How to (and Why You Should) Perform a Traffic Study
June 19, 2008 (Vol. 29, No. 13)
One New York City-based law firm will save $60,000 a year after it disconnects six unused T-1s – including saving on regulatory fees that shouldn’t have been applied to two of the T-1s in the first place.
Your enterprise probably has too many trunks relative to your traffic volume, too.
That’s the case in as many as 75% of the traffic studies done by Jane Laino, president of DIgby 4 Group Inc., in New York City, who worked with traffic study firm Bristol Capital, in Montvale, N.J., to complete the study for the law firm four months ago.
The main culprits behind chronic over-provisioning include telecom managers’ fear of having the CEO dial in to the business and get a busy signal, and carriers suggesting that buying more trunks is always a safe bet, says Walter Karopczyc, owner of Bristol Capital, which provides the InfoPlus Traffic Study.
Declining circuit prices have contributed to the trend, too, notes Art Yonemoto, president of auditing firm Yonemoto & Associates, in Dublin, Calif.
A traffic study is the key to determining whether you can save money by reducing the number of trunks you have in service, experts agree. It also can help uncover billing errors that can yield additional savings, as Laino found at the law firm.
Further, you might have cause to run a traffic study if you’ve gotten complaints from your enterprise’s customers because they get busy signals when trying to reach your business, Yonemoto notes. Or perhaps your end users can’t get a free line to dial out. [See related article for eight problems a traffic study can solve]
Step 1: Isolate (or Fake) Peak Traffic Time
There are a variety of options when you choose to hire an outside vendor to perform a traffic study [see related article], but automated programs use the same statistical probability math you would use to do your own traffic study, Yonemoto says.
No matter who does your traffic study, plan to conduct it during your enterprise’s busiest time of the year, Laino recommends. Your polling should capture call data during business hours for at least a week.
The old Bell System methodology is to use the busiest day of the busiest season to engineer trunks, so you’ll hardly ever get a busy signal (except maybe on holidays like Mother’s Day), says Christopher Lee, managing director of telecom consulting firm Source Loop, in Reston, Va.
Can’t wait until your busy season to complete a traffic study? Add an additional percentage to your current call data to approximate the busy season, Karopczyc advises.
Step 2: Capture Call Data
Don’t use your bills to collect call data for your traffic study. They’ll only show the calls you paid for, not the calls that got a busy signal, says Lee, a former MCI traffic engineer.
You might be able to leverage your on-premise or hosted call accounting system to collect the raw data to run a traffic study, Laino says. Contact your call accounting vendor to see whether the system can be set up to report on the usage of outside lines.
You also can request reports from your carrier’s switch, Yonemoto recommends. But note that carriers might refer to an “incoming” call as one placed from your enterprise, though your inclination might be to think of an incoming call as one your enterprise receives.
If your business is really sensitive to blocked calls, make your calculation more conservative by adding an additional percent – say 10% or 15% – to your actual traffic to account for spikes, Lee says. But trunking for conservative call blockage won’t translate into conservative spending, he adds.
Step 3: Translate Call Data into Erlangs or CCS
Now that you have your call data, calculate how many calls were made in an average hour and how long each call lasts on average.
Here’s an example: Users make 250 calls in an hour with an average duration of five minutes. That equates to 1,250 minutes of traffic in an hour.
From here, you can express the amount of traffic in terms of “Erlangs.” An Erlang is equal to 3,600 seconds of calls (3,600 seconds in an hour). Your Erlang figure is determined by dividing your minutes of traffic in an hour by 60, the number of minutes in an hour.
Using our example: You have 1,250 minutes of traffic in an hour, and 1,250 divided by 60 minutes in an hour equals 20.83 Erlangs.
Jot down your Erlang figure – you’ll need it in a few more steps.
The duration of the calls also can be expressed as “centi-call seconds,” or seconds of calls in 100-second increments. This is often denoted in a unit called the CCS; the first C stands for the Roman numeral for 100, and the CS stands for “call seconds,” Laino says.
Step 4: Determine Grade of Service
Now consider this: Is it OK that five out of every 100 callers to your enterprise get a busy signal?
While a department of motor vehicles might opt for cheaper telecom bills over low busy-signal rates, enterprises like stock brokerages and police departments need virtually every call to get through, Karopczyc notes.
Enterprises that need such an excellent “grade of service” scale their facilities to deliver a busy signal to only one out of every 100 callers, he says. This grade of service is written as P.01, which expresses the probability of getting a busy signal as one out of 100 calls.
Determine your business’ tolerance for delivering busy signals to callers, and write it as a probability (P.__). This probability figure is the second component you’ll need to determine how many trunks you really need.
Step 5: Consult Erlang Chart & Size Trunks Appropriately
Now look up an Erlang B table (www.umtsworld.com/resources/erlang.htm, for example).
Select your grade of service (P value) across the top of the chart. Then look up the value in that column that comes closest to your Erlang figure. The column on the far left gives you the number of trunks you’ll need to carry this traffic at your desired blockage rate.
Using our example, with 20.83 Erlangs at P.01, we’d need 31 trunks.
Warning: Check your contract before disconnecting a circuit, Laino advises. Otherwise you could run into shortfall penalties for failure to meet your minimum annual revenue commitment.
Step 6: Repeat Traffic Study Annually
How frequently you perform a traffic study depends on how stable your traffic patterns are. At most, you could do a study once every six months, though many will find an annual schedule to be a better fit, Laino says. (