September 29, 2013

Mayday for Mayday?

Jeff Bezos & Kindle Fire HDXBilly Joel famously sang that we didn't start the fire, but Amazon definitely has with their latest Kindle Fire HDX and its Mayday feature. While the feature itself has received lots of praise and amazement, the biggest question people have been asking is, how on earth does Amazon plan on making this thing scale?

Ground control to Major Bezos...

The Mayday feature is a single-touch connection to Amazon's support team, active 365 days a year, 24 hours a day, seven days a week. The service promise is 15 seconds from tapping the button to a service agent coming on to the line. It almost sounds too good to be true and that is what concerns so many people because what seems too good to be true usually is.

This does seem crazy when you think about the current state of technical support, with its long hold times and offshore support. If you've called support anytime in the last few years, you probably ended up waiting a long time to speak with a first level technician who read to you from a script of basic troubleshooting steps you've already tried. After getting through the hold times, the transfers and the seemingly ridiculous questions, your frustration level soaring, you probably ended up without a useful answer. Its no surprise that with poor experiences like that, people are incredulous that Amazon could pull something like this off.

The technical problem

No one seems to be questioning if Amazon can pull this off from a technical standpoint. Their web services (AWS) solution provides all the infrastructure they could possibly need to handle the Mayday service from a backend technology standpoint.

When it comes to the device side of the equation, once again Amazon controls the entire experience, so there is unlikely to be any technology issue. The device displays a small frame video and audio feed of the Amazon agent, streams audio from the device user to the agent along with a real-time screen feed from the user's device.

Amazon already has a great deal of experience with streaming video through their on-demand video service, Instant Streaming. They know how to compress video so that it takes up a minimum amount of bandwidth and since AWS is likely the back end solution here as well, this should pose no problem as it is likely far less of a stress than the number of people who are actively streaming video content from Amazon Instant Streaming.

Similarly, the video stream up from the device is likely highly compressed audio, similar to audio sent through VOIP solutions. Screen sharing services are also easily compressed as the video content doesn't change very often, at least in comparison to something like movie or live streaming, so there should be no issue here, either.

People to people

No, what really leaves so many analysts scratching their heads is how can Amazon staff up enough so that the 15 second response time is achievable for all the many people who are buying these devices. To answer that question, I think we first need to think about how many people are actually buying these devices. This isn't an easy question to answer as Amazon doesn't release sales figures for any of its Kindle devices, but there are some smart people out there measuring their supply chain.

The estimates put Kindle sales at $4.5 billion for 2013. Prices start as low as $229 and go as high as $409 for Kindle Fire MDX models. Lets assume that these represent 60% of sales for all Kindles and have an average selling price of $300. That gives us an estimate of about 9 million devices sold in 2013.

(For the sake of this exercise, I'm not going to deal with seasonality of sales and spreading all those 9 million device sales out over a full year. Lets just assume that these devices have a 3 year lifespan while we're at it. This obviously skews the costs to start earlier than they actually do, but for this exercise of figuring out a question of scale, that really isn't a factor because by the end of the first year, they have to be at scale already.)

But they're not going to sell 9 million devices this year and then discontinue the Mayday service. The next two years they are more likely to sell more devices (lets say 11 million in 2014 and 13 million in 2015), but that growing installed base is also offset somewhat by people learning more about how to use their Kindle Fire devices, so they need less support as the devices age.

Typical call center support rates are in the 2-3% of customers call about their purchase and those individuals who do call, usually end up calling more than just a single time. For the sake of this discussion, lets say that each of these individuals need help three times the first year, twice the second year and once the final year, for a total of 6 calls to support.

In 2015, we now have a total of 23 million devices in the field with around 1 million owners having made inquiries to support. If we consider the breakout from the last paragraph, that means Amazon will receive approximately 2.1 million calls in 2015. Sounds huge, but lets break it down a little further.

Lets assume that each call to support is 5 minutes long. Agents work an 8 hour shift, but they have to deal with email and other administrative stuff, have paid breaks, etc, so they're really only productive for 7 hours at a time. That means an average agent can probably take 6 calls an hour or 42 total calls in a shift.

If we take that 2.1 million calls and divide it by 42 calls per shift, we get 50,000 shifts in a year. If we assume an agent works 50 weeks a year, that comes out to 1,000 agents needed to take care of all the calls in year 3. That's a sizable call center, but nothing massive. We're talking a pretty average for a company the size of Amazon.

Now lets dig into some costs. These agents, for reasons I'll get into below, are not going to be your average agent; they're probably going to be on the high end of agents, so lets give them a going rate of $15/hour. Add in overhead costs (which are probably smaller than most call centers because there is no long distance charge and Amazon, as stated above, owns all their own infrastructure and add in management, reporting & building costs) and lets assume we're up to $20/hour. Divide that hourly rate by 6 calls per hour and you get an average call rate of $3.33 per call. Multiply that by our 2.1 million calls and we get a cost of around $7 million for the year.

Lets go back to our sales figures. During 2015, Amazon is selling 13 million Kindle Fires at an average selling price of $300 for a little under $4 billion in revenue. When you do that math, its a little less than a dollar per Kindle Fire sold to do this kind of support. I think Amazon can probably afford that.

Its all about the ecosystem

So far, we've focused only on the cost side of the equation. I did this intentionally because there is a larger point here that much of the knee-jerk analysis seems to be missing. I say that because, frankly, I missed it in the initial conversation around the launch as well. What is this last piece of information? Simply put, everyone is considering this to be a cost problem and no one is talking about the revenue question. We're all forgetting about the entire reason these devices exist and it has frankly little to do with providing cheap tablet devices to the masses.

The Kindle Fire is to Amazon the best storefront they could ask for. Its always near you, it quickly links its owner to Amazon's entire catalog of products and services and it makes buying something else extremely frictionless.

Why does Amazon make such a big deal about Mayday being so fast? Think about the last time you were in a brick and mortar retail store and you needed a sales associate to help you. You probably spent a couple minutes trying to find one if you were not so lucky to catch one walking by. Amazon is making the process of buying from them as easy as possible, far easier than dealing with a physical location. This is the genius of Mayday.

Lets go back again to those 33 million Kindle Fires with Mayday functionality in 2015. Lets again discuss those 2.1 million support calls with their $3.33 per call cost. If each of those devices is the source of an additional 5 orders per year, with an average of $15 per order, that means Amazon just made an additional $150 million in revenue, all because of Mayday.

Asking the wrong question

Asking how Amazon can scale Mayday is entirely the wrong question. The questions we should be asking are 1) how can Amazon not do this and 2) how long until their competition starts doing this as well? After running through the numbers, I'm actually surprised that Amazon has waited this long to create this feature and now I'm wondering how long it takes them to roll this out to amazon.com.