations, weekly batching could make sense. Flying one small package per day to a remote dwelling in Alaska would use both more packaging and energy than delivering a week’s worth of orders only once weekly. Perhaps Amazon’s models show that the USA is a spiderweb of many small deliveries to remote areas, just waiting to be weekly-batched into efficiency. More than 80% of people in the USA live in urban areas. This number is just over half worldwide and increasing.

Convenience and cluster. Prime customers choose to receive their weekly deliveries when it is most convenient to them. This causes a weekly crunch when weekly deliveries cluster on late Saturday morning or Sunday afternoon, as opposed to work days. Packing costs are reduced, delivery is less smooth, costs increase with weekend pay, and supply-demand imbalance for drivers.

Incentivized smoothing. Amazon recognizes that they want delivery batches to be smoothed out across the transportation system, rather than clustered on days most convenient for humans. Amazon gives people incentives to spread out the weekly delivery days. This could be a shipping cost discounts for choosing Thursday morning instead of Saturday… except that shipping is already free on for Prime members.

Leveled pull from buffer. Level-loaded delivery in the Lean model (heijunka) relies on pulling from a finished goods stock or some controlled stocking point. A buffer is essential, in other words, to avoid the whiplash effect from needing to fill larg orders one day and none the next. Since Amazon delivers everything off the shelf, lead-time bullwhip is a non-issue. Intra-week fluctuation and burden on warehouse labor would be the main issue. It is unlikely that Amazon would pick orders in advance and keep them in a box until closing it up on the weekly ship date, as there would be half-empty boxes everywhere, much double and triple-handling. This is a logistical problem Amazon is well-equipped to handle. However, should USPS and UPS need to level their load to smooth out the spikes caused by weekly deliveries, their task would be a far greater challenge.

It’s a wash. Assuming customers’ choice of the once-weekly delivery days did not cluster, but was distributed evenly, there would be savings in terms of packaging materials but the energy used for delivery would be mostly a wash. The same number of items are being delivered on a weekly aggregate basis. The only difference is, for example, that instead of every house on your street receiving between zero and three boxes per day, each would receive between zero and ten only one day per week. Amazon would need to start batching by delivery route, such that a truck visits a region or neighborhood only once per week, for this to yield energy savings. With even the USPS delivering on Sundays, it seems unlikely that this is a horse that will go quietly back into the barn.

Reduced cardboard, increased chaos. In some parts, orders will cluster and delivery will spike on certain days. In others, there will be no change. In yet others, a blend. If it was possible to manage these different delivery behaviors as coherent customer groups, or value streams, processes could more readily be managed and improved. If tens of millions of individual human consumers are allowed the choice of batching or not, disruption to the routines of warehouse workers and delivery companies may be the reward.

Emptier trucks on some days. The 80/20 rule. There are just over 100 million Amazon Prime members who are eligible for this service. It is estimated that 80% of them are in the USA, representing 63% of Amazon’s US customers. That may seem like a lot, abut if Pareto’s law holds true, all it takes is a small minority of customers who insist on daily deliveries to wipe out any savings from batching. The orders of the 37% non-Prime members would still keep the delivery trucks on the road 7 days a week, even if every Prime member opted to batch weekly. Trucks would be two-thirds emptier on many days, perhaps burning less fuel.

Heijunka is when the producer takes advantage of their agility to delay the decision to commit resources toward delivering the customer’s specific order until the last possible moment. This is enabled by having low transaction costs, very short queues and few non-productive process delays. Amazon Day allows the Prime customers to set the weekly delivery date. This is in effect batches the pick date for one week worth of orders. Amazon risks trading demand spikes for picking in their warehouse for savings to packaging materials. I struggle to see how giving two-thirds of the customers in the USA market the option to batch their orders for delivery by week, based on whim rather than on a heijunka algorithm, will deliver a net energy savings.

I have so many questions about this, but alas not the access to Amazon’s big data, computing power or thinking behind this. I do know how flow works. I also know that no amount of data and computation will make up for faulty modeling. Amazon are smart people so let’s give them the benefit of the doubt. They must have done the research on the effects of batch size on supply chains, developed algorithms that show smoothed delivery loads, and a high probability of energy savings. But my suspicion is that as a business idea, this is more Fire than AWS.

10 Comments

  1. Bob Williams

    March 4, 2019 - 4:16 pm

    From the outside, the message is this move is not about internal efficiencies. It is about environmental stewardship and customer choice/flexibility. But they have to be trying to solve for internal efficiency and this move will put stress on their existing efficiencies. They probably expect to fine-tune and iterate based on customer response(sign-up) to the offering.

    • Jon Miller

      March 5, 2019 - 5:55 pm

      I agree. Amazon is always looking at data and optimizing, so taking this as an experiment in the direction of less fuel and packaging wasted, it is good.

  2. Jimmy Mendillo

    March 5, 2019 - 5:47 pm

    Batching seems like a great waste and fuel reduction strategy, but just how many people are ordering products weekly? As someone who only orders something once in a while with my prime account, It would be much more beneficial to get my order in the normal 2-day delivery window. If they found a way to organize by extremely frequent customers, this would be an incredible idea.

    • Ron Pereira

      March 5, 2019 - 5:52 pm

      Hi Jimmy, my wife is a power Amazon shopper. And by power I mean… darn near daily shipments. I will be interested to see how she reacts to this new option. My sense is she won’t want to wait but I could be wrong… especially if there are discounts awarded to those who choose to batch and receive on a specific day.

      • Jon Miller

        March 5, 2019 - 5:54 pm

        Weekly does seem like an arbitrary cutoff point. I am curious as to whether this is based on data or just a convenient calendar chunk.

  3. Bob Emiliani

    March 5, 2019 - 8:11 pm

    Where will the tens of thousands of batches be stored while awaiting delivery? Will I have my own personal rack in an amazon, UPS, FedEX, or USPS warehouse? Warehouse space is now at near-capacity. Will they build more warehouses just for “Amazon Day?”

    • Jon Miller

      March 5, 2019 - 8:13 pm

      Good question. They will probably wait until the last minute to begin picking your order. That way no additional rack space is needed, but if it is a pull of 5 items on Friday vs. 1 per day, it creates a workload spike…

      • Bob Emiliani

        March 12, 2019 - 9:20 am

        RE: “They will probably wait until the last minute to begin picking your order.” Doesn’t that assume that all parts of my order exist in one warehouse? Amazon has 75 fulfillment centers (and 25 sortation centers) in North America. It seems like daily orders would be spread across different fulfillment centers.

        • Jon Miller

          March 12, 2019 - 9:34 am

          Good question. If pieces of a batched order had to be consolidated from multiple warehouses into one weekly delivery box, it seems like this would add handling and transport effort.

          • Craig Woll

            March 12, 2019 - 11:37 am

            Amazon has been transshipping for years between buildings to combine orders into a single shipment. They also cannot store the order in a bin and hold it. The fulfillment centers are not designed to do this. They use random stow throughout the building so there are no designated bin spaces for specific items. Any item can sit in any bin. The software sends the picker to the nearest bin with that item or brings the robot to the picker with that item.