Amazon is a logistics beast - A detailed teardown
Amazon is a logistics beast - A detailed teardown
Amazon is poised to rule logistics in a post-Covid world
Supply chains globally have been disrupted on account of Covid19. Yet, platform players like Amazon and Alibaba have demonstrated how their approach to logistics and supply chain management is not only scalable but also responsive and resilient to such disruptions.
This issue takes a peek under the hood of Amazon’s logistics capabilities and its potential in a post-Covid world.
In particular, a few insights before we get started
Amazon’s logistics playbook involves two mutually reinforcing flywheels: an asset infrastructure flywheel (Increase asset base and offer as-a-service) and a data-driven predictions flywheel (Gather data and improve predictions).
Mapping capabilities will be a key source of competitive advantage in logistics. Outdoor mapping for deliveries and indoor mapping for warehouse optimization. Amazon is well set up for both.
Amazon’s integration into demand-side data, not just as a retailer, but as a platform, is its key strength in managing utilization of its logistics infrastructure.
There’s a lot of good stuff here.
And if you’d like to discuss the ideas in this post further, please join the discussion on LinkedIn at this link.
Let’s get started…
But first, if you’re receiving this newsletter for the first time, please take a moment and sign up.
Also, check out the Platform Institute - an online masterclass on platforms which covers all these topics in depth.
First… some framing
Amazon’s approach to logistics is a masterclass in balancing vertical integration vs openness.
Amazon demonstrates that platforms don’t have to be asset-light, they just have to be strategic about asset ownership. Amazon increases asset ownership where such assets increase ecosystem dependance, and opens itself at other parts of the value chain where it needs third parties to bring in niche capabilities to complement its own.
Amazon stands to be among the biggest beneficiaries of the ongoing lockdown. Orders are up, boosting the company's revenues an additional $800 million per month. Amazon manages the delivery of the majority of this ecommerce volume without relying on 3PL service providers. It estimates a decrease of approximately $2-$4 in cost per package shipping using its internal network versus utilizing legacy carriers.
Let’s look at the key capabilities that uniquely position Amazon for a play in logistics.
In the age of AI, Prediction is king
Amazon’s integration into demand is one of its strongest control points in logistics.
Why is demand integration such a big deal?
To answer that, let’s look at a simpler example from another company: Netflix.
You could argue that there were many things that drove Netflix’s success in the DVD rental business. But the one thing that Blockbuster could never compete with was the integration of demand-side queuing data (users would add movies that they wanted to watch next into a queue) with a national-scale logistics system. All this queueing data aggregated at a national scale informed Netflix on upcoming demand for DVDs across the country.
Blockbuster could only serve users based on DVD inventory available at a local store. This resulted in:
1) low availability of some titles ( local demand > local supply), and
2) low utilization of other titles (local supply > local demand).
Netflix, on the other hand, could move DVDs to different parts of the US based on where users were queueing those titles. This resulted in higher availability while also having fewer titles idle at any point.
Queueing data improved stocking and resulted in higher utilization and higher availability. It allowed Netflix to serve local demand using national inventory.
Traditional supply chains need to manage the trade-off between utilization and availability. The ability to predict demand solves this trade-off and informs stocking and logistics.
Amazon’s prediction powerhouse
Amazon is well-positioned here, with multiple weapons in its arsenal.
Ecommerce analytics
Amazon uses collaborative filtering and other data analytics to build deep customer profiles. These profiles include behavior data, interest graphs, affinity scores etc. Amazon uses these profiles to predict, suggest, and drive purchases on its platform but also uses this to inform its stocking.
Anticipatory shipping
Amazon's patented anticipatory shipping model predicts precisely where, when, and how much of a particular SKU to be made available at any of its fulfilment centers, so they are ready to ship when ordered. Thus, Amazon can scale operations while ensuring high availability and high utilization.
Store analytics
Amazon uses in-store data to decide on product pricing, inventory management, store layout at Whole Foods, and Amazon Go stores. Sensors placed across the store detect the products that shoppers buy but also their interactions with the store layout. These stores will also double up as local collection points for Amazon's delivery network, especially for essentials in a post-Covid world.
Extra-fast shipping
Amazon also regularly tests what new products people might want with extra-fast shipping, and uses this to inform its stocking better.
SKU optimization
Amazon constantly decides what to stock by looking at every product detail. For instance, it would stock a shirt based on data about color, length, silhouette, sleeve length and purchase history for similar clothing inventory.
Prediction at ecosystem scale
Here’s where this gets more interesting. Amazon doesn’t merely act as the most data-aware store in the world. It also extends this capability to the rest of its ecosystem.
Seller analytics as-a-service
Amazon also packages some of these insights for its 3rd party resellers, enabling them to anticipate and plan for customer behavior. Planning for any customer behavior provides a significant advantage to merchants who usually rely only on post-sales data. To ensure better delivery, Amazon collaborates with its suppliers and manufacturers and tracks their inventory and provides recommendations on inventory management.
Store analytics as-a-service
Amazon has a history of doing something really well, developing scale (in assets and/or data), and then packaging that capability as a service to third parties. It’s does this with AWS (tech infrastructure) and with FBA (warehousing infrastructure). It could do the same thing with store analytics by eventually offering store analytics to third party stores as-a-service.
Supply chain management as-a-service
As manufacturers are move towards leaner operations, and as supplier audits increase in a post-pandemic world, Amazon is well positioned to provide end-to-end logistics and supply chain management as-a-service. Such a system would potentially analyze supplier data (e.g. delivery performance, audits, evaluations, credit scoring etc.) and create a reputation system for suppliers, enabling manufacturers to make better supplier decisions and reconfigure their supplier network. It would also allow suppliers and manufacturers to plug into Amazon’s logistics and delivery infrastructure seamlessly, removing the need for complex procurement.
All of these as-a-service models ensure that Amazon’s capabilities are used at industry scale and all the resultant data constantly trains Amazon’s prediction models, enabling it to develop the long tail of predictions.
If you’re finding this post useful, please share it with your network as well.
And if you’d like to discuss the ideas in this post further, please join the discussion on LinkedIn at this link.
Optimizing package interactions inside the warehouse
Amazon has built a massive warehousing and fulfilment footprint across the US, which now includes
1) Smaller warehouses closer to city centers where Prime Now promotes super-fast delivery options
2) Whole Foods locations for faster access to groceries and essentials
Amazon constantly invests in optimizing package interactions within the warehouses. This includes robots moving shipments inside the warehouses, gesture recognition to identify when a worker has placed a package on a shelf, automatic scanning of items that workers hold in their hands - all geared toward minimising the click-to-ship cycle time. Future patents suggest the use of UAVs for warehouse management and augmented reality enabled eyewear to increase warehouse worker's efficiency.
The rise of indoor mapping
In its bid to maximise warehousing efficiency, Amazon has developed strong indoor mapping capabilities. These indoor mapping capabilities may be eventually rolled out into stores as well. Indoor mapping, as a capability, will likely become more important in a post-Covid world, where contact tracing will require tracking of movement inside indoor spaces.
Amazon’s last mile play
Controlling the last mile is critical for control over customer experience. The last mile makes up ~30% of overall logistics expenses.
Amazon has a host of logistics services in the last mile, including crowdsourced deliveries from external contractors (Amazon Flex and Amazon Logistics), Fresh food delivery (Amazon Fresh), Amazon Key to allow deliveries into your home, delivery to car trunks, remote door access to Amazon couriers, Amazon lockers and apartment hubs (Amazon hubs), and distribution by drone (Prime Air) ensure customer convenience.
Again, data is the reason Amazon gets this right:
Route optimization: Optimal routes for delivery drivers are derived from the data aggregated across customers, drivers, connected vehicles, weather forecasts, traffic monitoring systems, and digital and satellite maps.
Fleet planning: Amazon’s fleet management system calculates how many drivers are needed at any given time. It evaluates the weight and number of packages headed to the same destination and matches packages and destinations to fleet availability. This includes determining the order of packing boxes into a vehicle to enable the most effective unloading based on delivery address.
Mapping metadata: Amazon is gathering delivery metadata that puts mapping features into context. For instance, one big challenge for Amazon Flex delivery personnel is parking. Amazon constantly analyses late deliveries and identifies patterns and correlations with:
1) Building type (single address vs multi-address),
2) Access (Deliver at door vs at reception vs in mailroom)
3) Parking facilities (at building vs not)
By correlating delivery times and delays with these variables, Amazon is creating a new layer of delivery intelligence on top of mapping data. This prediction capability can again be opened as-a-service for third party logistics firms.
Amazon’s logistics playbook
Amazon’s logistics play follows a common playbook that we’ve seen in other parts of Amazon’s business:
Gain asset scale through supply-side integration
Gain data scale through demand-side integration
Leverage supply-side scale to open out asset-as-a-service to ecosystem partners
Leverage data across the ecosystem to constantly improve prediction models
The more asset-as-a-service scales, the larger the ecosystem using Amazon’s logistics infrastructure and the greater the data capture for Amazon to constantly improve its prediction models
This virtuous cycle constantly strengthens Amazon’s logistics play.
If you’d like to discuss the ideas in this post further, please join the discussion on LinkedIn at this link.
—
And if you enjoyed this article, please share further
—
Thanks to Nishant Marskole for contributing to this strategy teardown.