Lies, Damn Lies, and Twilight Download Statistics

Stacey Higginbotham at GigaOm yesterday received a lot of attention for a post she wrote arguing that that metered broadband suffers a “Twilight Problem” based on a series of calculations based around downloading the movie Twilight. Our own Steven Hodson said that it highlights that “metered broadband will kill online video.”

As I’m lying here typing this, with my son and friend in another room streaming a movie, and my wife lying next to me in bed streaming a TV program, all on one of these “evil” metered plans, I though I might take a look at these “online video” ending statistics.

Higginbotham uses as her example the HD version of Twilight that she clocks in at 3.8gb. I doubt very much if Apple is the most efficient at file delivery (and I’ve argued before that scarcity drives innovation, in this case more efficient delivery) but lets use the 3.8gb as our starting figure.

Higginbotham says that a user on Time Warner’s cheapest plan of $15/ month for 1gb, with $2 for each additional gb will pay a headline grabbing $20.60 in download costs.

Sounds scary doesn’t it. Only two tiny little problems: first, she presumes people who download movies would be retarded enough to sign up to a 1gb plan to begin with; it doesn’t take Sherlock Holmes to work out that if you’re renting movies online, that you’d be better off on a higher plan. But here’s the even better part she didn’t mention.

That 1gb/ $15 plan only comes with speeds on 768 KB/128kbps. Even presuming that you got precisely the maximum 768kbps on your download (and I’ve never seen an internet connection yet that delivers the maximum it promises), this poor person paying $20.60 for their download would also have to wait 11 Hours 48 Minutes 22.28 Seconds (calculated here) for the file to download. Even allowing for a stream starting while the file was still downloading, you’d still have to buffer it for 9 hours to watch it in one sitting.

This is not to say that there are some severely mentally challenged people out there who would find this acceptable, but for the 99.9999% of the population who rents movies online, this is not going to happen because they wouldn’t sign up for it. Nice headline figure, complete BS in real world application.

Hidden Costs

Higginbotham then goes through and allocates the costs of downloading Twilight based on other plans, simply by allocating the percentage of the cap used by the download against the total cap cost.

There’s a problem with that: it presumes that the cap will be used in full. Example:

we’ll use AT&T’s U-verse rate of $55 per month divided by the 150 GB cap it’s said it will implement. That nets out to AT&T charging 36 cents per GB, which means the bandwidth for “Twilight” will cost $1.37. Adding in that $3.99 rental fee means my vampire fixation will cost me $5.36 to watch without leaving my couch.

The problem is that caps rarely are used in full, particularly at this level. Comcast claim that only 0.1% of their users use more than 250gb per month. One figure quoted regularly (example here) says that the average user uses only 2-3GB per month; obviously that won’t be a person who regularly downloads HD movies, but it’s an important point.

Pinning down average downloads of movies is a statistical challenge; in Australia for example, one major online DVD distributor says its “up to” 5 videos a month. If that applied to downloaded movies in the AT&T example, that’s 19gb per month. If it’s 10 movies, that’s 38gb per month. 20 goes to 72gb, 40 goes to 144gb, so you’d be watching 1.3 HD movies every day downloaded from the net to get up to your cap. Given that for a select few 40 movies a month is unlikely, the real cost of each download is actually higher, because the cost of the download would be the percentage of data used for the month, not the cap.

But the same economics apply for unlimited plans.

The equivalent 10mbps unlimited plan on Verizon costs nearly the same as AT&T’s 150gb capped plan (AT&T is $59.90, Verizon $59.90 without phone, $57.90 with phone without a one year contract.)

The only way you would save with the uncapped plan is to watch more than 40 HD Twilights, or to consume enough bandwidth doing other things (online gaming is surprisingly efficient despite claims otherwise) to use MORE than 150gb.

And yet this is what Higginbotham writes:

Verizon: Since Verizon doesn’t meter or cap its service (or plan to), the cost to watch Twilight only reflects the $3.99 rental fee. A user still pays for broadband (as they do even without a metered plan), but without a limit on data downloads, its impossible to calculate a per-GB cost for downloading content.

You can calculate the cost if you make a presumption about how many movies will be watched and draw a figure for non-movie consumption, not exactly challenging stuff, particularly given the other presumptions. But here’s where it gets better. Lets say you only want to watch your five movies (presumably all in HD), do some regular surfing, maybe some VOIP. Will you still save money with a uncapped plan?

Time Warner offers its Road Runner Standard package for $31.95/ month unbundled, or less in a bundle. It comes with 40gb per month. If you downloaded 5 HD movies, played some games, did some surfing, watched some videos on YouTube, you might just get to 40gb if you were trying really hard. You instantly save $28 over the “uncapped” plan with Verizon.

But is 40gb a month enough? As I’ve written previously, despite living in a house with 3x more internet connected devices than people, waking up to streaming radio on my Chumby, working from home online, including uploads of videos (note in Australia they count uploads to a cap), my son comes home from school and watches YouTube videos and plays online games, and then nearly exclusively we only consume online media of an evening, on a busy month we might hit 50-60gb. On average we do 30-40gb per month. But we’re not average: very few households would have anywhere near the non-stop internet usage we have across as many devices.

The world is not ending, and statistics can always be what you want them to be if you ignore certain facts that get in the way of the picture you are trying to paint, as is the case of the Twilight problem stats.

In an ideal world, there would be no caps; however in a free market where the cost of wholesale access is not unlimited, retail providers are free to pick their models. If people really feel so strongly about the rates being charged, don’t pay for them; if you’re in a large city, you’ll usually have a choice. But note this one thing: a one size fits all isn’t necessarily the superior model for all consumers. Uncapped plans can actually end up costing significantly more, and I’m yet to see a decent argument stating why users who consume the most should be subsidized by those who use little.