EXECUTIVE SUMMARY | It is extremely difficult to predict a declining product such as Interstate Minutes of Usage (MOU) in the telecommunications industry. This article shows how autoregressive EGRACH-M technique, combined with expert judgment, can significantly improve the forecast of such products. The case study is used to demonstrate it.
When is the last time you saw a movie where someone actually picked up a phone with a cord attached to it and made a call? Maybe a decade ago? Those oldtime phones, the landlines, are still out there but they are disappearing rapidly. From a rural telephone company's perspective, there are more than 1,100 of them serving 35% of the U.S.'s geographic area. The traditional long distance minutes flowing through landlines currently account for about 5% of their revenue stream. The issue facing us at Nation Exchange Carrier Association (NECA) is how to forecast long distance minutes for rural telephone companies.
NECA has developed a good deal of expertise in forecasting those minutes over the last twenty-five years. The older members of our staff remember when long distance minutes were shooting up by doubledigit rates inthe last fifteen years of the 20th century, and now, in the first ten years of the 21st century, saw them declining by double digits. Through every phase of this product life cycle, forecasters at NECA have experimented with just about any forecasting technique, from the most leading edge statistical tools to the simplest trend method. In this paper, we share our experience and offer suggestions on how to forecast a product, especially during its mature and declining stages. We have learned that times series techniques aided by judgment, built up over many years of experience, produce the most accurate forecasts. Here we wish to share those techniques, which cannot be found in a textbook. First we will provide more details of trends in our industry.
HISTORY OF INTERSTATE MINUTES OF USAGE
Interstate minutes of usage (MOU) are the amount of traditional landline long distance traffic measured in minutes that crosses state borders. Figures 1-3 map the up, flat, and declining stages of MOU for NECA pool members from July 1986 to December 2010. The categorization is based on the different trend patterns both in the raw data series and in the annualized monthly growth rates.
During its upswing stage, monthly MOU grew typically on an annualized basis by more than 10%, hitting 18.1% at its highest level, and 6% at its lowest. During the "stationary" stage, annualized monthly growth rates hovered at 0%, deviating slightly in both directions. During the declining stage, annualized monthly growth rates ranged from -9.1% to -17.4% at its lowest level, with the decline steadying in the -11% to -13% range at the end of the series.
WHY MOU'S HISTORICAL PATTERN?
The decline of landline long distance service should not surprise anyone. When you call someone, do you use a landline or wireless? If you say landline, you are probably dating yourself as someone over fifty. Even the question itself is misleading because it suggests voice is the way people communicate these days when in fact, most teenagers just text each other. Suppose you decide a voice call is necessary to someone overseas, would you make a traditional long distance phone call or use Skype? Any twentysomething person today uses the Internet, not the traditional telephone network to make these calls. Figures 4 and 5 and Table 1 show the growth patterns of landline's competing services: Wireless voice, texting, and Voice over Internet Protocol (VoIP). Notice that wireless growth shown in Figure 4 has flattened in recent years, so the continuing decline in landline traffic is now likely caused by other products. In earlier years, 20022007, wireless growth did coincide with a significant drop off in landline business. Texting, in contrast to wireless voice, grew by 31% between 2009 and 2010 (Figure 5), suggesting a shift from voice to texting probably explains the decline in landline long distance service in recent years. …