There’s been quite a surprising result from T-Mobile USA’s unusual tactic to offer free, unlimited video streaming to its customers. Traffic levels are said to have reduced in some cases, very much counter to intuitive expectations. We ask why, investigate what’s happening and consider some of the wider implications.
T-Mobile’s Binge-On Offer
Available to post-paid subscribers on a 6 Gbyte data plan or more ($65/month SIM only), video viewed from a range of popular sources is free for unlimited use.
It's opt-out, not opt-in.
Channels include films (Netflix, HBO), sports (Fox, NBC, Major League) and remote pre-recorded/time-shifted (Slingbox). Notable exclusions are Youtube, Facebook and Vimeo.
For those opting in to Binge-On, data throughput rates are capped at 1.5Mbps. That’s not just for those video channels above, but also for any other video streamed or downloaded. Non-video traffic may also operate at a reduced speed.
The impact this may have for end users is that when watching HD video on websites that don’t automatically adapt/adjust video data rates for slower connections, it may become choppy and break-up. Many sites do adjust though, especially the popular ones such as Youtube and Facebook.
There’s been quite a furore about the throttling in the media, worries that Net Neutrality is at threat. The Electronic Frontier Foundation published their test results illustrating the network behaviour. One startup video service has withdrawn from the program.
Arguably, the feature is an opt-in/opt-out service. Customers can choose to remain outside and enjoy full speed, HD video if they want to, using their data allowance. The feature can be turned on and off at will, if customers are savvy enough to understand it. Early reports are that customers who complained about poor quality of video streaming are told this may be due to throttling and have the feature turned off for them.
Video viewed over Wi-Fi is not throttled or constrained.
User behaviour is the key factor
It’s easy to become very focussed on the technology. What’s much more relevant is how people use and react to the service.
There’s a limit to consumption: Just as unlimited voice minutes don’t generally get abused (there’s only so much you can talk before you bore your friends and family), so there is a limit to how many hours you can sit and watch video. Other data Apps may continue to burn through data in the background, or make frequent status checks.
Snack video vs broadcast programs: Mobile devices are popular for watching a few minutes of video as part of social media, but I reckon less commonly used to watch entire films or TV programs. Most of us would prefer to watch those on a larger screen TV. Few would have blocks of time to watch those during the day when at work. Unlimited services on offer are generally oriented to longer, broadcast TV or films.
Much video is watched at home, where Wi-Fi is available: Many mobile devices would switch to domestic Wi-Fi when at home, and probably a higher proportion of those used by those who watch lots of video. Often (not always) resulting in faster speeds. Video is not reduced in quality by T-Mobile when viewed over Wi-Fi.
Car trips might be a high consumer: I’m told that cellular music streaming is very popular in cars throughout the US today. While drivers can’t watch video in the car themselves, DVD players are quite common to keep the kids quiet for those longer family trips. It remains to be seen how popular that becomes. Delivering high data rates to fast moving cars over cellular is much more expensive than in many static urban environments. If network performance isn’t up to it throughout the entire journey, then more likely this wouldn’t take off.
Video streaming is one of the most demanding and highest consumers of mobile data. Limiting it greatly increases network traffic capacity. We’ve argued before that the key metric of network capacity is available bandwidth (Mbps) at a specific location, not traffic handled (Mbytes). By constraining data rates for every user, the peak number of concurrent video channels that can be served increases.
The downside for the technical team is what the increase is in video consumption, and whether the savings in data rates more than offset it.
Presumably a net benefit is what T-Mobile’s strategists are banking on.
Longer term implications
Data pricing innovation likely: Gbytes of data don’t mean a lot to most consumers. The ability to separate off and price specific services for a fixed fee may be something we see more of in future. That might be anathema to those keen to preserve Internet neutrality, but I’d expect we’ll see pricing innovation linked to specific applications and M2M devices in the future.
Network traffic demand difficult to predict. As behaviour changes and video becomes more popular, strains on network capacity will evolve. Pricing and marketing policies above will further complicate matters, probably in ways that are difficult to foresee. Network densification will be essential, but even more so the flexibility to deploy additional capacity quickly where is it most needed.
Holistic approach essential: Network engineering has evolved in recent years to take account of complete end-to-end performance and cost, rather than being silo-based. There needs to be similar close co-operation with marketing teams to ensure that user perception of high value can be translated into cost effective network service. As with any large business operation, those networks that make the best choices and/or have the agility to react to market changes most quickly stand to benefit most.