Can you lay out the major stages of a typical fulfillment process for an online retailer?

Well, as you know, I was involved more so in the front side, where you have the fulfillment and ecommerce, which is the order dropping and all this kind of, stuff upstream. Then it becomes fulfillment in the warehouse. It becomes then, when to buy, like the scheduling of inbound, the deliveries, the inbound deliveries and then all the deliveries of vendors and then basically it's the inbound piece that comes first. The inbound, it's called inbound. The inbound operations mean the dock, all the operations have to process at the dock doors.

The appointment first, and the dock doors basically, and then basically stuff coming in, and then the various differences among let's say, fulfillment providers. What happened from the dock there because they're the earliest point, they can already plan according to certain value streams or process path and so forth, because then we can get the operations where you have purely palettes coming in where the stuff remains on palettes and you have stuff that you know that will be dissembled basically, and received one by one, and then go into totes or something, and it's been stored away into shelving for example.

The process at the dock is called receive. Receive means to get the stuff into the system that you get delivered, so in your warehouse system. That is your inventory now, and then after you receive the stuff, and that's typically receive this physically.

When you talk about the typical items like your books and all the other stuff, so you unpack that stuff. There's like certain quality procedures that take place in order to enter that or do you order this order also what came in. With the certain feedback loops to vendors and stuff like that, but in a normal fulfillment process you get stuff in, get into the system so you receive it.

There's different setups for that like layouts and different lines where you have receivers for example at their workplaces and after everything was put from the palettes to the receivers and the receiver basically take the items to scan the items into the system so to say, and put them, most operation they put into totes like which can go on a conveyor belt further downstream, or they are brought directly from forklift drivers to where they are stored anyway as bulk storage.

Then you have a full palette of one item like Mickey Mouse DVDs, a full palette. Then the operation's simpler you just basically get the whole stuff into the system. Then you already go directly into the stowing process which comes after the receive, so you have the first physically in, then receive you take into the system by scanning stuff, making this basically available for stowing and then after receiving, it gets stowed. Stowed means it gets put into the appropriate storage location. That was called put-away in the early days.

Yes, so the typical stow operation would be firstly to take everything into a tote (a box container). The tote has also an ID so I scan the tote ID and I scan every item. Virtually you can always check in this tote, there's like 30 items and this is Mickey Mouse, they'll send them back towards this and this. Then this tote comes either through conveyors or I can also receive into a cart, into physical trolley or something but it's always the same. It's a container that always sits right.

How does the software link to the hardware throughout the operation?

Basically, you mean registered that it's on your stock basically? That's happened by scanning. That's happened physically by you have a like a bill of lading and all this paperwork coming with the deliveries. The receiver basically takes this information and, in the end, it opens certain IDs and then it says, "Hey this delivery has this many items." Then you scan item by item and the item is basically moved into the container you scanned it into and that's mostly tote, a trolley, or something so that now the system knows, "Hey, this Mickey Mouse ordered by this and this vendor and is now sitting in container one."

You do that for like 30, 40, whatever, until the tote is full and then the store comes into play because the stuff is then moving to the stow process in the store. The only task of the store is basically to move basically this stuff from this container, to stow it from this tote into a storage location. The storage location can be a bin, a bin ID, it's always have a bin ID, but it could be different storage types. Like we had library shelving is the typical one where you have like a single items of all kinds sitting in different bins or you have a bulk locations, library D etc. and all kind of storage locations. That's the simple shelf in the store. Scan the item, scan bin, put the item into the bin.

In most operations this is a person doing this. It's already like in the receive for example, there's a certain process called License Plate Receive, for example, where already from the vendor, you have a certain code sitting there and you have a tote already delivered from the vendor and, you know, "Hey, in this code there's these and these items" and you directly put onto a receive tunnel which then drives through and if this stuff is then automatically received into the system.

There's some automation going on there however, the most operations like, I don't know 80% or something like this, is still put where people scan stuff into the containers but you have quantity features. If he have like 200 of the same you see just do 200 of them and then these things are moved into the container. Then the store is really everywhere in the warehouse and then they put it into the shelf.

80% of the volume is manual labor, pretty much?

I don't know how far they made progress now in the last days but it's the most, this is the operation to look at in the moment so you receive it and you stow it. Once it's been sitting then in the bin, when it's I think already at a certain point, you can already see when a customer orders against something. A customer can already order something when it says on stock. It's already when you have it already in the receive, when you already received them to container, and this is why it's important to have a low backlog so to say between receive and store.

You must quickly get it into a pickable location so to say, because nobody can take it from some containers sitting somewhere waiting to stow. You can take something because the customer, if they need something, it needs to go into the bin and this is why there should be low backlog and you need to get the stuff faster into a bin. And there's other software for these different things you can do with the bins, like to manipulate the hierarchy of which bins are, for example, by the system, taken first in favor of other bins because they're easier to reach and stuff like that.

For example, if you order a Mickey Mouse, and the Mickey Mouse is in five different locations in the warehouse, there's always fancy stuff in Amazon where it's optimized for picking later, so that it's not so labor-intensive so that they have the shortest route to get the item. There's a couple of settings you can do, but in bottom line you stow it into a bin, and then this is a faster process than picking or like stowing, I don't know how many stows per hour, but I think it can be a lot because you don't need to go to certain place to pick something, you just need space.

You need to go into certain lines in the warehouse, and then you deng, deng, deng, and you see that between 600 and 800 an hour.

Is this random, manual stow process standard across the board for retailers then?

Yes. I think there's a standard when it is random stow, because it's fast and this is why Amazon has it. Amazon doesn't say it's a chaotic, they always say the chaotic inventory places, it's not like where you say, "This needs to go there and this needs to go there." They just do it by areas where you have certain products like I say, "Media we want there. This we want there for a certain reason." but the main challenge is to find space during Christmas.

What's the difference in the stows per hour if you're organizing the warehouse randomly, versus organized by category?

They're different because you still have areas. I think because in most vendors it makes sense because they have pure media vendors and pure categories in certain parts of the warehouse. In this case it doesn't make any difference because you just route those things from receive into certain areas, and this is your media area for example. However, the customer doesn't buy categories together. They can buy media plus one household goods item for example. This is what you students need to understand. You need to look at the whole picture, not the one department optimizing their performance against the other, but the overall efficiency of the operation.

The goal should always be to ask, "How do I present the stuff to the picker?" so that you shorten the walking distances there downstream. This is why the random stow can into play and you had storage types within an aisle of shelving. You would have different sized bins, and for example you have typically one row. The lower row is called the A, then the B, then the Cs and the D.

Let's say you're shelving at seven levels, and you would have typically in level F or something, on your height of your eyes, will have a small layer, just a small-- as high as DVDs or CDs. That, for example, is a very small shelf, but this is where you would put in the CDs, for example, because it's hard to read and the titles and stuff. This is just an example where you want stand storage, you want to optimize the storage capacity, but you have space everywhere available for any kind of product, unless you're talking about clothing.

How does the inventory management system links with the buying of products for Amazon. How do they ensure their inventory is replenished and in stock as much as possible?

I think there's a visibility there, so depends on of course how you're setup of the entire cluster of warehouses, where for example you have warehouses like you have in the US, you have sort vendors and stuff where you already-- warehouses that do nothing but only buy everything into one location. You take everything into one location, you already unpack it and stuff, and do everything, all the pre-work basically, and then you assign it to certain warehouses. That means you already have pretty much done a lot, and every warehouse knows what's going to come, exactly in which quantity and all the stuff.

Other warehouses where you just order stuff and it comes to warehouse A, certain stuff, there's certain scheduling of stuff, and there's what's called DockMaster, where you have scheduled appointments, and you know those slots, what's coming and what appointments, but in reality there's always a huge need for vendor education. Vendors are sending more stuff they just maybe want to get rid of and know, "Hey Amazon's anyway, going to take it in their inventory, they're not going to send it back" and stuff like that.

There was always a big PQV, Purchase Quantity Variance. That was a big issue always and a big project, because they're talking about a lot of money, simply because of the variance of what you basically paid a vendor, and what you basically got, or what there should be, and what you in the end got into the system. This variance is quite a bit, and over time, it's been drastically improved, and there have been a couple of new platforms and new software things that were rolled out at the end of my time in Amazon. Which seems even better for scheduling for visibility of inbound that's coming in, and other stuff.

Traditionally Amazon hasn't been so good on the inbound, but Amazon was concentrating more and more on this. This also came from Bezos who was a big picking fan, always came from focusing what's closer to the customer, where you get things right, and this was more downstream.

What's the biggest challenge there? Is it actually that the vendors are not up to speed or able to integrate as required?

I think the important thing is really the predictability and the planning really when for example, stuff comes in later or earlier, so that you really make sure you have the fast lane process, where you want everything within two hours getting basically into the system. Sometimes if you have a huge backlog, and labor planning plays a big role on this as well, because simply if the volume traps up and you get more and more orders, and you don't have enough stuff and you have backlog on the dock, that's basically where everything is hurting downstream.

I think it's really the ability to plan, the predictability of what's coming, and the volume. The peak in inbound is always before the other peak of course. Exactly now as we're talking, now is the peak in Amazon. They're full now in front of Christmas, and we will get less as we're moving towards Christmas, while outbound will continue to grow until six, five, four days before Christmas when you have the absolute peak in the business.

To my understanding, Amazon has a network with larger mother nodes, large distribution centers, and then they typically smaller sister nodes closer to the customer which are more like sorting offices for delivery. Is this inbound process we are speaking about is in the mother node?

Amazon tries to maximize now the storage and the capacity of the existing FCs. We're talking about big nodes mostly. I'm talking of Germany, UK and so forth. You have nodes like the Marston Gate, LTN1 Marston Gate. It's big, and there's no smaller node besides that. You have like at Gourock, GLA1 in Scotland, you've got Swansea, at Peterborough and all this kind of thing.

If there's differences, then one is more like a non-sortable site for example, they only have big stuff, really big stuff, but the sortable sites they are all maxed out and the orders get assigned to where inventory is and whether that's close to a customer. That happens in the software by the supply chain teams. For example, you are a customer close to Peterborough, of course if you order something that's available in Peterborough, the system should assign your order to that fulfillment center where this is in stock in Peterborough, not to Marston Gate. That's all sophisticated on the upstream but what you said that mother node and children node, mostly in Europe at least, mostly you don't have that.

Is it a different set-up there then the US?

We think that they are all big FCs. In the US also you have a lot of most sites they ship like then 700,000-800,000 units a day in peak. Yes, one is double as big as the other but they all function the same principle which with inbound, they receive, stow, pick, pack, ship.

So we've gone through the inbound process. So the item has been stowed and it is in the warehouse. Can you explain in more detail the layout of the warehouse in terms of how products are placed where within the warehouse?

The chaotic approach means stuff can go anywhere you have space for it. That means it's more like the variable kind of storage solutions you have. There are aisles of shelving as in any warehouse.

You have such a construction and it's consisting of X aisles, depending on the capacity for the site and stuff. You can see like in a layout, you can see rows of shelving and that's where every shelf has a bay, it's like a bay. You have seven shelves typically, seven to eight shelves.

In every shelf you have between 4 and 10 bins. That's how you typically name, how the system works, so that every aisle, every row has a number. You have for example row one, two, three, four, five. If you go there you have a certain naming convention so that you could see what is the area, it's called or example area B, and then you have row one and then in every row you have a shelf and in every bay of shelving you have a sequence of bin ids.

For example, a bin in the beginning of the row might start with Bin 1. So you have for example, P-1-A, which is, P1A means area A. P1 is for pick area one and P1, typically one would mean on the floor. P2 would be upstairs in the second layer and P3 would be in the third layer. You have P1A is the pick area let's say. Every row starts with P1A but then you have P1A1. A1 would be row one. A2 would be row 2, and so forth.

Typically, we took one bay to the other, the next bay starts in multiples of ten. Let's say you have bin 1, 2, 3, 4, 5, but there is no bin 6. If you don't have on this shelf, if there's no bin 6, on the next bay you start with 10, 11, 12, 13, 14. Then on the next bay you start with 20, 21, 22, 23, because you need to leave some room if you want to add bins later and stuff like that. This is specifically what I did with my team, so we basically set the standard for the naming convention, which has a very high influence of how quick you can pick and find the item and all this stuff.

Basically, that's on the A1 and then you have shelf. A1 would be the bins A1 to A6 for example, then A10, and then the B shelf and then you have the shelf above, would be shelf B, shelf C, D, E, F, G and how many shelves you have. That's how you basically can locate the item. You scan it in right and then it's virtually sitting in there and you can see if there's an order against it or not. This is then virtually really in this order location.

When they're stowed, the bins have the products in by the dock door, after being stowed. In terms of transferring that bin to the shelf, is that manually or is that with the Kiva robots?

In almost every site, it's the stowing processor, where the store has a scanner in the hand where you basically scan the item, scan the bin, put it in. There is a process where directly at the dock, which was a quality problem first where it was called Each-to-Bin, where you directly already had what comes in at the dock, you would directly put it with one operation, the entire thing that's supposed to be, they put it in one huge shelf.

You receive the tack, you receive it in and everything is fitting there already the pickup location, but all the errors, the mismatch of virtual versus physical and all this stuff needs to be done, basically ironed out afterwards. This is why this is never something that really was followed on further. You're right. It's mostly manual operation and there's also directed stow, the new process. There was always random stow and you would've directed stow where the system tells you where to stow something. That only makes sense in case, where you can imagine, as you get more sophisticated and you have data about everything, about every bin, about space, about everything.

If every item you know that have never been scanned, of course we have all the measurements of every item and then it's stored upright in the bin, you should know how much space left is in every bin. You could have the system direct you where to stow, and where to put an item. Direct led stow is something that was also grown, first in non-sortable area, so it was easier to do and then certain areas where it's important that stuff goes into certain location because of restrictions or other specifics.

How does the density of random stow compare to other traditional retail warehouses with this method of direct stowing?

With random stow, of course because random means you can fit every item everywhere, it's allowed everywhere. The more random you do the higher is the utilization of the bins, because you're not restricting yourself to only PlayStations or something where then you have always something left in terms of room.

In Germany, one of the tricks that they always used, used to have more storage capacity than they needed because they simply wanted to have a good productivity so that you always found space you can easily put stuff in and stuff. They mostly had the luxury of having more space than needed, but now that you rather don't build a new warehouse and rather maximize to the extreme what you have, you would fill things really up and it's really getting tough to find the space. That means also that the bins are practically really full. With fill rates, let's say above 90% and stuff and that's very important for productivity and over density.

How does this method compare in terms of density or efficiency to traditional warehouses, this organized in different ways?

I don't know, it's very difficult question, but I think compared to the random stow and the bin full of everything, I think that's pretty much a real advantage in comparison to others who are mostly having not such a selection like Amazon who have specific product and therefore and for a certain reason, put that into certain areas and so forth. Mostly we can't compare so, you don't have so many good comparisons to Amazon with the selection piece and that makes it really tough to analyze.

I think in the beginning, the difference versus there, there was more traditional ones would be either be utilizing space better because at least in Germany I don't think, because the focus wasn't too much on, "Hey, let's save space and rather not too build warehouses and maximize everything out to the extreme." Now, in the meantime, I think Amazon got really good at it and it would be really mastering.

What's the difference in, let's say you stowed it, it's in the bins, you're picking, what's the UPH to Amazon typically?

Yes, picking. Well, picking depends a lot on the process path and that's totally different than the inbound. The inbound is more like people want to get stuff into the bins so there's label for customers to order. Now, in the outbound, in order to be efficient, it becomes a totally different game because now you're talking about different process path because there are different sizes of products that are going in different packaging and all the stuff. The question is how do you organize yourself to be efficient on the outbound side and therefore, the picking we had different process path reporting in Amazon like in the metrics and of course, if you have a small process path versus a large process path, the productivities differ a lot, but let's say pick single, so you distinguish between single and multi-picking.

Singles means you pick stuff where customers just ordered one item. The advantage of that is that the walking, the distance are much lower to walk from item to item because you can imagine, it doesn't matter where we stowed something, if there's 100 customers need to basically be shipped tonight at 20:00, I can just look, hey, which orders need to go out and the system drops all the items to your radio frequency scanners which you can pick from and the system knows in which aisles and all the stuff are these items. They assign it to you and if you are the picker, the certain area there, you just go add this; the first item then the system assigns you to the next.

It'll be configured pick path; it means what is the order of bins you would pick from in the entire warehouse.

What does the software stack look like between order and inventory management system to make this run smoothly?

The warehouse management system knows which items are ordered. There are certain shipment conditions. Say, if something is ordered and ready to be picked and so forth, the stuff is collated, made ready and dropped in order to be picked, so it's scheduled for picking.

There's a ton of different things you can influence how to assign all this bunch of work you have to the pickers because the denser the more you can use the pickers to pick the denser are workers, therefore the more they can pick can pick. It is all about density.

The UPH for singles picking is higher, so it's mostly in the normal item categories. Singles would be between 170-180 an hour.

How does that compare to other retailers in the industry?

In the industry, when you only pick apparel and stuff like this but typically this would be like 10% to 15% lower in the industry.

Well, apparel is an example. Apparel would be slow, because of the hanger and stuff. Suddenly you're talking about 70 or 80 items. I just wanted to say it's different. It's dependent on process path. Basically, on the overall, so if you have certain conditions and you have the whole, the picking software and the whole bin showing, "Hey, this is the work available." Then you can as a collator, as the planner, who's on the production and plans, you can influence a lot of what to give from the whole chunk, you need to pick out today for example what to give, releases to the pickers and that was automated through auto collate.

You always have some filters and now the system automatically says, "Hey, I'm now looking into everything that needs to go out today so we're not risking any shippers and now we're getting really sophisticated."

This picking software though, this is completely proprietary for Amazon?

Yes, it's Amazon. All programs by Amazon are proprietary and also the algorithms.

How much better is it than buying something from like Manhattan Associates?

I think there's a huge advantage when it comes to when you have such a volume and when I explained to you about not only singles picking which wouldn't be as huge advantage because singles is easy, right? I know then you pick them, but when you have multis where you have warehouse that's a one million square foot building and I need to create batches to get all these items from inbound into a package, so how do I pick those?

I cannot just pick like 10,000 items, they're sitting all in pack area and then, hey, which are the three items that X customer has ordered. This means I need to batch stuff and batching means how many custom orders are you looking at the same time for one batch? Typically, the batch was between 40 and 45 customer shipments, which was, then like 130, 140 units.

If you have those 140 units, it makes a huge difference if you randomly need to go through the entire warehouse for those 140 units then get the packing and to sort it back into the shipments for X customer and then pack it. If you have a smart system that would know, "Hey, I have entire warehouse I'm showing you all items that you need to ship today." Now I'm trying to be smart and say which of all the customer shipments, have all the items in a certain subset of the warehouse, like in one area?

This is where I came with the idea with the warehouse areas to subdivide the warehouse and have algorithm checking which can be in one area, two, three and all the combinations and stuff. This is where we tested and then save at least like 10% productivity just through being smart about how you can slice the work smarter and then the process, then such a multi-picker as a normal multi would pick like, let's say 140, 150 units an hour. All right, units an hour, but again here there's a huge difference if you having a more automated type versus own picking. That's when multiple pickers pick, we talked about this example, right? Where you have those 140 units for batch 1 let's say. If you were smaller, preparation would be a one pick-up a batch operation. That means one picker has a cart let's say, a cart goes with a trolley, has a scanner and the scanner says, "Here are your 140 units," and it gives you that work.

It could be that these 140 units are sitting in four areas, and the warehouse has 10 areas but now we've found a batch for you in only four areas. Still, you're going with this cart through these four areas, go to every bin, look where the item is, you scan the item, you scan the bin. In order to make sure it's from the right end, you put it on the cart. Go to the next, it shows you now go to P2 and you go, following the pick pass, the logic pick pass which is also sophisticated because you don't want to jump right, left, u, down. You need to follow a certain sequence. That would be, for example, now you're talking when it's used in a big, big site, maybe now you're picking 110, 120 units an hour.

Now, when you're more automated and now would have a trolley, you have basically conveyors from the pick tower, and you have now multiple pickers per batch. That means the system uses there's a picker is staying in zone one and another picker is staying in zone two. All you do is, in the one you pick, the system says, "Multi batch A, scan the code." Then you start picking. Only in this one area but that means now, it's just taking for this batch, for the 140 units you're just taking eight of the 140. When every item of this batch is over you first, you put it on the conveyor then you go for the next batch.

So as Amazon scales and volume increases, you get scale in picking as you can introduce greater density and more pickers per area of the warehouse then?

Exactly. One thing is, first of all, you can influence where you store stuff. You need to present everything as best as possible to the pickers. This is where I said for the stowers, this is where a student can learn a lot where you can present stuff. What products are strategically good to have where, how can you reach density? How can you store them? How can you have 8, 9, 10 layers in a shelf because the denser the better.

How do you do that? You have to realize customer buying patterns between certain products?

Yes, you have for example plans, some incentive plans from like where to certain vendors spend only medium stuff and you know high selling stuff coming in that you get that into certain areas. You know this is picked out firstly, whereas other stuff coming in this is a lot of slow mover in then you pick and you store in the store, the more you sell it the on more the filled areas.

There is also by the way consolidation process, which cost money. You could often time if your dead wood, they call dead wood sitting on the shelf, which means it doesn't turn, and it's just sitting there in the prime location. There is people going in scanning those things picking it out and picking all through a filed area and putting them back into the system but in a more filed location so that there is good areas already available for the good stuff.

Does Amazon really gain an advantage more in multi picking versus the single picking against competitors?

The single picking is a simple operation but the sophistication and software, which they build there for multi gives Amazon a strong competitive advantage.

Can you quantify the difference? In terms of density or UPH?

Well, I never did study against competitors really but I feel that Amazon would be in the 10%, 15% mark.

Of course, it gets better and better the more because there is other systems in Amazon like the Magic in Japan or the AFE the Amazon Fulfillment Engine. Rather than picking it to 140 units as a multi picking, you just pick into this huge raise, thousands, and thousands of items. They are all sorted in a different way, they are all picked into troughs.

They are all coming from the huge station there is a guy in every town and just taking the item out. These items are then going ahead and going to sit in shoots where they are married up together and stuff. There is some sophistication and the idea is always picking into bigger volumes because it created density for picking.

It's really the outbound on the picking, you can really make a difference there, and there are a lot of things to optimize.

Let's take an average order on Amazon, say $50. How does this cost break down for Amazon between these major areas in fulfillment?

In the warehouse as a whole cost I would say 65% outbound and 35% inbound cost. Then within outbound from this case, you have slightly more on the pick than on the pack, and they are both pretty expensive. I'm just thinking of productivities now. Like a single pack they're doing 100 units an hour and the multi pickers doing 180 or so. It could be almost 50/50, but it's both very intensive. Then between you need to know there's another space in between for the multi pick, and one of the main steps you need to get when you pick all those picks, you need to somehow sort the stuff back, you need to sort customer orders. Whereby somebody scans all those 140 units and then sisters with A1V1 and yet different slots from in the end. Everything sitting in A1 would be customer X’s order. Then another slot for another order and so forth. This is the reading process. This is a fast process where they're doing between 400 and 500 an hour on the scanning but it's an important process.

You're talking maybe about, you're talking something like 40% picking, 40% packing. I don't know 15% reading and 5% tote sortation.

So moving onto look at the delivery part out of the warehouse, how does Amazon optimize the final part of the outbound process?

All the decision making, what you pick, when and so forth is based on the stamp that every shipment gets. We calculate it. Everything is organized in order then prioritized by expected ship date. These expected ship dates always reflect a delivery slot or a pickup slot on the dock doors. So here for example ever one pickup at 1920 1921 1922, you have either expected shift of 1920 1921 1922 or for next day or something. That's how everything is prioritized and as you said, it is optimization in terms of the routes. For the classic postal services, they are doing the optimization in house on their end, so lots of their preview. For example, if you get a pickup from DHL or something, Amazon have this location system in the end. So after ship, you have a sort already mostly that it goes to slam and then goes and the down certain shoots for three.

That can be configured differently dependent on what you agreed with the pose and everything. For example, if you already do sort by the postcode areas for the margins or one is more Northern Germany one is west and then so forth, you get those different shoots and then pre-sorted by that because that's what the shipping cost cheaper because you agree certain prices or something like that.

What are the key things for students to really understand about how the process is structured?

I think it would help to look at the reflection of what is important, always the metrics because you need to measure really important, to improve all the time. So I don't mind, and we've discussed various metrics, but of course, this one is been variable cost metric. Which is how many units can you push through the whole site per hour. Then you break it down by inbound and outbound, and that's basically if it's a good indicator of the health of site. Then you have those customer facing metrics like backlog. How many days of backlog do you have?

Amazon Fulfilment Process

How each stage of the fulfilment process fits together for online retailers; inbound to delivery

Amazon Fulfilment Process

Former Global Director of Operations Benchmarking at Amazon

inpractise.com/articles/amazon-fulfilment-process

Why is this interview interesting?

Amazon Fulfilment Process(November 11, 2019)

Marcus Kolb

Former Global Director of Operations Benchmarking at Amazon

Interview Transcript

Related Content

Amazon Retail: A Valuation Approach

Amazon Retail: The Rails of E-commerce

Amazon Retail EMEA Strategy and Profitability

Former Finance Director at Amazon EMEA

Amazon Retail: 1P vs 3P Economics

Former General Manager at Amazon

Marcus Kolb

Former Global Director of Operations Benchmarking at Amazon

Audio

Amazon Fulfilment Process(November 11, 2019)

Related Content

Amazon Retail: A Valuation Approach

Amazon Retail: The Rails of E-commerce

Amazon Retail EMEA Strategy and Profitability

Former Finance Director at Amazon EMEA

Amazon Retail: 1P vs 3P Economics

Former General Manager at Amazon