Thursday, December 11, 2014

Gigabit Access Improves User Experience, But Only So Much

The growing embrace of faster Internet access in the U.S. market, and the growing supply of faster connections, despite all criticisms, has many drivers--on both the demand and supply side. But it is changes on the supply side that arguably are most important.

In the past, Internet service providers rightly have suggested there was not much demand for faster access (50 Mbps, 100 Mbps, 300 Mbps or gigabit Internet access), but that was in large part primarily because of retail prices.

It would have been more accurate to say there was, in the past, not much demand for Internet access costing $100 to $300 a month, when “good enough” services could be purchased for $40 to $60 a month.

All that is changing. When Google Fiber launched Google Fiber, offering a symmetrical access service for $70 a month, the key innovation might arguably have been the price.

True, Google Fiber offers very-fast access (two orders of magnitude faster than typical offers, and an order of magnitude faster than 300-Mbps services that had been available in some communities.

But the key innovation arguably was the price. A gigabit for $70 a month is not such a leap from the $50 a month level of “standard” Internet access.

Also, $70 a month is a huge change from the $300-a-month price of prior 300 Mbps offers, or $350 monthly prices for a few gigabit services that had been available.

So major changes on the supply side--dramatically lower prices and dramatically faster speeds--are spurring demand.  

In part, that is because the underlying technology is getting better.

“I joined AT&T in 2008 and I remember around 2012 looking at some charts and the cost of speed hadn’t really had a breakthrough, because 80 percent of your deployment in broadband is labor based,” said John Donovan, AT&T senior executive vice president for architecture, technology and operations.

“And then all of a sudden you have vectoring in small form factor stuff and all of a sudden a little bit of an investment by our supply chain a few standard things and we start to take a 25 meg on a copper pair and then we move it to 45 and then 75 and then 100 which is on the drawing board,” said Donovan.

The point is that the underlying technology used by cable TV operators and telcos has been continually improved, providing better performance at prices useful for commercial deployment.

Operating practices also are becoming more efficient. Google Fiber has been able to work with local governments to streamline permitting processes and other make-ready work in ways that can lower costs to activate a new Internet access network using fixed media.

Google Fiber also pioneered a new way of building networks, getting users to indicate interest before construction starts, and building neighborhood by neighborhood, instead of everywhere in a local area.

That changes gigabit network economics. As has been true for nearly a couple of decades in the U.S. market, for example, competitive suppliers have been able to “cherry pick” operations, building only enough network to reach willing customers, without the need to invest capital in networks and elements that “reach everyone.”

That makes a big difference in business models. A network upgrade that might not have made sense if applied across a whole metro network might well make sense in some parts of a city, where there is demand.

Also, every new supplier of Internet access goes through a learning curve, generally operating inefficiently at first, but improving as experience is accumulated.

“And then we are getting better at the deployment side of the business as well,” said Donovan. “So our average technicians and our best technicians are converging.”

But there is an important related issue. Customers who get gigabit service often cannot perceive a difference in experience as great as they might have expected.

“When we do the installs, we often have to stay and show them on a speed test is getting a gig,” said Donovan. That illustrates a problem we are going to be seeing more often, namely that gigabit access can only improve end user experience so much.

Access speed only improves experience so much because it is only part of the combined ecosystem, and only can affect a part of user experience. Remove the local access bottleneck on one end and all the other elements become visible.

Bragging rights and therefore marketing advantages are believed to accrue to Internet service providers with the highest perceived Internet access speeds.

Some--if not most--of that marketing hype apparently is misplaced, a new study by Ofcom, the U.K. communications regulator, might suggest.

The study found that “access speed” matters substantially at downstream speeds of 5 Mbps and lower. In other words, “speed matters” for user experience when overall access speed is low.

For downstream speeds of 5 Mbps to 10 Mbps, the downstream speed matters somewhat.

But at 10 Mbps or faster speeds, the actual downstream speed has negligible to no impact on
end user experience. Since the average downstream speed in the United Kingdom now is about 23 Mbps, higher speeds--whatever the perceived marketing advantages--have scant impact on end user application experience. Some 85 percent of U.K. fixed network Internet access customers have service at 10 Mbps or faster.

Investing too much in high speed access is, as a business issue as investing too little. The important insight is that it is perception that now matters most in the United Kingdom and United States, not the actual threshold required to provide reasonable end user experience of Internet applications such as web browsing and streaming video.

Average access speeds in the United States are 10 Mbps, according to Akamai. Average speeds are 32 Mbps, according to Ookla. Another study shows that average Internet access speeds in the United Kingdom and United States are equivalent, in fact.  

The quality of the upstream path and in-home network have some impact, at all speed ranges, but at a dramatically lower level as speeds climb above 10 Mbps.  

One finding was surprising. The Ofcom tests of end-to-end user experience suggest that web browsing is significantly affected by upstream and downstream access speeds, the home network and the Internet service provider’s network interconnection policies.

Both the upstream and downstream speeds affect user experience of streaming video, while voice experience is, relatively speaking, barely affected.

Those are important findings. The quality of the broadband experience is not solely dependent on
access speed. In-home wiring (including Wi-Fi performance) and peering arrangements etween internet service providers can also be important.

“Indeed, for connections with a download speed greater than 10 Mbps, access speed appears to become less significant than these other factors,” Ofcom says.

At connection speeds above the range of 5 Mbps to 10 Mbps, though, the relationship breaks down and broadband connection speed is no longer an important determinant of performance, Ofcom says.

The important observation is that elements of the end-to-end value chain--other than access speed--now are becoming greater bottlenecks.

No comments:

Many Winners and Losers from Generative AI

Perhaps there is no contradiction between low historical total factor annual productivity gains and high expected generative artificial inte...