Using AI to Reduce Small Cell Cost as Much as 4X

Mobile service providers are finding lots of ways to keep 5G infrastructure costs down, including using artificial intelligence to improve siting of small cells, thereby reducing the amount of small cell investment. 

Small cell capex can be as much as four times higher when the small cells are not located within about 20 meters to 40 meters (65 feet to 131 feet) of demand hotspots, according to a new white paper by the Small Cell Forum. So operators are starting to use AI to improve cell siting. 

“Where small cell placement was off by as little as half a cell radius from a given hotspot, the result – according to one operator – was that four times as many small cells were required to carry the same traffic,” the Small Cell Forum says.

Location accuracy is even more challenging when higher frequency bands are used, reducing the practical serving radius of the cells.

source: Small Cell Forum

“A well-placed small cell has its serving area covering locations with high demand,” the paper states. “A poorly located small cell is too far away from the hotspot; its serving area does not cover an area of high demand.” And that efficiency is directly related to the number of small cells that need to be deployed, and therefore capital investment. 

To maximize small cell return on investment, and minimize capex overall, small cell siting should be within 20 meters to 40 meters of the ideal theoretical placement (real estate considerations matter, in that regard).

No Quick Revenue Uplift from 5G?

Many observers are hoping for a relatively-quick uptick in firm productivity and capabilities driven by 5G, edge computing, internet of things and artificial intelligence. Fewer likely believe 5G will positively affect gross revenues and profits from consumer mobility services. Patience is the watchword. 

It often takes much longer to reap technology benefits than observers expect. Researchers call this the productivity paradox. Quite often, big new information technology projects or technologies fail to produce the expected gains, for longish periods of time, such as a decade to three decades. 

In fact, some argue that a productivity boost between 1995 and 2000 was not enabled by information technology. But it also is likely the case that better information technology allows some firms to take market share from other firms, though overall productivity might not shift much, if at all. 

Even though knowledge of electricity was widespread in the late 1800s, electric power technology did not significantly spread until around 1914 to 1917 in the United States. In fact, factories did not fully utilize electricity until the 1920s. 

It took two to three decades before electricity was being used in a productive manner: about 40 years, as it turns out. 

To note just one example, much of the current economic impact of “better computing and communications” is what many would have expected at the turn of the century, before the “dot com” meltdown. Amazon, cloud computing in general, Uber, Airbnb and the shift of internet activity to mobile use cases in general provide examples.

But that lag was more than 15 years in coming. Nor is that unusual. Many would note that similar lags in impact happened with enterprises invested in information technology in the 1980s and 1990s.

Investments do not always immediately translate into effective productivity results. This productivity paradox was apparent for much of the 1980s and 1990s, when one might have struggled to identify clear evidence of productivity gains from a rather massive investment in information technology.

In fact, diffusion of a new technology takes time precisely because people need time to learn how to use the technology, while organizations must retool the ways they work to incorporate the better technologies most productively. 

Computing power in the U.S. economy increased by more than two orders of magnitude between 1970 and 1990, for example, yet productivity, especially in the service sector, stagnated).

And though it seems counter-intuitive, some argue the Internet has not clearly affected economy-wide productivity. But part of the problem is that it is impossible to capture productivity gains using our normal measuring tools when products or services have zero price. And much of the consumer internet is based on zero pricing. 

Other traditional measures of growth also suffer when technology arguably improves efficiency and productivity (more produced by the same--or fewer--humans). Look only at the print content business, where revenues, profits and employment have plummeted in the internet era, even as the volume of content of all sorts has increased exponentially. 

Or consider the impact of software on information technology. The Bureau of Labor Statistics estimates that employment in information technology was lower in 2018 than it was in 1998, despite the obvious increase in software-intensive life and business.

Gartner, for example, recently said that enterprises will have to wait twice as long as they expect to reap incremental revenue from technology investments.

Through 2021, incremental revenue from digital transformation initiatives is largely unlikely, Gartner researchers predict. That will not come as good news for executives hoping for revenue growth from repositioning existing business practices for digital delivery and operation. 

On average, it will “take large traditional enterprises twice as long and cost twice as much as anticipated,” Gartner researchers predict. 

When, and How Will "Digital Transformation" Show Up in Statistics?

Despite a massive wave of capital investment now underway to “digitize” most aspects of business, it is a fair question to ask how long it will take before tangible financial benefits are reaped, beyond a shift of market share from some suppliers to others. That will be quite tangible, and will show up in gross revenue changes. 

There are several problems. First, there is almost always a long lag between major waves of technology investment and tangible changes in productivity. Also, digital transformation can cannibalize a firm’s revenue. 

“In a recent survey we conducted, companies with digital transformations under way said that 17 percent of their market share from core products or services was cannibalized by their own digital products or services,” according to researchers at McKinsey. 

The point is that heavy information technology spending to “digitalize the enterprise” might not show especially tangible benefits in productivity, incremental new revenues and new products for quite some time.

source: McKinsey

What will happen is that firms will slow the rate of market share shift from attackers, while some attackers will gain market share. While that might not alter long-term productivity or growth rates in an economy as a whole or within an industry, it will tangibly affect gross revenue, profit margins and market share within an industry, amongst competitors. 

The telecom analogy is the business impact of switching to fiber-to-the-home or other access network platforms. At least in competitive markets, where telcos were facing competition from cable TV operators in every core service, the decision to invest in FTTH was actually not driven so much by an expectation of overall increased revenue or market share, but fundamentally by the effort to retain overall revenues in the face of share loss.

Basically, and colloquially, the advantage of FTTH was strategic: “you get to keep your business.” The logic was that new video subscription revenues would offset loss voice market share, while FTTH would allow telcos to keep pace with cable TV operators in internet access speeds. 

That might seem like an awful expensive proposition. Investing heavily to “stay where one is” is not a traditionally sound investment principle. But lost market share really does matter as well, and many new “digital enterprise” investments arguably are of that nature: invest to limit attacker market share gains. 

That is not to say there will be no non-quantifiable gains for legacy or established providers. Better customer satisfaction, lower operating costs, better marketing platforms and other effects hard to capture on financial reports are possible. But the impact visible on financial reports might, in the near and medium term (several years to 10 years), mostly only be captured in a negative sense (market share not lost; market share not lost as fast), rather than in a positive sense (market share gained). 

The impact for attackers might arguably be different: market share taken from existing competitors. That noted, eventually, we are going to see value in the traditional metrics of productivity growth. 

Non-traditional measures, though, likely are needed to capture the benefits of innovations and value with a zero price, or reflecting quality improvements impossible to capture with price metrics. 

Traditional metrics do not capture increases in well-being and consumer welfare provided by zero-price quality improvements or zero-price products that often as substitutes for positive price products.

Can Policymakers Create Many More Silicon Valleys?

Many policymakers understandably would prefer if high-wage, high-skill jobs could be created more broadly across a country, instead of clustering in a relative handful of locations. In other words, can workers and jobs be dispersed more broadly, instead of remaining concentrated spatially, in Silicon Valley or Mumbai or Shenzen.

Local policymakers--especially--might prefer that well-paid “cognitive non-routine” workers live and work in their dispersed communities, and not in a relatively few cities. 

Rise of the Rest, the venture capital firm that aims to fund seed stage companies located across the United States, but not in Silicon Valley, New York or Boston, is one example of a bet on overlooked CNR talent.

Indeed, the argument that “good broadband” is the underpinning of economic growth for rural areas, small towns and cities is an unexamined claim very similar in its assumptions. 

But a new study conducted for the Federal Reserve Bank of Richmond suggests the difficulty--or futility-- of such efforts. In fact, the researchers actually recommend that concentration be encouraged. “The concentration of CNR workers in a few ‘cognitive hubs’ should be encouraged, not scorned,” they say. 

Basically, they argue, it is not possible to create big CNR job concentrations in most locations. Instead, 

“Contrary to some previous literature and much of the public discourse, the economics of the problem suggest that, with the appropriate transfers, small industrial cities in the U.S. should attract non-CNR workers and not try to become the next San Jose,” the study lead by Esteban Rossi-Hansberg of Princeton University, Pierre-Daniel Sarte and Felipe Schwartzman of the Federal Reserve Bank of Richmond says. 

Arguments about high-quality broadband take the same general approach to causation: good broadband creates the foundation for economic growth. The general economic problem, though, is similar to the issue of dispersing CNR jobs. There appear to be externalities that mostly prevent big CNR clusters from emerging in most locations. 

Though policymakers are correct to worry about occupational and wage gaps, the Federal Reserve study suggests the near-futility of efforts to create many additional dispersed concentrations of CNR workers and jobs. 

Since the 1980s the U.S. economy has experienced increased skill and occupational polarization based on geography. Large cities increasingly have more highly educated workers in “cognitive non-routine” jobs that also are paid more than workers in non-cognitive roles.

At the same time, many medium and small cities have suffered an exodus of skilled workers and population. Income inequality between occupations arguably has grown. 

This growing gap between top and medium and small-sized cities has motivated policymakers and city governments to advocate policies to attract CNR workers to smaller towns. 

If that sounds like the arguments for investing in municipal broadband, it is precisely so. The argument is that quality broadband underpins economic growth. But there is not much evidence that broadband--in and of itself--actually causes such growth. Some might note the productivity paradox as well, the finding that advanced technology actually does not clearly seem to drive economic results. 

Neither is there clear evidence that broadband improves productivity. You would be hard pressed to find any evidence for the thesis that broadband clearly boosts firm productivity, even if we all seem to believe that is the case. Some studies that find some small benefit cannot separate broadband from the other information technology introduced at the same time. But most of the time, it is hard to identify a clear correlation, much less causality.

As a practical matter, governments and others will continue to argue that broadband service has to be improved, because, you know, productivity will improve and economic growth will be aided. And, as a practical matter, firms will continue to deploy, and customers will buy, better broadband.

Still, it is worth noting that there is scant proof that broadband improves productivity.

“We find that the average effect of UFB (ultra-fast broadband) adoption on employment and... productivity is insignificantly different from zero, even for firms in industries where we might expect the returns to UFB to be relatively high,” say researchers Richard Fabling and Arthur Grimes,

One study found no correlation between broadband and productivity, when looking at digital subscriber line deployments. Another study also found no causal link between broadband use and productivity.

Yet other studies suggest that firm using more information technology, including broadband, do raise productivity, though it is not clear whether it was the broadband or the other innovations that contributed.  

Some studies note that it is difficult to tell which came first: a firm’s ability to wring value out of information technology, or broadband enabling that for a firm.

“One view is that good firms with good managers do most things in a better way, including use new practices at the right time,” note researchers from Stockholm University. “This makes studies of the impact of innovation, new management practices, work organisation and ICT use meaningless, since the good firms are much better in many other ways which are and can not be measured.”

The point is that social engineering, even when motivated by legitimate concerns, is tough to impossible.

%G Priced Above, Even and Below 4G

Chinese mobile operators plan to price 5G internet access plans at rates less than 4G, according to analysts at New Street Research. That means 5G retail tariffs soon will be available at a premium to 4G, the same as 4G, or less than 4G, by different servies providers in different markets.  

source: New Street Research

In the U.S. market, pricing has been set at no premium or at a higher cost than 4G.

Is Coverage or Demand the Big Issue for U.S., Europe Internet Access?

When assessing the degree to which internet access is unavailable to consumers, it always makes a difference whether we are considering services consumed by location, or by the person. Fixed services represent the former, mobile phone services the latter. 

The difference is important. For any service consumed by location (homes or buildings), the number of people that are unserved is inaccurate. It is the number of unconnected locations that matters, since the service is consumed by “locations,” not “persons.”

source: Intelsat

In the United States, where Intelsat estimates that about one percent of the population does not have a single supplier of internet access, the number of “locations” not connected is far lower, perhaps in the area of four tenths of one percent, since the typical number of persons per home is 2.6. 

The far bigger issue is locations that are able to buy service, but choose not to, estimated by Intelsat to represent 26 percent of people, or perhaps 10 percent of U.S. homes. Many assume the issue is supply, but a stronger argument actually is demand. 

Some households simply choose not to buy the product, even when it is available, often preferring to use mobile broadband as a substitute. 

Even in the 4G era, 15 percent to 20 percent of U.S. households have become mobile-only for internet access, while in Canada perhaps 20 percent of households already rely exclusively on mobile networks for internet access. 

In some instances, as with younger, single-person households and lower-income households, reliance on mobile-only internet access is 10 percentage points to as much as 15 percentage points higher than that.

source: Deloitte

Some argue take rates sometimes are low because the service is deemed “too expensive.” That might sometimes be the case, despite the existence of subsidized services for low-income households, especially in remote areas where the only option is satellite access. But a good argument can be made that non-buyers are making rational choices. 

They either do not value using the internet or they use the internet other ways (mobile, primarily). 

Absolute lack of coverage (no terrestrial networks or service providers) is a major issue in parts of Africa and South Asia, however.

Why Nobody Can Predict What Happens in 5G Era

Few industry executives are too certain about what big new 5G innovations will develop, and we should not be at all surprised by that lack of clarity. 

The reason futurists and industry executives have been generally unable to accurately predict the development of new use cases, applications and business models in the era of 3G and 4G, and will similarly find the same task difficult in the 5G era is that each of those network generations might be considered complex systems. 

And one attribute of a complex system is that they are virtually impossible to model reliably, as chaos theory suggests. 

A complex system is composed of many components which may interact with each other, and where small initial changes can have very-large, unexpected outcomes (the butterfly effect). 

Examples of complex systems often are said to include Earth's global climate, organisms, the human brain, power grids, transportation or communication systems, social and economic organizations (cities). 

An often-used example is an ecosystem. And it is reasonable to describe the entire connectivity business (fixed, mobile, satellite, other networks) as now existing as part of a larger internet ecosystem. And if a single complex system is intrinsically difficult to model, an ecosystem of complex systems arguably is impossible to model accurately over time. 

In large part, uncertainty about 5G exists because all complex systems are inherently highly dynamic, where no single actor in the ecosystem can direct progress. Also, the connectivity business as a whole now is part of the internet ecosystem. 

It is easy enough to point out the differences between the connectivity function and the applications and services that take advantage of internet connectivity. But even those important distinctions do not fully capture the differences.

Ever since the global telecom industry decided that internet protocol was the next-generation network, the industry has had to adapt to a business ecosystem where the creation of services and apps can easily be conducted by third parties with no business relationship with any connectivity provider. 

source: Detecon

And that changes everything. Yesterday, telecom was a complex system of its own. Today, the whole connectivity function is part of the broader internet ecosystem. The fundamental change is that connectivity providers no longer control the scale or pace of business development of the whole ecosystem. 

source: Nokia Bell Labs

Where today the assumption is that system conflicts must be resolved and resolved centrally and uniformly, a complex system actually requires decentralized control. 

Where a complex system inherently features conflicting, unknowable, and diverse requirements, today’s assumption is that requirements can be known in advance,  and will change slowly. That leads to the assumption that tradeoff decisions will be stable. 

But complex systems have inherently unknowable behaviors. 

Where today’s notion is that system improvements can be, and are, introduced at discrete intervals, complex systems evolve continuously. Where it once was believed that the telecom industry could make changes where the effects could be predicted sufficiently well, complex systems do not support such certainty. 

Where it was believed that the configuration information for any specific change was accurate and could be tightly controlled, a complex system actually means changes happen in an inconsistent environment. 

All of that explains why nobody really can predict what will develop in the 5G and subsequent eras of mobility.