Erlang 1. a basic unit of telephone traffic, one erlang is equal to one hour of traffic on a circuit. 2. a metric used to fine tune the degree of corporate theft by AOL and others.

Long ago and in a different life, I helped put together a technical proposal from EDS to the great Commonwealth of Massachusetts called REVS (Rapid Eligibility Verification System). Part of that process involved estimated the number of network POPs (Points of Presence) that would be required to meet the RFP requirements. Think of POPs as modems ready to take calls, because that's what they are.

Massachusetts specified that the maximum acceptable busy signal rate would be 2% of total calls. OK. I am fudging there. I don't actually remember what the specific rate was, but since I am only using this example for illustration, it doesn't really matter if it was 1% or 5%. That rate, by the way, is called the blocking factor, a measure of how many calls are blocked from getting into REVS because of busy signals.

I worried about how we would accurately guesstimate the number of modems/phone lines we would need to satisfy the requirements. But I shouldn't have. It turned out to be anything but rocket science. In fact, there were reference books we could turn to that had long ago solved the problem. Solved it a couple of different ways, at that. All thanks to Agner Krarup Erlang, who worked for the Copenhagen Telephone Company in 1908.

The problem Agner needed to solve was the very same one we were faced with in bidding on the REVS project: the number of lines it would take to provide acceptable service for a village in Denmark. A single line for the trunk into a village would be the lowest cost option, but of course then only one person in the village could use the phone at a time. That was unacceptable.

The best performance option would be to have the same number of lines in the trunk as there were phones in the village. That way nobody would ever have a call blocked and have to wait for a line. But the cost would be very high. That was unacceptable, too.

So Agner set about studying the trade-offs between cost and performance. The end result are the tables we turned to in those reference books. They told us exactly how many lines we would need given the expected highest call traffic in any hour of the day and the maximum acceptable blocking factor. Not only that, Erlang created one set of tables (Erlang B) for when callers just kept trying after getting a busy-signal and another set (Erlang C) for when they simply waited for awhile before trying again. It makes a difference in call distribution, and thus in the number of lines required.

We won the contract, by the way. the estimates must have been close enough for government work because it was a profitable one for EDS and Massachusetts health card providers were able to screen patients for eligibility more quickly than before. But that was in the late 80's. Let's fast-forward ten years.

In the late 90s, AOL was losing money. They needed to dramatically boost revenue. The answer was to make a fundamental change in pricing and an all-out push to gain new subscribers. It worked. All those millions of CDs they mailed out began to get used. People wanted the new lower flat rates instead of the steep hourly fees they were used to paying. The only fly in the ointment is that there was no way in hell they could support a massive influx of new users on their existing phone network. There simply weren't enough POPs.

There were complaints almost immediately from subscribers, but AOL turned a deaf ear to them and relentlessly, purposely, unethically, and illegally continued to offer and sell services it could not hope to provide. Finally class action lawsuits began to appear and AOL's hearing began to improve. ComputerWorld reported in January, 1997, that AOL only had 200,000 modems available to support a customer base of 7.5 million. AOL managed to settle the suit(s) by agreeing to partial refunds for a month or two of service, but only for subscribers who bothered to file the required paperwork. It's theft at twice the price they settled for.

The point of my personal history using the Erlang tables is to illustrate how easy and common and routine it is for anyone working in telecommunications to predict need. So easy, in fact, that there is no way in hell AOL could have been unaware of the impact its new subscribers would have. They may have feigned surprise that their infrastructure suddenly became overloaded, but they were stone cold lying through their teeth.

Is this practice uncommon? Hardly. And it's not just limited to ISPs. How many times have you been placed on hold due to "Unexpectedly high call volumes"? You've been lied to, each time. This is how accountants and business people control their costs for everything from tech support to customer service to just about anything. Except sales. It's easy to get connected to sales, and if those lines ever fill up it's amazing just how quickly they can be expanded to meet demand.

Erlang science is used today as a high-tech version of a grocer's thumb on the scale. The company decides how long they can keep you on hold - if they care at all, some hope you just hang up and don't call back - without you switching products or providers or whatever else it is they are selling. That decision drives just not the cost of the number of phones needed, but the number of people needed to answer those calls.

The thing that makes the AOL case so much more egregious than other types of business is that it is the phone time/access that they were selling. For ruthlessly stealing from AOL customers by selling goods it had no prayer of delivering, Steve Case became a wealthy and (sometimes) respected businessman. He probably giggled all the way to the bank over the non-punitive nature of the settlement of the class-action suits.

When he led the merger talks and the merger itself of AOL-Time Warner deal creating a 300 billion dollar monster, Steve Case, the thief and the chief liar at AOL, was named one of the "Men of the Century" by BusinessWeek Magazine. Never mind that he was forced to resign as chairman of the new firm a year or two later, and that his legacy in the post includes charges of accounting irregularities ala Enron and WorldCom.

Based on the success of folks like Case and Gates, I think it is safe to say that if you want your children to be really big and famous business-people, they need to learn to lie, to cheat, and to steal as well as being able to do everything else that is normally required. But that's not all bad, now is it. After all, it prepares them for a life in politics, too.

Internet Resources:

Explanation of Erlang b calculator

AOL: The right kind of busy signals

Frustrated AOL users file lawsuits

An Introduction to Erlang B and Erlang C

Some AOL members still eligible for refunds

Men of the Century