{audio:0:00:01} Philippe, a pleasure to have you with us. Could you provide some context to when you joined Amazon in 2000 and the state of the network, in Europe?

{audio:0:00:13} Back in 2000, we had just started in France. France opened 29th August, 2000. There were only two countries open, which were UK and Germany, which were both opened in 1998. When I say opened, it’s not really open, because Amazon bought two companies – one in the UK and one in Germany – to start up their business in those countries. In France, one of the particularities was that they couldn’t buy anybody because there was nobody to buy and, therefore, they started from scratch. At that time, we were only selling four products. Books, obviously, CDs, DVDs and software and video games. Those were the only things sold on Amazon, at that time, which was the case, by the way, everywhere in the world; in the US, the UK and Germany too.

{audio:0:01:17} Therefore, there were three websites and there were three fulfilment centers. One in the UK, one in Germany and the one that we had just opened in France, when I joined. I’d been hired to oversee the logistics for France and I set up the first fulfilment center and the first distribution network.

{audio:0:01:40} What are the specific differences between the way that Europe is structured, versus the US, in terms of logistics and networks?

{audio:0:01:51} Not too many, if you consider each country as a country. If you compare France and the US and UK and Germany, there is not much difference. Originally, there was one website, one fulfilment center. Now in the US, in 2000, in the US, there were maybe three or four fulfilment centers. There was Seattle, one of the historic ones. There was one Reno, in Nevada and there was one in Delaware, that I visited, when I joined. There were maybe one or two others, but no more than that. There was no difference in the structure. Each of the websites had their own catalogue and had each of the websites had one – or in the case of the US, several – fulfilment centers. In Europe, it was one for one; one fulfilment center per website. The products in the fulfilment centers were sent to everyone ordering on the website they had chosen. For example, if you were French and you wanted to buy on amazon.co.uk, you were able to do it. You had to log on to amazon.co.uk and you would receive, from the UK, the product that you bought. That was the structure and it was the same in Germany and in France.

{audio:0:03:16} Each website was sending all around the world, by definition. They were able to ship everywhere, but you had to go onto the website that you were sourcing the product from.

{audio:0:03:28} How has the network evolved, in Europe?

{audio:0:03:32} In Europe, it took almost 10 years, but we morphed it from this one to one relationship between the fulfilment center and website to the European fulfilment network, where every product in Europe could be seen in every website in Europe and could be fulfilled from any fulfilment center, in Europe, more or less. Of course, there were some restrictions. For example, with an electronic product, with a UK plug, you cannot sell that to continental Europe and vice versa. Also, when the product is not allowed to be sold, because it is not compatible in that market. But globally, at the large scale, I would say that it is a unionized market, where products are sold anywhere in Europe and fulfilled to anywhere in Europe.

{audio:0:04:29} Obviously, it started from the country’s structure and there are still huge ties between the website and the country fulfilment center, because you want to get the product closer to the customer, where they are buying from. Therefore, most of the French products are still in France and the same in UK and Germany. Now, also in Europe, two additional websites have been opened, in 2008 and 2009, in Italy and Spain, where we grew again. Some other websites have been opened, not necessarily with fulfilment centers, but fulfilled from elsewhere. The most recent one is the Dutch one, amazon.nl, where there is no fulfilment center, yet, in the Netherlands and most of the products come from Germany.

{audio:0:05:26} Is it very important to have a .nl website or a specific country website, in Europe, to give Amazon an advantage?

{audio:0:05:34} The only advantage is the language. That’s one of the big differences. If you want to compare the US and Europe, although Amazon talk about the US and Europe as regions, for Amazon, Europe is still an aggregation of countries. One of the big differentiators, is the language. We cannot have one website for all and it’s the reason we have those different websites. Where the Netherlands has been a very big customer, particularly from amazon.co.uk and some of the amazon.de, as well, because there are both some English and some German speakers in the Netherlands, for a long time, we didn’t think it was necessary for them to have their own website.

{audio:0:06:32} But when you get to this last thing of having the website in your own language, that helps a lot and greatly improves the conversion and it’s opened to more people in the country. That’s the only reason. It’s the reason why Italy and Spain have their site and Amazon is working, more and more, to open other websites. The key element is the language.

{audio:0:06:57} Because of that language advantage on the website, for Amazon, do you think they have a much larger competitive advantage in Europe, as opposed to the US, due to the network? How do you compare Amazon’s potential lead, versus competitors in Europe and the US?

{audio:0:07:17} The US is still ahead of Europe, in terms of competitive advantage. In the US, Amazon is very well-advanced, in several aspects. Firstly, it started earlier; it started about three to five years earlier and even earlier compared to Italy and Spain. The language is becoming an issue, however, because it’s not possible for it to be all over Europe, instantaneously. Within the US, by default, you can sell it anywhere in the US as they speak English. You can also get the amazon.com website in Spanish, for the Spanish-speaking population in the US, which is pretty large, particularly in the south. But in Europe, you have to open those sites country by country. Of course, as long as people speak English, you can go on every website, because on most of the website, there is the option to translate into English. But if you want it in your native language Amazon, unfortunately, has to open one, which they have done and they continue to do. It’s the reason that Amazon is quite well-penetrated into every European country, but still, I believe the fact that they lack some of the languages, they are missing opportunities. The opportunity will be even bigger when they start to get every website in the proper language.

{audio:0:08:47} There is, I think, still a lag of penetration, all over, if you take the entire population of Europe and the penetration of Amazon into households; I would say it’s behind. There are some countries, in the east, particularly, which are not yet covered. Today, Amazon covers, pretty much, most of the countries in the west. The Nordics are not covered the same way, but 70% of the Nordics speak English and there is good penetration there. I’m sure that they will have more languages. There is currently no Portuguese or Greek. Therefore, those will be pretty large populations, good markets to expand into. They are in Turkey now, however, so they have done that for the 80 million Turkish.

{audio:0:09:39} Taking a step back and looking, specifically, at inbound and how inbound was structured, originally, with only the first-party inventory. How did you look at organizing that process?

{audio:0:09:53} Not particularly differently from how it is today. The process of the inbound has always been structured the same way. There is a purchase order and you receive this purchase order in whatever country they are in. The classical one is that an Amazon order gives you the time to be delivered and the address to be delivered to, with a purchase order. You fulfil the order; for most of Amazon, you still organize most of the transportation. Over time, that has moved a lot, in the US, for example. We took a lot of ownership of the inbound so we came and picked up from suppliers but, nowadays, it’s a balance. It’s happening in Europe, as well but, in Europe, the tendency is still, you deliver to the fulfilment center, to the address you have been given. When the product arrives, we reconcile and scan every single item against the purchase order and put it in the inventory.

{audio:0:11:02} Before that, you have to make an appointment for your truck, for your delivery and you deliver, we reconcile against the purchase order. We take everything we have ordered and we reject things that have not been ordered. Ultimately, we pay the invoice and the product is stored in the fulfilment center. That has not changed. It has always been like that. The way that we store the product has slightly changed, over time. Also, the way that the scanning takes place, there are some processes that have been developed with vendors, where we had good accuracy on delivery, for example. It was called a license plate receive. When the vendor already has a good level of logistical optimization, they are able to tell you that, in a box, they put the items, they are able to deliver the box with a license plate number and they are able to give you, by EDI, what the contents of each box is. That enables you to start scanning only one box and automatically receive the inventory that is in the box. This means that the vendor has a very good level of accuracy, as well.

{audio:0:12:25} We have vendors, like that, where we have had a long-term connection with them and they have very strong logistics, as well. They are able to provide those numbers, which help a lot, with putting them in the inventory, in our fulfilment center, by streamlining the process and avoiding having to scan every single item.

{audio:0:12:46} They have always been scanned once and you need to also think about the fact that it is very rare for Amazon to receive 20 pallets of the same product. There are some products where we do receive them like that, but the reality of Amazon is that of the very large catalogue, where have we have very few items per SKU, which means that you always receive a few pieces of each product. It’s rare that boxes are full of the same product.

{audio:0:13:22} How did the inbound process change as you rolled out Fulfilled by Amazon?

{audio:0:13:28} It was interesting. It didn’t change much. The backbone of the process is still exactly the same. If you look at that purchase order, receiving the purchase order and storing, when you take the FBA, there is no purchase order in that case, obviously, because the vendor is sending the product. But the vendor has to create a shipment – I don’t remember the name exactly – but you could say it’s a purchase order. They are creating a shipment, basically, and saying that I am going to send all these products. For us, that will be the thing that we reconcile against, because that is the inventory which is given.

{audio:0:14:12} Therefore, it’s doing exactly the same thing. It’s creating those labels and it’s close to the FBA process, except that it’s the license plate process. It’s going to create these shipments and say, we got all this inventory to send to you. Every time that a vendor is assigned to a fulfilment center and they are going to deliver the product to. Now, it’s going to create this shipment, with everything on his own carrier and Amazon have a solution, as well. You have UPS or those kind of carriers that can come and pick up and deliver to us. Basically, you are going to create your shipment, with all your products; you’re going to record it in the vendor portal, from which Amazon automatically knows that they are going to receive those products. When your shipment arrives, we reconcile and scan everything, the same way that we receive 1P, except that it is not against a purchase order; it is just against your own shipment data. After that, we store it, in exactly the same way as we do the 1P, without any differentiation.

{audio:0:15:23} Therefore, the process has not changed much. The two differences are that there is no purchase order, because you create the purchase, but you cannot send products without creating anything. If you have not declared or entered into the system that you are going to send products, we will not receive the product; we will consider it as something that we don’t know about. The second difference is you don’t have, necessarily, an appointment when you start using the parcel carrier. We don’t necessarily know when it will arrive; there is no date, in the same way we have for the 1P. Therefore, there is a bit more variation, in terms of what is going to come when.

{audio:0:16:11} What was the biggest challenge in onboarding so many vendors, through this program, at Amazon?

{audio:0:16:18} Firstly, the people have to do the job themselves. They have to generate their labels and create their shipment. To really have a platform where people could manage everything by themselves, without having to interact with people, at Amazon – the system, yes, but not the people – that was one of the challenges. It was pretty well-designed, from the beginning, which meant there was not much to do.

{audio:0:16:53} The challenge, which remains today, is the accuracy of the people, to do the right job of sending products, which have labels – you have to label your product, when you send it to Amazon – and doing it correctly. Put the label on, but also put the correct label on. If you order from a 1P vendor or any supplier, there is usually a high level of accuracy of what you are going to receive. The barcodes match the product, usually, because they are manufactured by them more frequently. In the case of FBA, unfortunately, there is a greater likelihood of mismatching or there are some products missing. Not intentionally, but just because people are not used to it or it is an additional step for them. Therefore, this one of the big challenges which we’ve seen from the beginning and it is still the case now, as the number of vendors in FBA is still growing. We cannot say, we’ve now been dealing with these vendors for 10 years. Some of them, yes, this is the case. But every day, every week, there are more vendors that come in, and they all have their learning curve. So there is still a high level of inaccuracy from the FBA vendors, which is still a challenge for the fulfilment center.

{audio:0:18:31} Do you think there are any other real operational limitations, which will prevent Amazon from scaling FBA, apart from the challenge about mis-sorting or misplacing?

{audio:0:18:43} No. The only other challenge is the capacity. The idea of FBA was to utilize some of the empty capacity in both space and throughput, to optimize our own logistics, particularly off-peak. When that was very small, there was always enough capacity for that, because it was small. As it grew, obviously, the capacity now has to be well-managed and accounted for. Therefore, one of the remaining challenges for Amazon is scaling the network, to make sure there is enough space for those FBA vendors. What is important is, there is no limitation with an FBA vendor. When you subscribe, you say, of course, what kind of business you are going to do. But there is no commitment or limitation, necessarily so Amazon sometimes has to put some limitation, over time, as the FBA grew. The ones that were really shifting some considerable volume, above and beyond what Amazon had expected, there had to be some discussion.

{audio:0:20:07} So the one remaining challenge to scale, is to make sure that they always have the capacity for the FBA vendors. You pay for the inventory you have in the warehouse, and you have seen the inventory policy and the rates, which can change by time. If you leave the inventory too long, the rate increases, quite significantly, because Amazon has to manage that. Amazon can adjust stock, indefinitely, basically. That, I believe, is one of the remaining challenges, but it doesn’t prevent Amazon from growing this business; it just requires a little bit more attention on how to scale it properly. When we started, we were doing that on the back of our 1P plan. We thought, okay, it’s going to be a few percent on top; that’s fine. Now you can’t just do that. If you don’t plan for the growth of FBA, you won’t have the space, because FBA is quite large.

{audio:0:21:12} How would you approach that from a network perspective then? As you said, you can change the pricing of the inventory in the warehouse, to move it quicker, but would you look at adding different fulfilment centers or warehouse, specifically for non-Amazon customer, to increase the capacity?

{audio:0:21:32} That’s exactly the different way that Amazon has approached it. Now there is proper planning for FBA inventory, which is done by the FBA team. What they do, basically, is they work half with larger FBA vendors, who have a significant amount with us, where we can have a real vending process, where we ask them what they are planning for and what their forecast is. Then, they are also working, statistically, with onboarding and seeing what size of vendors there are and what they expect onboard. They are doing the planning. FBA, like any product line in the company, has its own planning, where it comes back to the fulfilment operation at Amazon, to tell them and booking some space, in advance, based on their need. The planning of the expansion of the network now, is done including those numbers.

{audio:0:22:33} Yes, today, I am sure, in Europe or in the US, there are many warehouses that are only for FBA. If you are taking the space used by FBA in each country or in the US or Europe, you will see that there are many sites which are entirely FBA or have been built only for FBA need.

{audio:0:22:59} I suppose, the next step could be non-Amazon vendors or customers or outside customers, to increase that?

{audio:0:23:08} I don’t think so, because FBA comes back to the philosophy of Amazon. Amazon does not do logistics because they’ve found it to be the most profitable business on earth, because the most profitable business on earth is called cloud computing. Therefore, Amazon just do logistics because that’s an enabler of business and retail. You need to think about the fact that Amazon do it if they can first benefit from it, as a retailer. Secondly, when whatever they had could help what they do for their retail business, to lower the cost, particularly, and enable a service. The max optimization of a fulfilment center is the box that Amazon is growing now, which is between 80,000 to 100,000 square meters. Those are the big fulfilment centers that have deployed. In the US, they are slightly bigger, because there is space and in Europe, slightly smaller, and in Japan, as well. But they are an optimal size for a fulfilment center that, when you fill one, there is no benefit from having other one. You need to start from scratch. You need to have a new director of the fulfilment center and an entire structure.
{audio:0:24:33} Therefore, there is no scaling effect of that. In that case, why would they start opening fulfilment centers to give to others? They would give others the fulfilling excellence that Amazon is able to provide, with no other benefit. You can charge for excellent fulfilment, but it would, mostly likely, be more expensive than 3PL. It would, maybe, be done better but, most likely, more expensive. I think, as well, if you are a real competitor of Amazon, you’re not selling on Amazon. Why would you give your competitor your logistics and be depending on him? That makes no sense. It’s the reason why I don’t think that Amazon will.

{audio:0:25:19} What Amazon has done, historically, was for example, worked with sellers who were selling on Amazon and wanted to do FBA, but have their own website. You didn’t want to split your inventory. Amazon went back into that, because when it started, it was very frequent that we could ship the order from your own website. It’s still being done, for historical sellers, but it’s not offered everywhere, on every website, anymore. Again, it’s not something that Amazon are willing to do too much. I would say that Amazon will not step out and start to just do logistics, because that makes no sense, because it makes no sense to do that directly, if it doesn’t benefit their retail business.

{audio:0:26:16} When you see them moving into the Amazon Freight business, that’s because it reduces the cost for their internal business?

{audio:0:26:28} Exactly. Transportation, that’s different. Fulfilment is a smaller-sized business optimization. Any kind of transportation, it’s very, very large and the more you bring, you can optimize. If you want to fill something with a container, but you are not able fill it completely, you have leverage to negotiate the price. In that case, you might want to take other product to fill it, because you will lower the cost for you and for the other party and there is a benefit. The last mile is exactly the same situation. The more you have a certain density, the better your costs will be, in delivery. You are better to aggregate other volume, when you have not yet reached the density which is the best for the ROI. It is the reason why, yes, in that case, the slight difference will be that Amazon open, if there is an opportunity to take advantage of lowering its own costs, by doing so. In that situation, then it might be a way for having lower costs in distributing their own product and that benefits their retail business.

{audio:0:27:43} Inside the warehouse and picking and sorting, can you just outline the principles that Amazon use, to organize inventory, inside a fulfilment center?

{audio:0:27:56} Amazon has different types of buildings, which are mostly linked to the type of product or, at least, the size of the product. Amazon has a big split between what they call sortable and non-sortable. Roughly, the sortable is 30cm by 30cm by 40cm. It’s a large box, just below a microwave oven box. This is sortable, because it was meant to go on a conveyor, whereas the others were too large for that. It’s a purely arbitrary choice, which was made, very early, by Amazon, based on the sub-optimization that they had. Therefore, in the Amazon network, you have sortable buildings and non-sortable buildings. In non-sortable building, most likely, there are racks and they can manage large sized boxes, which can sometimes be heavy.

{audio:0:29:09} There is a third category that they call the heavy-bulky, which are most likely to be washing machines and fridges and stuff like that. It is definitely two-man handling and sometimes, even needs to be handled by equipment. Those are the three main categories.

{audio:0:29:26} Amazon has also segregated the last one, which is shoes and apparel. There is a very specific fulfilment or storage method. That’s upstream. Each of the buildings, after that, they are pretty much the same, for each type of the product. The biggest part is still the sortable. More than 85% to 90% of what is sold by Amazon is sortable, so you will find a huge number of sortable sites. Those are still the typical products for Amazon, because they started with books. They have shelves, mostly, multi-story shelves, where the people walk into, to pick up the products. The shelves could be different sizes – there is a small shelf and a large shelf. They have a few pallets, as well and they also have some pallet racks, for replenishment, for the high-rotation ones. That is the structure for sortable.

{audio:0:30:43} For non-sortable, there are a lot fewer shelves. There are some shelves, because some large products can still go on shelves, but mostly, it will be pallet racks or pallets on the floor. More traditional picking like you could find in any traditional warehouse. The heavy-bulky will sometimes be with mass pallets and no racks, for the very large items. Of course, when you understand that volume, as I say, 85% to 90% being sortable, the non-sortable are larger, because of the huge volume. In terms of space, it might be today, in Amazon that 50% of the sites are sortable and 50% are non-sortable, because of the space taken up by the products, in each category.

{audio:0:31:34} That is how it is organized. Inside, you have a receive area, which is pretty common again, adapted to the products. For the sortable, you have conveyors and people are putting the products on a conveyor, which go in this big tower. In non-sortable, it’s done more by forklift. That’s the major difference.

{audio:0:31:55} The ultimate change, which came later, was the automated side and the less automated side. There are a few historical manual sites, where people go around, even with carts. There are still some of them and they are more than 10 or 15-year-old. The new ones are the Kiva enabled sites, where nobody really goes in the big tower. The robots go to the shelves, to pick up the product and bring to the people.

{audio:0:32:32} Is there any advantage that a vertical specific competitor could have against Amazon, in terms of sorting and managing inventory, in a fulfilment center? For example, pet food comes to mind, where you have Zooplus in Germany, or Chewy in the US, where they sell online pet food. Can a company or vertical like that, have an advantage over Amazon, due to the designs and organization of the warehouse or fulfilment center?

{audio:0:33:02} I’m sure they are. One of the things I didn’t mention, which was very key is that, historically, companies like Amazon, some things are stored randomly, for three months. There is no order. You don’t find a certain product there; it’s random. The only constraint that you have is fresh products, which needs temperature control and not necessarily fresh or frozen, but even some of the food, which is dry food, but needs a certain temperature, not in excess of 30 degrees. Hazardous material needs to be in a certain area in the warehouse, created for products that are inflammable, which have to be stored separately. The rest is entirely random, within the warehouse.

{audio:0:33:53} Of course, the fact of narrowing your product will improve, for example, your density in picking. In Amazon, you can buy a toothbrush, a book, your pet food. You will do that one time and the next time you will order something completely different. Therefore, the randomness of the storing comes out there, because there is no algorithm that you can manage what products are going to be sorted together. If you have a narrow range, with fewer items, the connections of things that sell together are going to get better. Secondly, size wise, you may also have a more standardized product size. In pet food, for example, you have the big bags and you have all the tins and that’s it. The boxes are pretty standardized. In that case, if you have a more standardized product range, you may be better optimized in your fulfilment, there is no doubt about that. Indeed, if you have less SKU, particularly, which is one of the key elements, as well. In the large fulfilment centers, there are more than a million different SKU. There are not that many manufacturers, or even vendors, who are not distributors or retailers, who are managing more than 50,000 or 70,000 SKU.

{audio:0:35:32} How does Amazon manage between 1P and third-party inventory, in the fulfilment centers?

{audio:0:35:38} Third-party is FBA inventory. There is no 3P inventory at Amazon. There is not any difference. As I said, I described the process of inbound, but inside, it’s completely mixed. There is no difference in treatment between items. We know which is which, because one will have the specific label that I mentioned before. But there is no difference inside. They are stored side by side, in the same bins, sometimes. They are picked in the same round, for 1P and 3P. Sometimes they go in the same package, to a customer that has ordered one of each. There is absolutely no difference in treatment.

{audio:0:36:22} Before we move onto the last mile, when you look at the whole process, from order to shipping and delivery, what is the split in the total cost of the package and delivery? Is it 10% on inbound, 20% on picking and 70% on outbound? How do you look at where the cost and real challenge actually lies, in that process?

{audio:0:36:52} Firstly, it depends if you talking about fulfilment cost or the total cost, including picking up at the vendor and delivering to the customer. The in-the-box cost, from the door of the fulfilment to the outbound door of fulfilment, roughly 30% is spent in inbound and 70% in the outbound, of the total cost of process. It depends; it could be 25%:75%. Two thirds of the outbound is going to be the packing and shipping activity and one third is going to be the picking, roughly.

{audio:0:37:40} Again, that really depends on which type of building. It will vary a lot, depending on if there is a Kiva system, where there is a little bit less time spent in picking. There is a merge between picking and storing, because the same people are doing both, so that’s blurred a little bit. But roughly, I think, it would be that split, so 50% is in the packing and shipping.

{audio:0:38:11} Which is where scale matters?

{audio:0:38:16} I don’t know. The picking is where scale matters the most, or the density. We mentioned the density in delivery and the density in picking is one of the keys. Whether you are in automated or not, it’s still an optimization, because if you are in manual, that matters a lot, because that is the time that the picker has got to spend in walking, from location to location. But even when you are automated, it matters. The less movement that the robots are doing, the more movement they can do and the less distance they cover, the more available they are for the next pick. Therefore, if you can bring your shelf to a picker and have 15 items to pick at the same time, on the shelf, it’s better than having a shelf and pick one and having to then bring another one. The productivity will be better. Picking is maybe where the density matters most. Not the size because, of course, the bigger you are, the more challenging are your pick paths. But if you have a lot of volume, that’s where it matters.

{audio:0:39:33} For the rest, a little bit less, because the storing is random and so people are storing in the same area. When they go, they store and, as they don’t have any rules, they just store in the same area and they don’t move much. For the packing, in that case, no. It is not the case that if you have 10 packing tables it will go faster than one packing table on its own. For the whole organization, yes. With the conveyor, automatically loading trucks, those kind investments which could ultimately help in lowering your costs.

{audio:0:40:13} Just thinking back to the pet food example. It seems as if you really organize your warehouse, specifically for the category and you have some automated, Kiva-like systems to pick this stuff, at density, you can really compete with Amazon, inside the box, inside the warehouse, on efficiency, for your category?

{audio:0:40:36} Yes; I’m definitely convinced that there are people who may be more efficient, particularly when their product categories are very similar and the product types are very similar. I’m sure that you could be even more efficient than Amazon. Amazon is very efficient for the kind of structure of catalogue that they have. That’s the key. I don’t think of Amazon has the best fulfilment cost ever. I think some other people will have a better fulfilment cost. But nobody could reach it with the same catalogue and the offer they have; that’s the difference. Unfortunately, you cannot reproduce that. We started, at one point in Amazon, to do benchmarking in different countries, where we would benchmark activities, end to end. You start to find people who can compete with Amazon and, usually, the best competitors of Amazon are the ones that are specialists. It’s a choice, but you will never be the size of Amazon, if you are a specialist. To sell hundreds of millions of pet food, that’s not possible. But yes, you could be very efficient in doing so and become a very strong competitor.

{audio:0:41:55} Basically, in those benchmarking activities, originally, we were benchmarking against people who were doing the same as Amazon, such as online sellers. In those cases, we were usually more efficient than most of them, in every country. But then we started to benchmark with specialists, and decided to benchmark Zalando, for example, for shoes and apparel and, of course, that was tighter. For shoes and apparel, Zalando was pretty efficient, in just doing that. We developed some fulfilment centers that we focusing on those products because we discovered that they were rarely cross-selling. They were sometimes, but it was less important and it was better to optimize the warehouse. With those kind of things, I am definitely convinced that some people, being specialists, could reach better cost on their kind of product.

{audio:0:42:57} Can you elaborate on how Amazon’s strategy with last mile and with carriers, has evolved, over the last decade or so?

{audio:0:43:07} Our carriers weren’t involved much. Historically, Amazon has always worked with several carriers in every country, over time. Of course, we started, in every country, with the postal network, particularly in Europe. La Poste, in France; Royal Mail in the UK; DHL/Deutsche Post, because Deutsche Post bought DHL in 1999, just at the time when Amazon started. After, we expanded from the need for more capacity and to get some competition between the carriers, we used all the carriers, depending on how much they were available, by country and so already had a network of carriers. Up to the point that, in 2008, 2009, we were starting to face issues, particularly for the next-day delivery. Amazon started the Prime program, in Europe, as a one-day service – not as a two-day service, as many people think, particularly due to the recent announcement in the US, where they’ve moved from two-day to one-day. Europe has been a one-day service in Prime, since day one. When we started Prime in 2007, 2008, in the three main countries, we started it as a one-day service.

{audio:0:44:33} That created a lot of pressure on the one-day delivery, in each of the countries. Two years later, as Prime was growing, aggressively, in each of the countries, we started to face some capacity issues and this pushed us to think about, why should we not invest in that, because no carrier was really succeeding to invest properly in doing that. That’s how that started, basically.

{audio:0:45:01} In the UK, particularly, it was one of the first times we started an initiative, not in the US. My team and I proposed it and to run the first pilot in the UK, because the UK was a market which was very hurt by the service. Royal Mail was not a one-day service and first class was not. Many carriers tried and didn’t succeed to deliver what we needed. Therefore, we started to do it. We started to do it because we were really facing a capacity issue, to which carriers had decided not to respond, at first. We said, we’re going to start and very early carriers, like Royal Mail, knew about it. Ironically, I would say, they didn’t believe it. For a long time, these initiatives were grown out of our transportation department. I was vice president of transportation, for Europe and, at the same time, I was negotiating with all the carriers. On the other hand, I was growing this initiative with our own team. For most of the time, people did not realize what we were doing and weren’t thinking that it would be successful. Some of them even told me that it was good that we were trying it, because then we would realize how difficult it is.

{audio:0:46:46} Every time that they missed their target or tried to improve, they tried to argue and push back and negotiate by saying, you don’t know what that costs. So some said, it was good we were trying so we could see how hard it was. Yes, we did feel it and we understood how it worked. But ultimately, I don’t think that anybody really believed that we would be successful in growing the network. That helped a lot, because there was no animosity. Unfortunately, when they realized that we were quite successful and doing quite a large number, it was too late for them to react. For sure, if you asked Royal Mail, DHL, La Poste now, they would not tell you that they were happy that we are doing our own last mile. But they didn’t do anything to fight against it and when they started to be worried about it, it was already too late; we were engaged and we were efficient. Then there was a, kind of, what could they do? Put them out of business or continue to deal with us and manage it?

{audio:0:47:59} At what density or scale does it make sense to actually insource the last mile?

{audio:0:48:07} That’s a difficult question, from several aspects. Firstly, it depends on what your strategy is. In 2007, I’d just taken over the transportation in Europe and one of my first negotiations – with one of the carriers in the UK – and the negotiation was not going well and the MD of one of these companies told me, you know what, when you have 60 million parcels a year, in the UK, you will be able to run your own business. At that time, we were not 60 million in the UK but, of course, based on the number that we had and the projection, we would soon be at that figure. But I wrote that somewhere and I parked it. After the crisis, the acceleration of the sales at Amazon really went up, after 2007, 2008 and with the Prime program helping. Two years later, literally, I though, wow, we are just above this 60 million parcels; should we start looking at it?

{audio:0:49:20} So I asked some of my business analysts to have a drill down and have a look at it to see if could be real. Indeed, they came back and said, yes, that makes some sense. We can almost optimize people every day, in a town like Glasgow – we did the analysis in Glasgow; I didn’t want to do it in London, which was a bit more obvious. I wanted a medium-sized city for the analysis, because if it works in London, it doesn’t tell where it works elsewhere. If it works in Glasgow then, of course, it would work in London and many other cities in the UK.

{audio:0:49:59} I don’t even remember the real number of the density, in package per square meter, but it also depended on which city you were in. But the reference point, that I kept for a long time, was this 60 million. Is it 60, 70, 80 million? I don’t know. But for sure, in any European country, you need to have multi-million parcels, except obviously, if you decide to be very local. That’s if you want to have a network, at a country level. Now if you start to say, I want to deliver in Paris only, you may not need 70 million for Paris. You can start much lower than that. I don’t remember the actual number of parcels per square meter, but people can easily do the analysis of what they have today. I think it should be around two or three parcels per square meter today and you start to have something that might work.

{audio:0:51:10} How do you look at the challenge of Amazon owning the last mile, in the US, versus FedEx and larger players, versus Europe?

{audio:0:51:22} It’s difficult to call out. Today, Amazon still doesn’t have the network to cover the full US anyway and it will take a long time to achieve that. You have the question about the density and the location where Amazon deploys. My belief is, the smart way would be to still keep both. If you consider the less dense areas of the country – within the US and Europe, it’s the same; within countries in Europe, it’s the same – at a certain point, there is a density at which you there is no benefit to going on your own. It’s better than one carrier goes and does the job for all the people selling in this area. This would have the best cost. Every retailer would have to accept the cost of it.

{audio:0:52:33} But the opposite applies, as well. When you take the big cities, they are very, very dense and there is a level at which adding more parcels doesn’t help. There is a level of density which is pretty high, but you could have more. But there is a time where even if a guy was delivering in his own area, without taking his car or truck back, he will still need the time to take the parcel, go to the door, ring the bell and stuff like that. At the level there, you could have more and the benefit is almost nothing. In that case, work with both, basically. Today, Amazon works with both, even in the areas where they are. If, for some reason, they are beyond this density, they may give the work to another carrier. They don’t want to set a new route and split the route in two as there is a scale effect. For example, if you have a good route, where you can deliver 200 parcels in a day and you are there for eight hours, that is a very, very good route. But if you are down to 50, you are struggling. Therefore, it’s better to give the 50 to another carrier, which may help their density and doesn’t impact yours, because you know you are maxed out at 200. If you have 400, you will keep the two routes.

{audio:0:53:46} If you want to be optimized all the time, you need to manage this. In that case, you will always need a partner to do it. I believe, for the Amazon business, the best way for carriers such as UPS and FedEx, is to really work in partnership with Amazon, understanding where they will be needed and, obviously, adapt their rate to that. But Amazon will not do better. One of the follies, for example, in the low density area, is that there will be a rate that will be more expensive than the high density and Amazon will have to accept paying more to those carriers, to do it, which ultimately, should have a better cost than Amazon would, because Amazon plus all the other parcels will always be better than just Amazon alone.

{audio:0:54:31} How do you see this relationship evolving, between the two, over the next decade?

{audio:0:54:45} I think, maybe one or two will drop. In the US, there is UPS which is there. Of course, there may not be enough room for everyone. I think FedEx have already given up a bit, with Amazon in the US. I believe there will be room for one or two, per country, with whom we work in partnership. When I say in partnership, that doesn’t mean agreeing. It’s just making sure you are giving them the right number of parcels, where you don’t deliver. We did it in the UK, back in 2015, 2016, when Amazon Logistics was already big in the UK. We’re working very closely with DPD, in the UK, for example. Of course, this business, as you develop the network, will not grow extensively. However, remember that Amazon are still doing 15% to 25% growth, depending on the region. So every year, Amazon adds a quarter to its existing business. Ultimately, every year, you get a little bit more, despite your share decreasing, for some of the business. I don’t think Amazon will do a great percentage of their own parcels, because economically, it will not work.

{audio:0:56:20} If, in a country, every carrier decides not to work with Amazon, for whatever reason, yes, maybe Amazon will do and it won’t be optimized for the 10% to 15% where they will never have the density. But they will do it themselves. For the country where a carrier is smart enough to say, okay, I know I’m not going to have the best parcels, in the middle of the city, but I work in partnership and they give me a pretty accurate forecast and they guarantee me at least a certain level of business and I will work around that. I think people will still be able to grow their own business. Now, it will never be as good as it was, many years ago, except for one thing. If you compare the long range, I can tell you that Royal Mail is doing more parcels today, with Amazon, than they were doing in 2005. Okay, it was 15 years ago but, despite the fact that, in 2005, Royal Mail was 90% of the Amazon business, today they’re maybe 20%. But that’s not what matters. What matters is the total parcels they’ve been given and I’m sure, today, they’ve been given more parcels than they were, at that time.
{audio:0:57:38} That’s what some carriers should think about. Of course, they lost a lot of opportunity, but they’re still growing their own business.

{audio:0:57:47} How have you seen the cost per parcel decline over the last two decades?

{audio:0:57:52} It will not. In delivery, it will not. I don’t believe the cost will decline, at all. I believe, today, the level that it will reach in many places, in Europe or in the US, the levels of ecommerce in general, are close to the optimal, for many carriers and for Amazon, as well. Therefore, this cost will only grow by fuel, wages and stuff like that. There are a few innovations that are coming to help, such as lockers and pick-up points. But I don’t believe it will decline. I think it will be capped, as far as possible and flat for a few years and, after that, slightly increase, over time.

{audio:0:58:46} So you think Amazon, particularly, will have to wear the increase in the cost, potentially, if they go to one-hour delivery?

{audio:0:58:55} Yes, absolutely. That’s a no-brainer. The one or two-hour delivery increases the cost of delivery. Even the evening delivery that we have, it’s just mathematical, but reducing the window of when you deliver, that decreases the density. The problem is, in a delivery network, you have to at least come from the place where you picked the parcels up. This is called the stem time, in this business, which is the time between the location where you pick up the parcels and the first delivery, has to be amortized in the round. The rest could be very dense, but this first part, you need to deal with. You don’t always deliver around your delivery station; you sometimes have to do 20 or 30 miles before you get to your first delivery. This could sometimes take an hour. In that hour, you don’t deliver any parcels, so therefore, the more you reduce this window, your productivity in delivering gets worse. I don’t believe that we will see decreasing costs in parcel delivery.

{audio:1:00:10} What do you think is the biggest challenge for Amazon’s network evolution, going forward?

{audio:1:00:16} The same that it has always been: scaling. Making sure they are able to grow as fast as they need, as required by their business need, at the same time, their quality and costs. Because you can grow and spend a lot of money but that’s going to make your costs jump. Or you can grow and keep your costs low, by sacrificing the quality. But that is not the Amazon way, either and it creates a big risk of endangering your business. I think it’s always been like that. If I take my 17 years at Amazon, that has been my biggest worry – hiring the number of people that you need and the right quality of the people you need.

{audio:1:01:05} Not compromising, as you grow, I believe, is what many companies do very badly. There is a range of companies which are good at maintaining a good level of business growing 3% and doing so very efficiently and even saving costs. I believe that there are many companies like that, who are doing a great job and delivering great results. But there are companies that want to be aggressive in growth but, unfortunately, sacrifice something. Either their bottom line and sometimes, you can hear companies that say, we are going to slow down the growth, because they know that they can cope with it, as they can see their costs going in all directions. Or sacrificing quality, which is the most frequent. My experience with carriers is where I’ve seen the most people who wanted to grow, but are not able to grow. Every time they got a new contract, with a lot of business coming, six months or nine months later, there was a discussion about the quality and there was a problem. We had a few like that, in the UK particularly. People were eager, but didn’t know how to do it. When you don’t know and you’re not structured for it, you fail. That is where, I believe, Amazon is one of the best in class, in the world. Going back to our comment about pet food, yes, you will find people doing a fantastic job, at a cheaper cost, but you don’t find people growing 25%, growing to the size of Amazon, doing it. That’s the difference.

{audio:1:02:44} With the size, you get some disadvantage and Amazon has still been able to cope with this growth, at their level.

Amazon European Logistics Buildout

Evolution of Amazon's European versus US logistics network

How inbound is organized and adapted for FBA

Sorting and picking of both sortable and non-sortable products inside the warehouse

Amazon European Logistics Buildout

Former VP, Amazon Logistics Europe

inpractise.com/articles/amzn-logistics-buildout

Why is this interview interesting?

Executive Bio

Philippe Hemard

Former VP, Amazon Logistics Europe

Interview Transcript

Related Content

Amazon Logistics: Challenge Scaling

Former VP, Amazon Logistics Europe

Amazon Logistics: Inbound

Former VP, Amazon Logistics Europe

Amazon Europe: Logistics, Hiring and Leadership Principles

Former VP, Amazon Logistics Europe

Amazon Logistics: Sorting and Picking

Former VP, Amazon Logistics Europe

Audio

Amazon European Logistics Buildout(May 11, 2020)

PARTNER

Speak to Executive