Path Dependence in Telecoms and Technology

In this guest post John Naylon, AWS Principal SA for Telecoms, talks about path dependence—the idea that solutions are sometimes subject to historical accident—and why the AWS notion of “reinvention” in the cloud is a possible solution. While his examples are drawn from the telecom industry, the point is an important one for all of us.

—Mark

By John Naylon, Principal Solutions Architect for Telecoms at AWS

Path dependence is the name for the phenomenon of a de facto standard enduring, not because it represents a currently optimal choice, but because of historical evolutions toward the current state.

One example of this answers a question that puzzles some travellers to the UK: “Why do so many British houses have a separate hot and cold faucet?” The answer is that internal plumbing became commonplace in British houses well before the advent of central heating and hot water generation. Then, decades later, when hot water became available on tap, the simplest way to add it to existing kitchens and bathrooms was by adding a second set of pipes and an additional faucet. Because this was therefore the norm, typical baths and sinks were manufactured with two sets of holes, and therefore even newly built properties would fit a hot and a cold faucet, so perpetuating the pattern.

I was reminded of this phenomenon when reading a recent McKinsey report about telecom operators. A stark figure in the report shows the industry average return on invested capital (ROIC) dipping ever closer to the weighted average cost of capital (WACC), with a spread in 2019 of less than 1%. What does that have to do with path dependence?

To answer the question, we have to look at the evolution of telecom operators. The oldest are the traditional wireline operators like AT&T or British Telecom, whose origins lie in building copper-based fixed networks to homes and businesses. By its very nature, this was an undertaking which required the companies to build physical infrastructure in many locations. Even mobile operators, relative newcomers to the telecom world and not burdened by the need for a physical bearer to the end user, initially had to build their own cell towers, switching and data centres, and backhaul networks.

Networks subsequently underwent a series of transformations from analogue origins, to engineered digital systems, to COTS digital systems, to virtualised network functions, and most recently to a service-based architecture. These evolutions did not occur particularly rapidly—think of the approximate 10-year lifespans of 2G, 3G, and 4G, for example. But the serviceable life of a building that once housed a telephone exchange might be 40 or more years.

During each of these technological transitions, it is theoretically possible to make financially optimal choices about how to host these technologies. However, there is also the temptation to make the minimum necessary change—or in other words, in the plumbing analogy, to add the second faucet. One cause for this preference is that it is harder to quantify the benefits of making greater, more optimal changes than it is to quantify the immediate cost of change. Another possibility is that a slightly altered status quo appears to be a lower-risk option, because it is well understood.

Returning to telecoms, and mobile networks in particular, the 5G system is a service-based architecture which naturally lends itself to efficient implementation as loosely coupled services, cost-effectively hosted in the public cloud. But, for operators with existing infrastructure (both physical and digital), fully adopting that architecture does not represent the minimum necessary change. Path dependence is at work! This can result in operators not realising the full benefits of successive generations of technology, due to implementations arbitrarily following the precedents required by earlier generations. Software developers may recognise this as being similar to the accumulation of technical debt, and organizational theorists might refer to the process as “imprinting.”

Coming back to the McKinsey report, when we consider that network operators typically operate multiple generations of technology simultaneously—2G, 3G, 4G, and 5G being a common mixture in 2021—we can begin to understand how network investments are becoming ever less financially efficient. Nonspecialists may be surprised to know that 2G-era systems are often a critical dependency of an operator’s mobile network!

In the AWS model of cloud adoption, we refer to the final, ongoing state of cloud adoption as “reinvention.” This captures the idea that organizations in this goal state are continuously adapting to a changing environment using the tools the cloud provides, paying only for the IT resources and services consumed, and experimenting in a low-cost, no commitment way. This approach makes it easy to iteratively optimize architectures, painlessly abandoning previous versions without financial penalty. Reinvention inoculates against the perils of path dependence, firstly because the cost of change is lowered: sunk costs and the attendant fallacy are avoided altogether. Secondly and even more significantly, experimenting, iterating, and abandoning previous generations of architecture become business as usual, with the result that organizational imprinting cannot take hold. This allows the organization to maximize the potential of not just endogenous innovations but also generational technology changes in the wider environment.

New entrant, or “greenfield,” network operators present a very interesting counterexample to the general observation that network architecture is path dependent. Such operators often run just a single generation of technology, and their architectural design inherently proceeds from a blank sheet of paper. Two of the most prominent such operators are Rakuten of Japan and DISH network of the USA. Amazon Web Services and DISH recently announced a strategic collaboration to build a cloud-based, nationwide 5G network.

Over time, it will be extremely interesting to compare not just the technical performance of these new market entrants to the existing networks, but also the financial performance. I expect this comparison will allow the true cost of path dependence to be shown more accurately, and ever more operators will realise the benefits of adopting a fully cloud-native approach to their intrinsic operations. However, even now it is by no means the case that only greenfield operators are cloudifying their networks. For instance, Swisscom, an organization with roots dating back to 1852 and currently Switzerland’s largest telecom company, recently announced that they will use the AWS cloud to power their 5G network.

You can read more about Swisscom’s cloud-first strategy and planned cloud-native 5G core here, and about AWS Telecom in general here. At the recent Mobile World Congress 2021, we were delighted to be joined by executives from both DISH and Swisscom to discuss their network cloudification plans. You can view these presentations and many others by visiting the AWS Virtual Village.

About our Guest

John Naylon is a Principal Solutions Architect at AWS, in the Telecom IBU. Before joining AWS, John founded two disruptive startups in the fixed wireless access (FWA) space. This blog was written using a QWERTY keyboard, iconic emblem of the power of path dependence.