Interview Transcript

Disclaimer: This interview is for informational purposes only and should not be relied upon as a basis for investment decisions. In Practise is an independent publisher and all opinions expressed by guests are solely their own opinions and do not reflect the opinion of In Practise.

I am interested in both Amazon and Wayfair for different reasons. Amazon is the dominant player in the market. It simplifies commerce for everyone. Initially, it was a product-based business, but now it seems to be more of a service provider for sellers, offering cloud computing infrastructure, fulfillment infrastructure, and logistics. Amazon excels at scaling these services and solving complex business problems for smaller businesses. For example, my wife runs a brand and we've had to navigate through logistics, supply chain, ocean freight, port drayage, warehousing, product distribution, and shipping. It's a complex process for a small business. Amazon simplifies this process if you use their infrastructure. I'm interested in understanding how this fulfillment infrastructure has evolved, how standards are created and implemented, how it's continuously improving, and how it compares to previous years. I'm also curious about how you foresee its evolution.

The second aspect is related to Wayfair. I'm interested in the furniture market and the handling of large, bulky, odd-sized items. These items don't fit well into the general merchandising infrastructure. I've visited Amazon's fulfillment center in LA where robots move around a grid and everything fits into a small cubby. Humans are involved in picking and packing, but these are relatively small items. I'm not sure what Amazon's infrastructure for larger items like couches or tables looks like. Furniture and bulky items are notoriously difficult to handle in fulfillment. There's a high rate of breakage and damage. The more these items are handled, the higher the damage rate. I'm curious about how a company as large as Amazon has addressed this issue. Wayfair has its own distribution centers and has realized the need to control more of the middle and last mile to reduce damage rates and improve customer satisfaction. I'm interested in how the largest player in the market handles this. These are the two areas I'd like to learn more about.

I'll provide some background information about myself, which might help you ask more relevant questions. I'll primarily focus on my experience with Amazon. I started in 2013 in India, which is relevant because I was part of the initial business setup for Amazon in India. I served as the general manager for the first Fulfillment Center (FC) launched there.

Over time, my role evolved, and I started managing multiple FCs for Amazon in India. After some time, I moved into the middle mile network, which encompasses the entire transportation network you're referring to, including our line haul, air haul, and sortation centers. I spent four years in India, during which I was part of launching the heavy bulky business from scratch. Interestingly, this aspect of the business evolved very quickly and is more advanced than what we have in the US.

So, the Indian operation is more advanced in handling large, bulky items than the US businesses?

Yes, that's correct.

That's interesting.

In 2017, I moved to the US with the task of starting the last mile network for delivering these heavy, bulky items to our customers. Until then, we had only relied on third-party carriers who specialize in heavy, bulky items.

The transition began at the end of 2017 when we launched our first three last mile delivery stations. This marked the beginning of my journey into last mile delivery for heavy, bulky items. Over two years, I expanded that network to cover all the key metro locations. After that, I moved to running FC operations.

Initially, we operated almost entirely through third-party FCs because we believed they were cheaper and more experienced. We only had five FCs at that point. However, we started improving our efficiencies and eventually became cheaper than third-party carriers. This led us to close the third-party FCs and build more of our own. In two years, we increased our FC count from five to 20, and now we have over 30.

Just to clarify, you're referring to the large and bulky FCs, right? This is a separate network?

Yes, exactly. We have over 200 FCs in total.


These large FCs are similar to what you've seen with Wayfair. They primarily consist of racking and a lot of PIT equipment. That's the setup for all these large FCs.

Now, I'm with Walmart where my role is very different. I run the last mile, but since you're more interested in heavy, bulky items, FC operations, and last mile operations with Amazon, I can focus more on those areas.

Just to clarify, you started in India, but did you start with general merchandise or was it always large and bulky items?

I was responsible for the operations of large and bulky items, but my overall responsibility extended to the middle mile and first mile FCs for general merchandise. Essentially, I was overseeing the whole business.

So, managing large and bulky items was just a subset of the entire business you were overseeing. Let's start with discussing the US operations. Later, I'd like to understand why India's operations are more efficient and evolved than the US. I'm assuming it's because they were built from the ground up at a more advanced stage, but I'd love to hear your thoughts on that. In the US, can you explain what drove the internal decisions to transition from third-party carriers? Were you using carriers like FedEx and UPS for delivery? Or were you using Less Than Truckload (LTL) or truckload carriers for the middle and last mile? What was the setup when you were primarily using third-party carriers?

Yes, FedEx and UPS were part of our carrier network, but we also used other carriers who specialize in handling large items. When dealing with oversized items weighing more than 200 to 250 pounds, these carriers are better equipped than FedEx or UPS. So, we had multiple carriers involved in the deliveries.

And did you also use these carriers for storing the items, or where was the inventory held?

The inventory is held in third-party fulfillment centers, or warehouses. We have contracts with multiple companies that operate these independent warehouses. They hold our inventory and deliver it to the carriers that we appoint. These carriers then deliver the products to our customers. We pay the warehouses for handling charges for each product they handle.

It's important to note what we experienced during this phase.

When we evaluated the costs, we found that these third-party warehouses were cheaper in terms of both variable and fixed costs, especially when we were operating only five fulfillment centers. However, when we factored in the costs associated with damaged goods, the expenses were significantly higher. One critical aspect of handling large, heavy, bulky items is how they are managed in your system and throughout your supply chain. This led us to consider moving to our own infrastructure.

In 2018, we began to consider investing more in this area. At that time, heavy and bulky items, along with groceries, were the only two areas where Amazon did not dominate the online retail market. Jeff Bezos was closely involved in this process. The decision was made to purchase Whole Foods to enter the grocery market and to invest in the heavy and bulky category to expand our offerings.

We didn't anticipate the significant returns this would generate. Previously, customers would go to Best Buy for their TVs or Home Depot for other appliances. However, in 2020, this changed dramatically. Our investment in this infrastructure allowed us to meet the changing needs of our customers.

By 2021, we were not only investing in fulfillment centers but also building our last-mile and middle-mile networks. This gave us full control over the supply chain from end to end. We found that even if we could ship out from the fulfillment centers, we were still failing to meet customer expectations during the middle-mile transfers. Many items were damaged during this process because not every carrier was equipped to handle these special items correctly.

In the heavy and bulky category, handling a couch is very different from handling a refrigerator or a treadmill, or a ping pong table, which is also a popular item. Once we started creating and building our last-mile network, we gained the ability to properly handle these items. We even acquired our own trucks and began to figure out how to set up each item for the last mile.

The third crucial aspect is servicing these items. If this part is broken, where we deliver the item to the customer but can't install it or offer support, it creates a disjointed experience for the customer. They would then have to seek help elsewhere. However, by building our last-mile network, we were able to provide these services, which was previously inefficient for the end customer.

To summarize, in the early stages when there was not much investment in this area, third-party fulfillment centers (FCs) or 3PLs were more cost-effective from a fixed and variable cost perspective. However, you realized there were additional costs to consider, such as damage rates, returns, reverse logistics, and customer needs due to supply chain imperfections.

As a result, you began to invest and take control of the entire system, optimizing for heavy and bulky items, which reduced damage rates. I presume that as you expanded the FCs and capabilities, and put all the necessary assets in place, your variable costs dropped below those of the third parties.

I'm not sure about the scale of investment required to achieve this, but I assume it happened relatively quickly. It's likely that your fixed cost per unit also decreased compared to the third parties. Could you tell us where you think you ended up relative to the third-party costs of doing this, considering the entire end-to-end process, including the netting back of returns for damages and such?

Yes, you're correct. Both our fixed and variable costs improved. In fact, the fixed cost is largely dependent on how quickly you rotate your inventory in your warehouse. This is particularly important for heavy, bulky items because of the large space they occupy, which significantly impacts the fixed cost.

The variable cost, on the other hand, is a function of efficiency, which we improved through automation. Another critical factor in reducing our variable cost was the layout of the FCs. We completely transformed the setup. Unlike Amazon FCs, which are typically unidirectional with docks only on one side, we designed ours to be bidirectional. This reduced the movement of large items, which is costly, and allowed items to come in one side, get racked, and move out from the other side.

This might seem simple, but it required a complete transition, including having trucking yards and doors on both sides. Despite the design change, we were able to improve our efficiencies. I won't quote exact figures, but within two and a half years, our variable costs were lower than those of third-party carriers. Although we had made significant investments, which meant our fixed costs were higher during that period, they will pay off over time.

At that point, our variable cost was at least 10% lower than the third parties, while our fixed cost was about 10% higher than the third-party carriers.

The scale does increase over time, so it's not a significant concern. If you were to expand from five to 20, I understand that Amazon has recently regionalized their FC footprint. Jassy has mentioned this. I'm uncertain how large and bulky or heavy bulky items fit into this model. What do you believe the ultimate, albeit ever-changing, mature footprint would need to be for this heavy, bulky infrastructure? Would it be 50, 100, or 200 before it begins to grow with the business instead of investing to reach full scale for this operation? Or has it already reached that point?

It has indeed already reached that point. We were still transporting items from a longer distance, but we had regional FCs set up, even with 20 FCs at that time. However, what's more important than having a regional footprint is a different matter. Regardless of the size of FCs you can build for these heavy, bulky or large items, it will still not be enough, unlike with smaller items where you can hold much more inventory.

What's even more crucial for larger items is to ensure that you know your fast-moving goods and make sure that those are available in all your regional locations. You also need to consider the long tail, which you still need to keep in your warehouses. You could leave it with the sellers or the merchants where you can drop ship, and they can send it out directly. Alternatively, you could decide on the amount of long tail you keep in your warehouse, but you would keep it in a more central location, which is at a lower cost.

While the transportation time would be longer, the cost of keeping it for an extended period will compensate for it. The fast-moving items are where you are making business and profit, and those are available in your region. The closer you are to the customer, the lower your handling cost and the fewer damages you'll have. The number of times you have to shift or transfer an item to maintain cost efficiency will impact your damages to that extent. Thus, having the right balance is very critical in this part of the business.

This is quite intriguing as it involves the intersection of merchandising and what you display on the website, along with the speed badge, such as Prime. It's a balance between having a wide selection and maintaining a good service level so that customers aren't waiting for three months to receive their delivery.

For instance, if you're Amazon, you probably have a good idea of how much toilet paper or paper towels will be sold within a certain period due to their consumable nature. However, for large and bulky furniture, which is purchased less frequently and has stylistic elements that can fluctuate with trends, it's harder to predict sales. This affects inventory management, storage, and ultimately the service level for these items, making it much more complex.

How did you coordinate with the front-end merchandising team? For example, if you decided to stock 200 items in every fulfillment center because they are fast-selling, would those items be displayed first on the website within their categories? Was there a coordinated effort between the merchandising teams and the infrastructure or supply chain team? This seems like a unique category where more coordination is required.

I find your question amusing because it's not just about coordination. It's more like a struggle for space for these items. It's not just about fast-moving items, but fast-moving items for a season. This is what makes this space more challenging. For instance, treadmills and grills are not going to sell in the summer. They sell a lot more in the winter. So, how do you manage that inventory throughout the year? You might have a kayak or a canoe stored in your warehouse, but it won't sell during the winter season.

That's fascinating. The large and bulky category has so many different items, and there's a lot of variance in demand patterns and how that affects inventory management. It's quite challenging. How would you assess the profitability of this category? It seems much harder to manage than something like paper towels.


Do the challenges make the margins better or worse?

That's an insightful question. Let me rephrase it to help you understand the nuances involved. This is also where your competitive advantage lies. If you can figure this out, you gain a significant competitive edge over dealing with minor issues, which can keep changing and won't consume much of your space.

For instance, consider a canoe you have. The storage rack for a canoe is vastly different from a rack where you can store a mattress. The challenge is that different seasons require different items. We had to determine how to create a flexible racking system to hold multiple items together and identify which items should be kept in certain types of racks. If not done correctly, you risk damaging the items in your warehouse.

These are the nuances you deal with in this space. However, this is also your competitive advantage because you are innovating in all these areas, allowing for the flexibility of items that come into your warehouse with the same design. The cube's size won't change. You have to use the same cube but manage different items. This is a function of how well you utilize your racking. If not done correctly, this is where you lose on cost. For example, you have the same cube, but some of the racks are only 30% utilized during a certain part of the season, while at other times, they are fully utilized, and other racks are left vacant. This increases your fixed cost.

If you can manage this efficiently, that's what will give you the competitive advantage you are ultimately seeking, correct?

Did Amazon have to innovate around racking systems to make them more versatile so that a canoe could be held in the same rack as a mattress? Did those systems exist, or did Amazon have to create some of those to increase the utilization of the space?

Yes, absolutely. We were able to improve on those aspects over two to three years. Initially, we struggled. During peak times, our overall cube utilization would show that we were at 90%, but we were actually struggling at 120% in certain spaces. In other areas, we had available space, but we couldn't utilize it.

That's when we learned our lesson. We worked with our vendors to make those changes. What ended up happening is that we would make small adjustments in the rack seasonally, which would help us accommodate new items. We also designed new types of racks for certain items to reduce damages. These are all the different things we were able to create. It helped us become more efficient compared to our third-party carriers. Third-party carriers would do things in a very general manner because they are not just serving Amazon, but also Wayfair in part of their warehouse. So they have to keep it more open and flexible.

In our case, because we had data, we knew what kind of items come at what time. We were able to make our racking more efficient to accommodate those changes throughout the year.

What about robotics and automation? I know it's a significant part of the smaller merchandise fulfillment centers. How much could be used in the large and bulky, the heavy and bulky?

Heavy and bulky items are typically stored in warehouses with racking. These warehouses use equipment such as reach trucks and forklifts to move these items. PIT operators usually operate these machines. Although some amount of conveying and sorting can be automated, manual labor is still required, especially for heavy items.

At that time, we were exploring automated PITs. These machines can navigate to the correct rack, pick up the item, and move it to another PIT for palletizing and dispatch. Not only does this increase productivity, but it also enhances safety, which is a significant concern in our operations.

It's interesting to hear about this evolutionary process. You started with five facilities, expanded to 20, and now you have 30. I'm sure the 20th facility, like you mentioned, had a different configuration with two docks. Initially, you might have started with a C-shape design, typical of a Fulfillment Center (FC).

How are these changes implemented? Do you build a few FCs each year with similar designs, and then try new things the next year to improve efficiency? Or does each FC evolve individually, with the latest one incorporating the most recent innovations? How are these improvements deployed throughout the network? How often do you go back to older facilities to reconfigure and update them?

That's an excellent question. We do have different generations of FCs. Traditional FCs were set up and built in a certain way, and we've seen various iterations even within these traditional FCs. For instance, we moved from 500,000 to 600,000 square feet FCs and then to 1.2 million, before realizing that 600,000 is the most efficient size with the automation we've introduced.

The same principle applies to our FCs. We created a new design that became the standard for that year. All new FCs built that year, around five to 10 of them, followed this new standard. We then evaluated the efficiency of the new standard and calculated the cost of modifying older FCs to match it.

We did go back and make modifications to previous FCs, which proved beneficial. For instance, when we were planning to build another five to seven FCs the following year, we realized that in certain regions, we could modify three existing FCs instead of building a new one. This approach allowed us to increase the cube in an existing FC and save on the cost of building a new one.

Could you explain the payback or return thresholds required to make the decision to build a new FC? If a new FC is built, it would add aggregate capacity to the system, which would likely be needed in the future. If the decision is made to forego building a new FC in favor of making three or five existing ones more efficient, how does Amazon conduct this analysis? Investors are often curious about decision-making at Amazon, given the significant capital expenditure and investment. I'm interested in understanding how this analysis was conducted, what the trade-offs were, and what return thresholds were considered in making these decisions.

That's a good question. It's always a challenge when planning for the next two to five years. There are two aspects to consider. Firstly, adding a new FC brings you closer to your customer, provided it's in the right location. After reaching a certain regional location and footprint, any additional FC will just add more cube, but it won't bring you closer to the customer. Therefore, it won't save much on middle mile and last mile costs.

Some of the decisions we made were even more difficult as we also had to evaluate how much we could reduce in terms of middle mile, last mile cost per cube added in a particular location. Typically, when we look at modifications in an existing FC, we want to see how much cube we can add as a result. For instance, we considered changing our normal aisles into Very Narrow Aisles (VNA), which can increase the cube in the same racking system by almost 30%. A 30% addition essentially means that if you can do that in three FCs, you don't need to build a fourth one.

Right, you've essentially added one.

However, the challenge is that this requires significant capital investment into new types of PIT equipment, as the VNA aisles require narrow reach trucks that can only operate in those aisles. So, we had to balance the cost of forgoing one FC, which we could achieve with the cube addition, with the additional investment required for these new types of PIT equipment. For these types of investments, we typically look for a three-year payout to decide whether to make the additional investment. But for FCs, the payout is much longer if you're building an entirely new building.

Is there a rule of thumb in terms of getting closer to the customer and reducing middle and last mile costs? Is there a correlation between the distance to the customer and the aggregate middle and last mile costs?

It's particularly critical for heavy, bulky items as the cost of middle and last mile delivery is comparatively higher than general merchandise. If your operations are efficient, typically, 70% of your cost is spent on variable costs in the middle and last mile, and 30% is spent on your FCs. The majority of your cost is in the middle and last mile, more so in the last mile than the middle mile. The middle mile cost is much lower. As a rule of thumb, the cost is a function of what is changing and by how much. If you already have two FCs and you're adding three more in the same region, your cost differential would be lower compared to when you just have one and you're adding one more.

We typically do a clear calculation. We have a customer base that we serve through these FCs. We simulate the alternate network that would be created with the additional FC and calculate the change in the middle and last mile cost. To reduce last mile cost, we also have to build delivery stations, which is done in parallel. It's a function of how many last mile delivery stations you have, as that's how you get closer to the customer. We optimize the middle mile based on the last mile delivery network that is being created, and the FCs that are set up. For us, it was about running this simulation to make a decision, to determine the cost benefit from the third location to reduce our overall variable cost. The second consideration was the need for this cube to feed the growth that we expect in the next year.

I assume this is all software-driven. Is there much human intervention in managing the middle and last mile planning and scheduling? It's such a complex system. Can you run those simulations easily with the software and systems in place or does it require a lot of effort?

When you start with the last mile routing, we initially used simple map views to do the routing manually on a small scale. However, this doesn't scale. You can probably do it manually for 30 to 50 large packages a day, but beyond that, you need more sophisticated software to build the last mile routing logic. For the middle mile network. it's essentially a network analysis, connecting nodes most efficiently. Many companies have built this logic, and any new company can buy that logic to use it for building their middle and last mile network. What's more critical is having the right data and customer information to predict where the demand will come from and by how much. That's where we, at Amazon, had the advantage because we had our customer data, which we used to build things more efficiently.

Everyone can understand the logic, but you have more data to train the systems and extract more relevant, superior insights from it. That's interesting. I'd like to return to the topic of India for a moment. From our conversation, it seems that the US wasn't very mature when you began to build the heavy and bulky sector; it was mostly handled by third-party carriers. It appears you were responsible for building the heavy and bulky sector in India, almost from scratch, and similarly in the US. You mentioned earlier in our conversation that the Indian sector is more evolved. Why is that? What differences between India and the US contribute to this? I'm curious.

Yes, absolutely. Let me clarify what you said. When I joined, the US business was large, but it wasn't operated by us.

The Indian business was non-existent, and then we started it. I would say it's more evolved because we started operating that business entirely by ourselves from the beginning. By the time I moved to the US, we had built a lot more things. Gradually, we have started building many of those things here. For the Indian business, we had multiple nodes in different formats. We had FC operations, Seller Flex operations where we owned Seller Flex in smaller locations across the country. We had last-mile locations where we stored some of our sellers' inventory. We created many of these smaller nodes to operate the business.

One of the reasons we had to do this is because the competition in India was very innovative, and we had to act quickly to try out different things and build them. Beyond that, for India, we set up returns for these items much sooner. We handled the return ourselves, the older item, how to get that item back. We enabled customers to sell their old items. In the US, people can sell their old items, but no one is going to buy them to dispose of them. In India, you can actually get it back, employ someone to refurbish it, and sell it as a refurbished item. We got involved in this process at a much earlier stage in India. Many of these things are still not in place in the US.

I meant to ask about the reverse logistics and returns of heavy, bulky items. For example, mattresses are a complete nightmare because you can't typically resell a mattress. So you basically have to throw away the returns. These are traditionally difficult reverse logistics items, especially if you're using third-party carriers to handle them. How does Amazon in the US handle this? Is it managed by Amazon or third parties? Or do you tell the customers to keep some of these items because the returns are too expensive? I know Overstock does that a lot. They just say, "Oh, we sent you the wrong thing, just keep it," or "It's broken, just keep it," because they don't want to deal with picking it up. It would cost them too much. So how does Amazon handle it?

One of our major challenges has been dealing with customer returns. Traditionally, if a customer returns an item, it goes all the way back to a return FC. Amazon has separate reverse logistic FCs where these items are categorized and then sent back to the FC. This process also applies to large items like tables or grills.

However, we began to realize that the cost of returning, recycling, and getting some items back wasn't worth it. It also depended on the state. If the box is unopened, it makes sense to return and resell it. But if it's been opened and used for a few days, it's not worth taking it all the way back. We started considering using our last-mile delivery stations as points for initial sortation. Here, we could determine if it's cost-effective to return items to the supply chain. If not, we could sell them to third-party vendors who manage liquidation.

That's an interesting and smart approach. Is there anything we haven't discussed? Any particular challenges or differentiators in this market segment that we haven't touched on?

One significant area in this space is deciding what inventory to keep. To be efficient and profitable, it's crucial to determine what inventory you own as a company, what inventory you want to bring in from your sellers, and what inventory you want the sellers to hold.

Another aspect is managing transportation. If you've established a network, you could ingest seller inventory and handle their transportation and last-mile delivery. If they can fulfill orders efficiently, you could even give them a Prime tag.

I was going to ask about that. How significant is the uplift in this category for items with a Prime tag?

We typically see a 30% increase. However, this is a ballpark figure as it heavily depends on the category.

So, if they can meet the Prime service level agreements (SLAs), they don't have to hold the inventory, but they can benefit from the improved conversion of the Prime tagging. It seems like a tricky system to manage. Was there a lot of variance in performance over time? How did you gauge your performance over time and make improvements?

Yes, absolutely. It's not just about inventory, but also about maintaining quality control.

This has been extremely helpful and interesting. Thank you again.