161 11 AIRTIME PRICE PROJECTIONS 11.1 GREAT EXPECTATIONS When Bell Atlantic Mobile Systems (BAMS) launches its first . . . CDPD . . . system in Washington . . . it will initiate a new business that will impact the . . . bottom line by 1995. . . . The average monthly bill for cellular is dropping, and . . . new services such as data will boost those bills. 1 This BAMS management expectation was common to all nascent data carriers of the period. But were their expectations realistic? With multiple carriers bringing increasingly untapped capacity to market, is it reasonable to believe that data prices will permit high-revenue goals? 11.2 SUBSCRIBER CAPACITY POTENTIAL ARDIS reported 2 a year-end 1991 capacity for 800,000 field service class subscribers on a system rarely having more than two channels per city and using the inefficient MDC4800 protocol. By April 1993 ARDIS had secured 70 channel pairs in the top 20 cities 3 (6 each in New York and Los Angeles) and begun conversion to the faster, more efficient RD-LAP protocol in order to build capacity for the expected Mobile Office/PDA boom. That application did not happen. Many base stations were mothballed until the 1997 win of the UPS contract when buildout resumed. Application profiles also changed. Often, traffic intensified. Some users who began at two messages per hour moved to one every 10 minutes. The early morning/end-of-day message spikes associated with field service were smoothed by The Wireless Data Handbook, Fourth Edition. James F. DeRose Copyright © 1999 John Wiley & Sons, Inc. ISBNs: 0-471-31651-2 (Hardback); 0-471-22458-8 (Electronic) the addition of applications with different message profiles. While there are always individual hot spots, ARDIS subscriber capacity with the new, more message intensive profiles, now exceeds 1 million users, less than 10% of current network utilization. BSWD has up to 30 channel pairs per major metropolitan area. These channels are only half the width of ARDIS channels. and the maximum bit rate in a channel is only 8 kbps. BSWDs cellularlike reuse plan reduces the number of channels at a single, complex base station location to about four. However, BSWD is installing an extremely large number of single-channel base stations. Nevertheless, it certainly has an untapped capacity potential as large as ARDIS. Ricochet, with high bit rates and unusual load dispersion, has very good capacity potential for static users. Both GSM and MIRS technologies give PCS and SMRS users latent data capacity of unusual magnitude with dynamic capacity allocation between voice and data. CDPD, if rolled out to all cellular base stations on a dedicated-channel, sectored basis, has enormous potential. Much of this will be consumed by TCP/IP overhead; much more will be devoted to new application classes featuring image transmission. It remains to be seen whether the necessary rollout will actually occur but, if so, the subscriber capacity potential can be high. What does all this unused capacity coming to market mean for airtime prices in general and, as an interesting surrogate, CDPD in particular? 11.3 LIST PRICE HISTORY: CDPD BAM was first with a priced CDPD offering, structured to land squarely atop BSWDs (then RAM Mobile) packet offering. When CDPD acceptance lagged, each successive carrier price plan flattened the curve in an attempt to make data more attractive. Figure 11-1 illustrates this phenomenon without dwelling on the sometimes important Figure 11-1 CDPD price change history. 162 AIRTIME PRICE PROJECTIONS distinctions such as perfect packet size requirements, national versus local rates, and prime time versus time-insensitive plans. The May 1998 culmination was the AT&T Wireless Unlimited Service plan, which puts CDPD in internal competition for voice channel resources. 11.4 VOICE/DATA CHANNEL RESOURCE COMPETITION: CDPD 11.4.1 Choosing a Representative Carrier Channel availability is not a constraint to carriers providing voice service in relatively low density metropolitan areas such as Racine, Raleigh, Richmond, and Roanoke. But it is a major problem in thickly populated urban locales such as New York City. The two cellular carriers in Greater New York City have long begun their migration to digital technologies to increase voice subscriber capacity. AT&T Wireless has chosen TDMA; BAM has opted for CDMA. Both carriers are also encountering significant New York area competition from the GSM PCS carrier Omnipoint as well as the ESMR carrier Nextel. This competition has led to the first true voice cellular airtime price reductions. AT&T Wireless in Greater New York City was chosen as the surrogate for the value of a voice channel. Its choice of TDMA permits a granular substitute of digital for analog channels. With seven pricing plans, its user charges are very competitive to both Nextel and Omnipoint when minutes per month approach 100. Beyond that usage pointwhere business users are likely to beAT&T Wireless trounces its competitors. Its 1998 Digital One Rate plan is untouched by the cellular alternatives. Figure 11-2 is an example of the competitive New York City offerings. The first AT&T plan shown is the $49.99, $.50/minute beyond 180 level; the Digital One Rate plan is also plotted. Figure 11-2 Representation voice price plans: New York City. 11.4 VOICE/DATA CHANNEL RESOURCE COMPETITION: CDPD 163 11.4.2 Establishing a Representative Voice Profile The granularity of the AT&T Wireless plans forces a strawman choice. For this analysis the business user is assumed to consume 290 ± 110 voice minutes per month. This is a range of 920 minutes per working day. Within this wide range the monthly service charge is fixed at $69.99. This selection is shown in Figure 11-3. For capacity purposes, call duration is assumed to be the national cellular average, 2.32 minutes, which is likely a bit light for business users. However, AT&T rounds up to the next full minute in its pricing; that fact tends to compensate for some of this lightness. Thus, the business user is assumed to average ~4.8 calls per working day, each averaging less than 3 minutes in duration. The wide variations inherent in the $69.99 plan apply. 11.4.3 Estimating Voice Capacity per Cell Without dwelling on the technical intricacies of Erlang tables, the number of analog AMPS calls was estimated with the following assumptions: 1. Erlang B blocked-calls-cleared tables. 2. Grade of service is 2% blocking probability for busy-hour call initiation. 3. Number of user analog channels per sector: 18 (54 per omni site). 4. Average call duration: 2.32 minutes. From the Erlang tables the offered load present at the intersection of 18 (channels) and 0.02 (blocking probability) is 11.5. Since we have assumed an average call duration, the maximum number of AMPS calls that can occur during the peak hour can be calculated by Figure 11-3 AT&T wireless voice price plan comparisons: New York City. 164 AIRTIME PRICE PROJECTIONS Calls during peak hour = o f f ered load × min u tes / hour a v erage call duration = 11.5 × 60 2.32 = 297.41 To simplify matters, assume that we are dealing only with business users. Assume further that one of the average 4.8 business calls per day is generated during the peak hour. Thus, each sector can support ~297 active subscribers with our business profile; each trisectored cell can support ~892 such users, as shown in Figure 11-4. Because cellular subscriber growth has been so relentless, AT&T began the conversion of existing AMPS channels to TDMA some years ago. This conversion had to be started before the number of users saturated the analog channel capacity since, using our call profile, the conversion of the first three channels to TDMA actually reduces maximum call capacity slightly. With four channels converted, the break-even point on call capacity is reached. As more and more channels are converted, total call capacity rises steadily, as shown in Figure 11-5. 11.4.4 Estimating Voice Revenue per Sector Assume that six channels on each sector have already been converted to TDMA. This is a convenient number that marks the point where overall call capacity has risen ~17% (to 348 per busy hour), and both the analog and TDMA channels are each carrying about half the voice traffic. Assume further that every user fits our business profile and makes one call during the busy hour. Thus, an average of 348 subscribers (1044 per Figure 11-4 Trisectored cell capacity representation. 11.4 VOICE/DATA CHANNEL RESOURCE COMPETITION: CDPD 165 cell) use that sector. This conversion point yields an average per sector revenue of $69.99 × 348 = $24,356 per month. As voice traffic grows, and TDMA conversions continue, sector revenue would theoretically head toward the $60,000 per month mark. 11.4.5 Calculating the Required Data Revenue AT&T Wireless has made the business decision to dedicate one channel per sector to CDPD data. Occurring at the same time as TDMA conversion, this must have been a difficult choice. With one channel unavailable to the voice pool, the number of peak- hour callsand thus subscribers per sectorare down to ~328 at our six-channel conversion point. The monthly revenue contribution required from data to reach the break-even point can now be estimated (see Table 11-1). That is, for data to be an interesting business contributor, this sector must have enough subscribers to yield $1394 per month. If data can not deliverand assuming the demand for voice is still growingAT&T Wireless would be better off sticking with voice. Figure 11-5 AMPS/TDMA sector trade-off: average cell duration, 2.32 minutes; 2% blocking. Table 11-1 Required data revenue per sector Voice Only Voice and Data Source Subscribers Rate Revenue Subscribers Rate Revenue Analog 170.9 69.99 11,961 151 69.99 10,568 TDMA 176.9 69.99 12,381 176.9 69.99 12,381 Data 0 1,394 Total 347.8 24,343 327.9 24,343 166 AIRTIME PRICE PROJECTIONS 11.5 ESTIMATING FUTURE DATA PRICE LEVELS Assume that all CDPD data activity is Internet oriented. Yes, there will be the Otis Elevator telemetry applications, but per-unit revenue is small. It is not particularly valuable to divine the future through narrow differentiation between $5.00 and $5.50 offerings. Let us concentrate on the big money instead, and guess at how low monthly prices can go. If all the Internet business users are local unlimited, they pay $54.99 per month for Wireless IP (CDPD) service. The break-even point is thus $1394 ÷ $54.99 = 25.4 average data users per sector. Said another way, ~8% of the business voice users must also subscribe to data. While worst case scenarios can always be constructed, the required user count should not present a technical load problem even if all 25 data users are engaged in realistic Internet access. One valuable characteristic of the Internet profile is that it presents little load to the inbound side, which is the first choke point on CDPD. With a dedicated channel there is no hopping overhead. Thus, the principal constraint on the number of users is the capacity of the outbound channel to deliver information to, say, the mobile professionals laptop. Internet access will create data usage far beyond AT&Ts estimates for the mobile professional. 4 Assume that once during the busy hour each of our 25 users makes a single inquiry to the Web for something like price information, as described in Chapter 8. The outbound channel must deliver ~100 kbytes to give the user what is desired. Naturally this much information must be parceled out by the gateway so that a single user does not monopolize the channel for inordinate periods of time. AT&T studies indicate that achievable CDPD user bit rates range from 4400 to 6400 bps during long file transfers 5 ; I have personally tuned file transfers over CDPD to achieve ~7150 bps when the device is not in motion. Assume that the compression techniques like those employed in Nettechs Smart IP middleware 6 can squeeze 35% of the bits out of my efforts and achieve an effective rate of 9600 bps. This beautifully tuned CDPD channel, with address compression in place, no retries, no hopping, . . . is now capable of delivering ~1200 user bytes a second. Thus, each user will require 84 seconds of pure transmit time, possibly broken up after, say, 2-second intervals, to give someone else a shot. During the busy-hour all 25 users combined will require 2100 seconds. There will be conflicts and delays, but a combined channel utilization of less than 60% (2100/3600) is a reasonable place to be at this juncture. In time, and with historical measurements to aid the calculations, AT&T may be able to load the channel toward the 90% point, ~38 users per channel (ARDIS currently has ~50 per channel per base station, with a much less intense message profile than Internet access). After this point a new channel is required, and a dedicated/hopping hybrid is certainly feasible. But if voice demand is being comfortably accommodated by the TDMA conversion, another dedicated data channel may be deployed. Conversely, if the data subscribers do not appear, AT&T Wireless can ease its prices down again to attract users. Since the channel may be able to handle ~38 11.5 ESTIMATING FUTURE DATA PRICE LEVELS 167 Internet users per hour, each requiring a 100K message, the local unlimited price could theoretically fall all the way to ~$36 per month. This is close to Metricoms current $30 Ricochet price, while offering vastly broader coverage and flexibility. In fact, a precipitous plunge to ~$36 per month is unlikely. Rather, incremental steps to $49, then $44, then $39as the business circumstances dictatedwould be a more reasonable expectation. But if PCS and ESMR impact forces lower voice prices, the value of the voice channel declines and so will data prices. Will these price levels boost those monthly bills as the carrier executives originally hoped? Not likely. REFERENCES 1. B. L. Scott, EVP and CEO: BAMS, Telephone Week , 6-21-93. 2. J. Blumenstein, (then) ARDIS President, Mobile Data Report , 1-27-92. 3. T. Berger, (then) ARDIS Vice President of Radio Network and Product Technology, Lexington Conference, 7-27-93. 4. www.attws.com/nohost/data/order/pricing.html. 5. JFD Associates (Quantum Publishing), ARDIS/RAM Coverage Analysis, CDPD Preview, p. 6-6. 6. Nettech press release, 6-22-98. 168 AIRTIME PRICE PROJECTIONS . Nevertheless, it certainly has an untapped capacity potential as large as ARDIS. Ricochet, with high bit rates and unusual load dispersion, has very good capacity potential for static users. Both GSM and. addition of applications with different message profiles. While there are always individual hot spots, ARDIS subscriber capacity with the new, more message intensive profiles, now exceeds 1 million. reasonable to believe that data prices will permit high-revenue goals? 11.2 SUBSCRIBER CAPACITY POTENTIAL ARDIS reported 2 a year-end 1991 capacity for 800,000 field service class subscribers