Price – Who’re You Calling a Freeloader?

One of the major objections to flat-rate pricing of communication services, including Internet access, is the danger that a small percentage of users will absorb a huge percentage of the resources either degrading service or raising prices for everyone else. Providers call these users “freeloaders” even though they are paying something.   

It’s easy to make a hypothetical case for this argument; but, in my experience, freeloading is not usually a problem even when a service is unmetered.  In fact, so-called freeloaders may be important to the value of a network service.

Network engineers know that you build a network for its anticipated peak load (actually just for the xxth percentile peak load depending on what kind of reputation you want).  Since most networks including the power grid, the phone network, and the Internet have peak loads far in excess of their average load, most network resources are idle most of the time.  BTW, this is one of the reasons that WiFi modem sharing (see previous post) can work well.

Unlike the power grid which must be fed expensive power, the cost of running a communication network (other than customer service) is almost independent of how much the network is used.  Much of the cost for communications networks is in amortization of the original capital required to acquire rights-of-way and to build the infrastructure.  There are also operating costs for a network control center, maintenance, and repair costs; but these costs aren’t affected by actual usage – electronic routers don’t wear out after so many packets and switches no longer have mechanical parts subject to wear.  Electronic stuff ages and gets obsolete whether it’s used or not.

In the US (except New York City) we’ve had unmetered local phone service as long as I can remember (and I’m old).  Did some people talk more than others?  Of course.  Did this make local phone service terribly expensive or cause long waits for a dial tone? No. (In all fairness, there are both governmental subsidies and a subsidy from long distance which reduce the cost of some local service so the argument here is a little shaky).

More recent experience is with flat rate all-you-can-eat dialup access pricing which I helped to popularize when I started AT&T WorldNet Service.  I wish I could take the credit (or blame) for inventing it but I didn’t.  I just latched onto it and used the AT&T brand to popularize it.  We had plenty of backbone bandwidth available from several previous failed AT&T services but we didn’t have an unlimited supply of the ports that growl at your modem when it dials in (remember those days?).  We had to buy equipment for each port and we had to rent a line from our then rivals, the baby Bells, for each port.  Ouch.

Besides brand, one of our main selling points over our well-established rival AOL was that they were then known as “America on hold” both because of the many busy signals encountered when trying to dial into them and the long waits for customer service when you wanted to complain about the busy signals. 

We were determined that users would usually (80% of the time during peak periods) find an open port the first time they dialed in.  So we had to have enough ports.  We also engineered a pretty clever hack to bulk haul calls from one place to another as peaks hours rolled across the country.

We were far from perfect.  We did have cities where we were short of capacity.  We did find that we had to provision more than the one port for every twenty registered users we’d baked into our original business plan.  And, at first, we blamed the freeloaders, the people who stayed online tying up a port 24x7.  I forget the actual numbers but it was something like 5% of our users accounting for 50% of our use.  What to do?

Fortunately, before we took any drastic action, we looked at the PEAK hour numbers.  Turns out that there were lots of users who only showed up during peak hours.  Many were on our hourly plan but we weren’t billing them a lot because they didn’t use many TOTAL hours.  But the size of our network and the cost of running it were only affected by the peak hours used, not the total hours.  Eliminating flat rate pricing would not have solved our problem and it would have driven some of our most loyal customers away.  The so-called freeloaders accounted for a much smaller percentage of peak hour usage than they did of total usage.  Obvious when you think about it since everyone else is also using the network during busy hour.

Broadband providers don’t need to provision unique physical ports for all of their users.  They recognize this by encouraging their users to be continuously connected.  But broadband providers also have no incremental cost for actual usage.  Doesn’t really matter if you download a little or download a lot – except during peak hours.  They still have to build their networks for peaks.

In fact, this same logic holds for phone networks – wired or unwired.  All the actual costs depend on the size of the network and not its use.  That’s why we’re starting to have flat rate long distance and even international calling plans.  The only reason we have ANY metered phone plans left is that, in many parts of the world, monopolies control the last mile of access to the called party.  These monopolies CHOOSE to charge for access to this last mile by the minute because that’s lucrative – but their network costs are determined only by peak hour demand.

You could make a case for charging differently for peak and off peak access.  The difference would have to be significant to get customers to accept the complexity.  Some providers do this simply by tacitly allowing their customers to use more bandwidth than they’ve paid for except during peak periods.

You can certainly make a case for the current practice in many places including the US of gating access by the width of the pipe.  Makes sense to me to pay more for 10 meg (if I could get it) than for 2 meg.  The telephone equivalent is the single line.  You can’t use it to call for more than 1440 minutes per day.  If you know how much access you’ve sold, you can guess how much backbone you’ll need to provision.

Now back to the freeloaders that carriers love to hate.  If someone keeps an access line humming 7x24 as part of a file sharing network, she is probably doing the network operators a favor by moving files around at off-peak times so that they will be more easily accessible with shorter network paths during peak times.

She, the server operator, is certainly making the network more valuable by adding content to it that all users can access.

The first post in this series on price is about the cost of complexity.

The second post is about FON, the WiFi sharing scheme, as a great experiment in pricing.

Comments

[Tom Evslin]

Great bunch of comments. Thanks

Brian: Cellular calls are drifting towards flat rates with bundle pricing. But they will be most resistant since the providers have good customer lockin - so far. Look for mixed mode phones including wifi to break that.

Jim: You are correct that there are sometimes explicit peering costs, especially in voice. Subject of an upcoming post.

Bob: I should have been clearer. In capital costs I meant to include everything that a carrier would capitalize - not just the cost of initial build but also the the later capital cost of expansion.

Scott: I think the "extra" that customers will pay is actually the disaggregation of today's cable bill - some for faster access and some for paid content - but the total less than the cable bill is today.

Tom

[Scott]

I agree with most of your points, but there is an important difference between dial-up and broadband pricing models that is worth noting. Broadband access pricing depends upon a model whereby users’ actual transfer volume, even during peak period, is a very small percentage of available bandwidth. A sample user may download 100Mb of data in an hour, using the backbone for 35 seconds during the hour (assuming full throttle on a 3Mb port). As a result, peak traffic on a backbone is several orders of magnitude less than the aggregate potential volume of the broadband ports it supports. This allows ISPs to provide very high speed connections at a low cost. New apps, notably IPTV, will drastically alter the financial model because the data volume per unit time is much higher. Backbones will have to be scaled up considerably, increasing port pricing. The question becomes who pays - the consumer, the content provider or the government (through subsidies). And while the telco’s motives are suspect for other reasons, my guess is that they believe consumers will not pay extra for the additional data volume IPTV creates.

[Bob]

"Much of the cost for communications networks is in amortization of the original capital required to acquire rights-of-way and to build the infrastructure". Is this entirely accurate for wireless? With wireless, as usage increases, carriers split cells with lower power towers to increase bandwidth usage within smaller areas. I've heard some ancedotal estimates that with wireline access, 75% of the capital cost was in initial deployment, and 25% was in scaling for load, but the exact opposite was true in the wireless case. Full disclosure - as an application developer, I'm a big fan of flat rate pricing, but I do want to understand the accurate scaling costs of the wireless network.

[Jim]

Your point is compelling, but the statements "the cost of running a communication network (other than customer service) is almost independent of how much the network is used" and "network costs are determined only by peak hour demand" are not exactly right. Due to peering agreements between ISPs, changing your total data transfer amounts with regard to your physical capacity can in fact change your costs.

[brian]

if flat rate pricing brings in the heavy users but doesn't necessarily cost the provider more is there a reason why celluar calls are still rated by the minute or that is isn't the option for flat rates?