You all know Monsieurs (MM.) Moore et Metcalfe. But do you know Monsieur (M.) Zipf? I made his acquaintance whilst researching infinite long tails. Why does he matter, you inquire? Because M. Zipf brings some respectability to Moore et Metcalfe, who can get a little out of control from time to time.
Monsieur Moore is an aggressive chap who doubles his strength every 18 months or so and isn’t shy about it. Monsieur Metcalfe has an insatiable appetite, and every bit he consumes increases his girth substantially. Many people have made lots of money from MM Moore’s et Metcalfe’s antics over the past 30 years. The first we refer to generally as the silicon or processing effect, the latter as the network effect. Putting the two together should lead to declines of 50-60% in cost for like performance or throughput. Heady, rather piggish stuff!
Monsieur Zipf, on the other hand, isn’t one for excess. He follows a rather strict regimen; one that applies universally to almost everything around us; be it man-made or natural. M. Zipf isn’t popular because he is rather unsocial. He ensures that what one person has, the next chap only can have half as much, and the next chap half that, and so on. It’s a decreasing, undemocratic principle. Or is it?
Despite his unpopularity, lack of obvious charm and people’s general ignorance of him, M. Zipf’s stature is about to rise. Why? Because of the smartphone and everyone’s desire to always be on and connected; things MM Moore and Metcalfe wholeheartedly support.
M. Zipf is related to the family of power law distributions. Over the past 20 years, technologists have applied his law to understanding network traffic. In a time of plenty, like the past 20 years, M. Zipf’s not been that important. But after 15 years of consolidation and relative underinvestment we are seeing demand outstrip supply and scarcity is looming. M. Zipf can help deal with that scarcity.
Capacity will only get worse as LTE (4G) devices explode on the scene in 2012, not only because of improved coverage, better handsets, improved Android (ICS), but mostly because of the iconic iPhone 5 coming this summer! Here’s the thing with 4G phones, they have bigger screens and they load stuff 5-10x faster. So what took 30 seconds now takes 3-10 seconds to load and stuff will look 2-3x better! People will get more and want more; much to MM Moore’s et Metcalfe’s great pleasure.
“Un moment!” cries M. Zipf. “My users already skew access quite a bit and this will just make matters worse! Today, 50% of capacity is used by 1% of my users. The next 9% use 40% and the remaining 90% of users use just 10% of capacity. With 4G the inequality can only get worse. Indignez-vous! Which is the latest French outcry for equality.” It turns out Zipf's law is actually democratic in that each person consumes at their marginal not average rate. The latter is socialism.
Few of us will see this distribution of usage as a problem short-term, except when we’re on overloaded cellsites and out of reach of a friendly WiFi hotspot. The carriers will throw more capex at the problem and continue to price inefficiently and ineffectively. The larger problem will become apparent within 2 years when the 90% become the 10% and the carriers tell Wall Street they need to invest another $50B after 2015 just after spending $53B between 2010-2014.
Most people aware of this problem say there is a solution. More Spectrum = more Bandwidth to satisfy MM Moore et Metcalfe. But they’ve never heard of M. Zipf nor understood fully how networks are used. Our solution, extended as a courtesy by M. Zipf, is to “understand the customer” and work on “traffic offloading” at the margin. Pricing strategies, some clever code, and marketing are the tools to implement a strategy that can minimize the capital outlays, and rapidly amortize investment and generate positive ROI.
We’ve been thinking about this since 1996 when we first introduced our 4Cs of Wireless (cost, coverage, capacity and clarity) analyzing, understanding and embracing 10 cent wireless pricing (introduced by French Canada's revolutionary MicroCell). As a result we were 2-3 years ahead of everybody with respect to penetration, consumption and wireline substitution thinking and forecasts. Back in 1995 the best wireless prices were 50 cents per minute and just for buying a lot of local access. Long-distance and roaming charges applied. So a corporate executive who travelled a lot would regularly rack up $2000-3000 monthly phone bills. The result was less than 10% penetration, 80 minutes of use per month, and ARPUs declining from $45 to $40 to $35 in analysts' models because the marginal customers being added to the network were using the devices infrequently and more often than not putting them into the glove compartment in case of emergencies. Fewer than 3% of the population actually used the devices more than 1x day.
We used to poll taxi drivers continuously about wireless and found that their average perceived price of $0.75 per minute was simply too high to justify not having to pull over and use a payphone for $0.25. So that was the magical inflection point in the elasticity curves. When MicroCell introduced $0.10 late in the Spring of 1996 and we polled the same set of users, invariably we were just able to avoid an accident they got so excited. So we reasoned and modeled that more than just taxi drivers would use wireless as a primary access device. And use it a lot. This wireless/wireline substitution would result in consumption of 700-800 minutes of use per month, penetration hitting 100% quickly and ARPUs, rather than declining, actually increasing to $70. The forecast was unbelievably bullish. And of course no one believed it in 1996, even though all those numbers were mostly reached within 5 years.
But we also recognized that wireless was a two-edge sword with respect to localized capacity and throughput; taking into account the above 3 laws. So we also created an optimal zone, or location-based, pricing and selling plan that increased ARPUs and effective yield and were vastly superior to all you can eat (AYCE) and eat what you want (EWYW) plans. Unfortunately, carriers didn't understand or appreciate M. Zipf and within 2 years they were giving away night and weekend minutes for free, where they could have monetized them for 3-6 cents each. Then some carriers responded by giving away long-distance (whose marginal cost was less than a minutes without access; but still could cost 2-3 cents). Then AT&T responded with the One-rate plan, which destroyed roaming surcharges and led to one-rate everywhere; even if demand was different everywhere.
Here’s a snapshot of that analysis that is quite simple and consistent with Zipf's law and highly applicable today. Unfortunately, where my approach would have kept effective yield at 8 cents or higher, the competitive carriers responded by going to all you can eat (AYCE) plans and the effective yield dropped to 4 cents by 2004. Had intercarrier SMS not occurred in the 2003-4 timeframe, they would have all been sunk with those pricing models, as they were in the middle of massive 2G investment programs for the coming "wireless data explosion", which actually didn't happen until 3G and smartphones in the 2008-2009 timeframes. It was still a voice and blackberry (texting and email) world in 2007 when the iPhone hit. With ubiquitous SMS and people's preference to text instead of leaving vmail, minutes dropped from 700 to 500, lowering carrier's costs, and they were able to generate incremental revenues on SMS pricing plans (called data) in the 2004-2007 timeframe.
All that said, the analysis and approach is even more useful today since extreme consumption of data will tend to occur disproportionately in the fixed mode (what many refer to as offload). I let you come up with your own solutions. À bientôt! Oh, and look up Free in France to get an idea of where things are headed. What is it about these French? Must be something about Liberté, égalité, fraternité.
It's All in Your Head by Robert Metcalfe
JD Power Surveys on Wireless Network Performance
The Law of the Few (the 20/80 rule) from the Tipping Point
Briscoe, Odlyzko get it wrong because of one dimensional thinking and mixing apples and oranges
See Larry Downes' Law of Disruption
Law of Bandwidth (but only in a period of monopoly)
See our own home cooked Law of Wireless Gravity
Keck's law of fiber capacity (may be coming to an end?)
Kurzweil's 20 Laws of Telecosm (how many right and wrong also depends on timing)