Usage-based pricing

May 24, 2002 12:00 PM

Much debate has arisen regarding pricing for broadband service, particularly high-speed Internet service. While other business factors have harmed broadband service providers, the model of flat-rate pricing has left networks open for abuse by heavy bandwidth users. This has created a situation where the cost of providing service can surpass the revenue that is generated. There is no simple solution to this problem, but the answer may lie in migration to usage-based pricing models.

In contrast to the public network, which has a clear-cut business model based upon the exchange of minutes of usage, the Internet has been built around the "Sender Keep All (SKA)" model. Under an SKA plan, various service providers agree to interconnect their networks for the benefit of all subscribers, with each service provider retaining 100% of the connection fees for their own subscribers. For simplicity, most ISPs have adopted a flat-rate pricing structure for broadband Internet connectivity to consumers. Under the flat-rate plan, service providers are motivated to deter subscribers from using the network, since doing so maximizes the potential oversell. Common policies include prohibiting multiple computers from using the same access line or reducing available bandwidth for a subscriber if usage is too high. However, by migrating to a usage-based pricing plan, service providers can shift from discouraging network usage to maximizing profits by encouraging network usage.

Any implementation of a usage-based pricing strategy is the sole discretion of a specific access network or single service provider (because of the SKA model). There are three basic metrics, which can be used to develop a usage-based pricing model: megabytes transferred, time online and percentile pricing.

• Megabytes transferred--Many ISPs during the late 1990s attempted to institute a per-megabyte pricing model to protect themselves against potential loss by heavy users of the network. Usually, a ceiling was placed on total data transfer allowed per month between 500 MB and 2 MB, which, if exceeded, was billed at a variable rate that could be as high as $0.75/MB. The major problem with pricing based on megabytes is that no clear distinction can be made with respect to who is the sender and receiver of Internet traffic. Consequently, subscribers are charged for information they do not want or may not have requested, such as "spam" emails. Furthermore, as Web developers continue to add graphically rich content to Web sites, this subsequently causes users to experience higher access charges. 

• Time online--Pricing models that measure subscribers' time online (in minutes or hours) are basically an extension of circuit-switched pricing models such as those used in the public network. Time online is a technically effective billing method for dial-up networks, where network capacity is determined based upon the number of simultaneous connections supported. Dial-up networks are dependent upon circuit-switching to establish their connection to users, and therefore billing based on hours of usage is a fair and viable method for equating network costs to subscriber usage. However, for broadband connectivity, which does not rely on circuit switching, but rather provides dedicated access (full-time), billing based upon hours or minutes online becomes irrelevant.

• Percentile--The most abstract and least common of all pricing models, percentile pricing uses statistical metrics to define network usage, and therefore the price. Percentile pricing has primarily been deployed in data center operations as a method of billing co-location and hosting customers for the backbone bandwidth they consume. Very few, if any, Internet access providers have based access pricing on percentile measurements, although many top-tier ISPs collect percentile data for network planning and engineering purposes. The most common implementation is the 95th Percentile plan.  Under this plan, the customer is given the highest port-speed connection available and is free to use as much as they like. Data is collected on average bandwidth consumption over 5 minute intervals. After all the data is collected over the period of one month (8640 data points) the top 5% is tossed out as peak traffic. The next highest data point, which is the 95th percentile, is then used to bill the customer. Essentially, the customer receives 36 hours of bursting for "free."

Which of the above models is ideal? Ideal is quite subjective, but I believe the greatest potential resides with migrating to a percentile model. Percentile pricing is a very attractive proposition that can yield higher margins for service providers while at the same time providing greater flexibility and functionality to the customer. Customers pay for the ability to have uncapped bandwidth at their fingertips, while service providers derive revenue based on the 95th highest measurement, whereas their costs are based on the average bandwidth usage (50th percentile). What emerges is a true win-win scenario.

While most broadband service providers have been preoccupied with attempting to prevent unauthorized overuse of their networks, perhaps their efforts would be better spent educating the public as to the benefits of usage-based models. In time, subscribers will come to accept it just as they accept the kilowatt-hour; how many individuals really know what a kilowatt-hour means? Rather than limiting the functionality of the broadband access line by reducing its port speed from multiple megabits per second to hundreds of kilobits per second, why not give us an open pipe and let us consume that which we desire? I certainly am willing to accept such a deal, just bill me.

I have to go now, the phone is ringing again; maybe it's the electric company calling to tell me to stop using my toaster.