Blogs

RedMonk

Skip to content

The Coming Mobile Data Apocalypse

When Apple launched the iPhone in 2007, they correctly anticipated that having what was a reasonable proxy of the real internet on their phone would cause users to consume more data. It can be assumed, therefore, that it was Apple rather than AT&T who pushed for and got unlimited data plans for their customers, to avoid the overage horror stories that would have been the inevitable result of a launch with traditional plans. Thus it was that every iPhone could browse the web with the comfort of knowing that they had, at least in theory, unlimited data.

By now, everyone knows how that turned out.

AT&T’s network was overwhelmed by the “Hummer” of cellphones, its broader network availability declined, and its brand became synonymous with poor performance. Paradoxically, the exclusivity that AT&T bargained for and which won it millions of new subscribers became a liability for the carrier.

From an industry perspective, there are essentially two responses to this problem. One, overprovision network capacity. Two, end unlimited data plans and cap usage. The wireless carriers, at least domestically and in the short term, are trending towards the latter. Whether the primary driver is network reliability or ARPU is debatable, but the end result is inarguable: bandwidth costs are going up, dramatically.

While AT&T would have placed me among the 98% of regular users, the usage patterns of the original iPhone bandwidth hogs – the ones who brought AT&T’s network to its knees – probably looked something like mine. Massive consumption of webpages and regular usage of streaming services, intermingled with higher bandwidth video tasks. Mostly YouTube, back in the days before the availability of NetFlix on what is now called iOS. Remember, the original iPhone didn’t even launch with 3G: it was EDGE only. How much video do you think people were really consuming over a connection that’s a single digit multiple of dialup speeds?

How much data does that usage represent? For a six month period for me last winter, it looked like an average of 624 megabytes a month.

AT&T Data Usage
This year, my average consumption from August through December on my grandfathered “unlimited” data plan was up to 1.2 GB. Still comfortably under the 2 GB cap of AT&T’s new tiered medium sized data plan, in other words.

Then came January, when I discovered NetFlix Watch Instantly.

Pre/Post NetFlix Watch Instantly

The jump in this graph is attributable to NetFlix’s online movies, but it just as easily could be MLB.tv consumption (assuming they would remove their antiquated blackout restrictions). Or Hulu Plus. Or Rdio. Or Amazon Unlimited Instant Videos. Or Google Music, if or when that gets here.

Expanding device storage capacities notwithstanding, streaming is going to define the mobile experience for many, either because the content is being transmitted live (e.g. MLB.tv) or because you’re only renting it (NetFlix, Rdio, et al). Or at least it would, if mobile pricing permitted it. Which it most certainly does not.

AT&T, of course, no longer has an unlimited data plan for smartphones; their largest is 4 GB ($45/mo) with a $10/GB overage. Of the major domestic carriers, three have unlimited dataplans for smartphones still available: Sprint, T-Mobile and Verizon. What’s not clear is just how unlimited their unlimited data plans really are. Here’s the fine print from T-Mobile’s T’s&C’s, for example:

To provide the best network experience for all of our customers we may temporarily reduce data throughput for a small fraction of customers who use a disproportionate amount of bandwidth. Your data session, plan, or service may be suspended, terminated, or restricted for significant roaming or if you use your service in a way that interferes with our network or ability to provide quality service to other users.

The available plans for laptop or tablet users are even less user friendly. Precisely none of the carriers dare provide unlimited connections for devices larger than a smartphone.

Verizon’s largest non-smartphone mobile data plan is 10 GB ($80/mo), with a $10/GB overage fee. AT&T’s is 5 GB ($60/mo), with a brutal $0.05/MB overage fee; Sprint’s is the same at $59.99/mo. T-Mobile’s basic plan isn’t much better, with their maximum size plan capped at 5 GB ($39.99), but at least they have no overages: after 5 GB, “data speeds may be reduced.”

To put this in context, here is what my January bandwidth consumption would have cost on the various carrier’s mobile broadband plans, assuming an overage of 24.8 GB against 5 GB service plans. Yes, the X axis is in dollars.

Mobile Broad Band Overage Costs

The irony of the trendline here is clear from a customer perspective: the demise of unlimited data plans roughly coincided with the introduction of services (NetFlix, Hulu+, etc) that made them attractive. Network operators must take measures to secure their networks, of course, but it is worth questioning whether the carriers are over rotating in an effort to not become the next AT&T. There is little debate that the iPhone swamped an underprovisioned network; the question instead is how much of this was driven by the artificial market factor of Apple and AT&T’s exclusive deal rather than organic growth in data consumption.

Whatever the answer to that question, it is clear that the appetite for mobile bandwidth will grow exponentially over the next twelve to eighteen months. With high volumes of smartphones shipping, more and larger form factors entering the market, and the accelerating build out of streaming services, bandwith consumption is set to spike. Equally apparent is that the carriers are ill provisioned to address this demand, both from a network capacity perspective as well as with their pricing structures.

The takeaways for developers?

  • For better or for worse, data consumption is going to be a user concern moving forward. Applications that assist users in their management of same will find a market.
  • Developers should aggressively explore native DRM frameworks, because if content streaming is cost prohibitive while on mobile data, local secured storage may be advantaged.
  • Developers of streaming applications should a.) heavily optimize for minimal bandwidth, b.) be cognizant of the differentiated data pricing between device types (e.g. smartphones vs tablets), and c.) be network status (wifi or 3G) aware. Customers that are presented with large overage fees due to the consumption of streaming services are unlikely to be happy customers.
  • Even for web applications, caching, minifying and offline persistence – historically important to maximize performance on mobile clients – will remain so to minimize bandwidth consumption.
  • For certain use cases, native clients may be advantaged against network enabled alternatives.

For everyone else? Repent, for the mobile data apocalypse is upon us.

Categories: Mobile.

Tags: , , ,

  • Danno

    I’d like to, again, point out the irony that if they charged, at most, an order of magnitude more than their amortized and realized cost for each byte then there would be no problem.

    Though, uhh, you might’ve gone a *little* overboard with the Netflix there in January.

  • Heather

    SOG — can you give us a ballpark idea of how many hours of netflix that spike represents?

  • Pingback: A Big Problem: Carrier Support for the iPad 2 and Other Mobile Cloud Devices | JetLib News

  • Pingback: Carrier Support for the iPad 2 and Other Mobile Cloud Devices | Finding Out About

  • Chuckie

    Hi Stephen,

    Thanks for calling out what the carriers and carrier vendors have been saying for the last five years. Perhaps devs will listen if it comes from outside of the industry, because they certainly haven’t shown any signs of listening to those within the industry.

    Due to the cost and lead time required to add base stations, most (if not all) carriers build out on a just in time basis and that will likely not change. AT&T has stated that nobody foresaw how popular the iPhone would be nor how it adversely affect network capacity planning.

    There never really were any “unlimited” plans. Most were capped at 5GB and that was felt to be adequate. Now all of the carriers have learned a hard lesson (at AT&T’s expense) about what real mobile data can look like on their networks. Hence the appearance of tiered pricing for LTE networks (which is likely to find its way into the non-LTE customer base at some point, imho).

    I do agree with Danno though, you might have been just a wee bit aggressive with the Netflix in January… ;-)

  • Pingback: La economía de la escasez del ancho de banda | La Pastilla Roja

  • David Hamilton

    Interesting post.

    I think that one solution to this issue that the TelCos will deploy are ‘open femtos’. The are mini-cells deployed in, say, public buildings and which then channel the traffic via the building’s broadband connection.

    I can see TelCos making a major push with them in the next couple of years in order to help with this problem.

  • grey

    Who are you a shill for? Claiming bw costs are going up when anyone who pays for transit sees the continuing opposite trend? Espousing DRM in 2011 when even Apple realized this is a bad idea and they’re the largest online music retailer in the world?

    Either you’re severely misinformed, or you’re deliberately misleading the public into believing the misinformation that profiteering carriers are of course invested in lying about.

    AT&T et al have their bottom line in mind, not their customer’s best interests. As their cost-of-bandwidth continues to go down in line with transit providers on the backend (e.g. unless they’re acting like comcast and trying to charge people for peering with them; thus causing unneeded strife with level3; and trying to profit off free bw because it’s peering).

    Increased bw utilization does not imply increased costs, unless someone doesn’t know how to arrange peering agreements, or hasn’t renegotiated transit costs, or doesn’t know how to do least-cost-routing; all of which have been standard practices for decades.

    Albeit, hardware costs are a capital expense and you need to keep upgrading your network; but aside from that the cost of running it is basically pure profit with that many subscribers; that they are trying to foist artificial expenses on their subscribers when they’re already profitting madly is simply because they see opportunities to increase profits by charging more and metering, even as their back-end bandwidth costs continue to go down; as they have since before TLG even started in the early 90’s.

  • http://redmonk.com/sogrady sogrady

    @grey: your argument is difficult to parse. let’s take a few of the points in order:

    Who are you a shill for?
    If it’s not obvious, the more obvious explanation might be that I’m not shilling anything.
    Claiming bw costs are going up when anyone who pays for transit sees the continuing opposite trend?
    AT&T used to have an unlimited data plan available for $35.00. 4GB, which I exceed, now costs $45. The cost of bandwidth, then, has gone up from a consumer’s perspective significantly. What the transit costs are is irrelevant to me.
    Espousing DRM in 2011 when even Apple realized this is a bad idea and they’re the largest online music retailer in the world?
    First, Apple realized it was a bad idea for downloadable music only. Video is still DRM’d, as are streamable music resources like Rdio and live content like MLB.tv. Given that none of the latter anticipates operating DRM free within the near term, yes, it seemed reasonable to recommend developers familiarize themselves with native DRM stacks. Regardless of my personal distaste for DRM.
    AT&T et al have their bottom line in mind, not their customer’s best interests.
    Just as I’ve argued above.
    that they are trying to foist artificial expenses on their subscribers when they’re already profitting madly is simply because they see opportunities to increase profits by charging more and metering, even as their back-end bandwidth costs continue to go down; as they have since before TLG even started in the early 90′s.
    Where do you see me arguing otherwise?

    If your point with all of the above is that the carriers are not passing on lower wholesale bandwidth costs to their customers, we don’t disagree. I think you underestimate substantially the spike in consumption, but the point stands either way.

    The point of the above is that whatever you or I may think about the actual costs of mobile data, the facts are that the cost per consumer is higher. Which consumers and developers alike need to adjust to.

  • Pingback: What we’re reading: 3/11/11 | High Tech Forum

  • Andrew

    That Netflix spike is crazy–I have a hard time imagining how to even schedule 26GB of video time into a month of my life. Does this imply you were paying a $10/GB overage cost in January to the tune of about $250? Ouch.

  • http://stevesouders.com/ Steve Souders

    I love the reinforcement of caching, minifying (esp. gzip compression), and offline persistence. Edmunds.com cut their CDN requests by 35% when they added caching headers. Netflix cut their outbound traffic by 50% by turning on gzip. Combined, these techniques could reduce bytes downloaded for a website by nearly 67%. I just found out about Allegorithmic’s game textures reduction technology. There’s a lot of opportunities out there to reduce downloads while maintaining content quality.

  • http://www.allconnect.com/blog jessica@data plans

    Mobile IP is a potential solution in this area as it will allow operators to switch users connections (data, video, VoIP)seamlessly from a 3G to a Wi-Fi network. This removes the need to build extra network capacity and provides users with a better QoS when an alternative network is available (eg. in their home or office).