Amazon Flex

Like Uber, but without the drunk people

Drivers hate dealing with passengers who swear, smoke or throw up in their vehicle. That’s the bet that underlies Amazon’s recent logistics endeavour, Amazon Flex.

Amazon Flex is a delivery system whereby independent drivers sign-up to deliver packages around their local area. It’s like Uber, but without the human passengers. Drivers are incentivised by guaranteed earnings of $18-25 an hour and all without the hassle of late pick-ups, small talk and ungrateful four-star ratings.

Your typical Uber driver

The video below describes how Flex works but, put briefly, it consists of the following[1]:

  • Drivers schedule ‘blocks’ of time (typically 2 hours each) when they’re available to deliver packages
  • At the start of a block, drivers check in to their local Amazon warehouse to pick up packages
  • The Amazon Flex app then routes drivers to various locations to drop off their wares


For Amazon, it just makes sense

Shipping costs have perpetually been a key concern of Amazon. In 2015, Amazon’s shipping costs exceeded $5bn[2] and they have been growing as a proportion of revenue[3]. Experts estimate that last mile (i.e. local delivery) accounts for about 50% of all shipping costs[4]. With Flex, Amazon is able to save around 30% of these last mile costs that would otherwise go to local couriers[5]. Furthermore, Flex gives Amazon greater control of their supply chain to ensure packages are delivered on time and customers are happy.

Creating the platform is made all the easier by the fact that there’s an existing pool of drivers, conditioned by the ride-sharing world, ready to take on this work. Flex has been selective in scaling its operations, targeting dense cities with high Amazon and ride-sharing penetration, thus utilising local network effects already in place.

Amazon Flex has an advantage compared to ridesharing companies

The tension between Amazon Flex and ridesharing services is clear – they’re fighting for the same pool of drivers. On top of this, Uber is also scaling up UberRush, a service which uses independent drivers to deliver packages.

However, Amazon Flex has an advantage because they currently only need to manage one side of the platform. The other side, delivery customers, is all just Amazon. More precisely, it’s selected delivery orders that end customers have placed on and are being fulfilled by Amazon.

Amazon Flex’s current platform ecosystem

Given the volume of these orders will far exceed Flex’s capacity in the near term, Flex can guarantee drivers reliable work and compensation for their scheduled blocks. The economics and reliability are therefore more attractive to drivers than other ridesharing and delivery services that are ‘on demand’ – allowing Flex to attract drivers away from established ridesharing players. (Compare, for example, how the UberRush deliveryman below waited for 5 hours with no work[6])


A steady flow of ‘customer’ demand also allows Flex to schedule drivers in fixed blocks. This provides stability in capacity forecasting and discourages driver multi-homing (at least, multi-homing simultaneously while on a Flex route).

But managing the platform still has its challenges

Despite these advantages, the future isn’t cut and dried. Amazon Flex still needs to manage constituents on all sides of the platform. Drivers are a tough bunch to satisfy and retaining them is critical given the low switching costs for drivers to move to a different delivery or ride-sharing platform. Thankfully for Amazon, driver complaints seem to revolve around their not having enough blocks of work, boding positively for availability of drivers on the Flex platform.


Typical Amazon Flex reviews on Glassdoor

Other issues will also arise if and when Amazon decides to use the Flex platform as a third-party logistics services. In such a world, Amazon would not only need to learn how to grow and serve third-party delivery customers, but also manage the tension between using Flex for their own delivery needs vs. prioritising third-party delivery customers’ requirements.

Amazon Flex’s future ecosystem?

However, with the ability to use deliveries to fill-in the ‘customer’ side of the equation, Flex faces less pressure to immediately scale its third-party delivery customer base. This gives Flex more time to figure out its third-party model and correctly position itself to capture part of the $250bn global courier market[7]. And with Flex’s success, Amazon can progress one step closer towards world domination being earth’s most customer-centric company.











Ethereum: fueling the hype around blockchain?

Student comments on Amazon Flex

  1. Great write up, thanks! I knew about Flex but didn’t know exactly how it worked. I also didn’t realize that Flex could pay so well, I imagine there must be an algorithm that prices it to be comparable to the average Uber earnings. Quite impressive that the Flex app also provides specific GPS directions for where to go, for reason I (naively) thought it the delivery driver would have to plan this out themselves.

    1. Ahaha thanks, Yun – glad it was educational 😀

  2. Great post! Interesting that they are launching in the biggest cities. Intuitively this makes sense for most of Amazon’s pilots, but in this case I would imagine that a Flex service would actually make the most sense in smaller cities with a smaller number of Amazon customers. In bigger cities with a larger Amazon user base, wouldn’t the asset utilization of 3rd party carrier trucks (e.g. – UPS) make more sense?

    1. Hi Felix, this is an interesting question. I see where you’re coming from. From Flex’s perspective, however, if these smaller cities don’t have a big ride-sharing base already, Flex would need to cultivate drivers from scratch. Furthermore, I believe Amazon pays fixed prices per package to the 3PLs. Consequently, it’s advantageous for Amazon to leave the 3PLs with the higher-cost-to-serve areas and start cherry-picking denser locations for their own internal logistics operations. In the long-run, this might leave the 3PLs with higher costs and cause them to increase their prices. But in the long-run, well, who knows what the state of Amazon deliveries will look like?

      1. Meili, this is a great post! Thank you so much for sharing!

        To give another take on Felix’s question, Amazon Flex is currently mostly used on Amazon’s quick delivery services (i.e. Prime Now, Prime Restaurant, etc.) that are only available in large city centers. This allows them to make the most out of the population density and guarantee both short term delivery (2 hours or less) as well as sufficient deliveries within the two hour driver window to make the compensation sustainable for the company.

        1. Thanks, Gil! Yes, this is a good point. Drivers have an option to choose between whether they want to deliver for short-term services (Prime Now, etc.) or regular deliveries

  3. Great post, Meili! I had never heard of Flex, and this is a great summary of how it works.

    I think your observation that Flex competes with Uber (and others, like Postmates and Instacart) for drivers is an interesting observation about how “on demand” delivery markets work and might evolve in the future. Uber’s expansion to deliveries (via UberRush) suggests that drivers might be indifferent (drunk riders aside) to particular delivery tasks that they perform—that delivering passengers, parcels, prepared food, or groceries are fundamentally the same “job.” With that in mind, I’m curious: do you think that the end state of this market may be a dominant platform for virtually all “last mile” deliveries? And if so, will that firm control the consumer experience, or is there still room for other firms to “sit atop” a dominant delivery platform and manage the purchasing process? To put it another way: might we rely on a firm like Amazon or Uber to both order and deliver groceries, or might we place our grocery orders through Instacart, and receive them via Flex or UberRush?

    1. Thanks for your comment, Micah! This is a really good question and one that I’m really struggling to get my head around. I suppose we can try breaking it out into two parts and see what factors influence each:

      1) Will one player dominate all ‘last mile’ deliveries?

      My opinion is that there will still exist multiple platforms for delivery, for two particular reasons:

      First, the job to be done is slightly different for package delivery vs passenger deliver. Passenger delivery is inherently a very ‘on demand’ job where passengers can’t/won’t book cars in advance. On the other hand, package delivery is typically scheduled ahead of time (and should be to maximise delivery efficiency). Attempting to integrate the two types of service is possible but difficult and would probably be quite inefficient.

      Second, I don’t believe delivery platforms is a winner-take-all market. Whether or not it is depends on the strength of network effects and multi-homing, right? Currently, multi-homing is incredibly pervasive in the market both on the driver and the passenger side. Moreover, network effects appear to reach diminishing returns quite quickly. Once you can get a driver in 3 minutes, the incremental benefit in having him arrive in 2 minutes is minimal. (To put a little nuance on it, this network saturation applies to big cities such as New York and Boston where there are far more drivers than would ever be needed to sustain a delivery platform. However in smaller towns, the network of drivers may only be large enough to sustain one platform at critical scale.)

      2) Will delivery players also control the purchasing process?

      I guess answering this requires thinking about the fundamental strategy question: are there sufficient synergies in vertically integrating? Optimistic me sees a version of the world where the answer is ‘yes’. Perhaps one day, Amazon’s data processes will be so sophisticated that they can use information on purchase patterns to route packages at a cost/efficiency that no one can compete with. More realistic me thinks that world is probably quite far away. Moreover, if you think about a company like UberRush that doesn’t have a large retail platform, the synergies aren’t strong enough to justify building a retail platform for the sake of the logistics platform.

      1. Thanks for sharing, Meili—great insights all around! I’d be curious to know how well ridesharing firms like Uber can forecast demand, either immediately in advance of spikes or several days out. I imagine that the ability to do so, combined with the ability to incentivize drivers to plan their availability around those times, would make passenger transportation more like parcel transformation. But it’s hard to see those forecasts matching the power of knowing which actual packages need to be delivered. In any case, thanks again!

  4. Meili – Great post. Had heard of the sad state of UberRush performance, didn’t know it’s so bad.
    You mentioned about last-mile delivery side of Amazon Flex. Wondering if they also are doing / plan to do first-mile pick-up (for Amazon returns) as well? Should help boost the cost savings further!

    1. Thanks, Mohit! I haven’t heard of anything along these lines. It’s a possibility though probably not a huge priority (currently, customers complete the first-mile, no?)

Leave a comment