Making Complexity Simple: How it works

This is the second of four articles that will describe why taming complexity offers the biggest opportunity of our generation. This article explains how complexity works and why its impact is so powerful and widespread.

A recap on ‘What Complexity Is’

My first article described what complexity is and isn’t, and what it will and won’t do. It defined complexity as “the propensity for emergent phemonema to arise due to the interconnected nature of a system.” It explained that complexity is created when systems become interconnected in a non-trivial way, and it described how complexity acts as a multiplier that turns common occurrences into global events. It went on to say that complexity is a real and measurable entity that can be understood and tamed, and it described how complexity is different from ambiguity, difficulty and uncertainty. It concluded by briefly stating how taming complexity will impact humanity.

How Complexity works…

Complexity works by allowing the probabilistic network of a system, created by local rules, described by a multitude of variables, to create emergent properties that influence performance.

To understand what this means we need to explore five core principles:

1. Networks describe systems;
2. Structure drives function;
3. The more complicated the structure the greater the effect of complexity;
4. Local actions drive global behaviour;
5. Complete control is impossible.

Principle 1: Networks describe systems

Networks are graphical representations of systems that capture components and interactions as ‘nodes and links’.

The first network, constructed by the mathematician Leonhard Euler in 1736, was small, simple and static. Many modern, real-world networks are large, interconnected and highly dynamic. Networks enable us to visualise and interrogate the workings of complex systems and are the starting point for understanding how complexity works. All systems that contain connections can be represented as a networks.

Principle 2: Structure drives function

Networks have ‘emergent properties’ that dictate the performance of the system they represent and limit our ability to understand and control them. By taming complexity, we can control these emergent properties and increase the performance of our networks and systems.

Emergent properties are created from non-trivial interactions within the network and produce global behaviours that are not necessarily observed at a local level. In small, simple and static networks these properties are minor and trivial, but in large, interconnected and dynamic networks they are significant and rarely obvious. The more significant and obscure the emergent property is the more difficult it is to understand and control the performance of the network. Some examples of emergent effects are as follows:

• Human ‘consciousness’ created from the interactions of neurons within the human brain; and,
• Systemic risk in banking and project networks arising from the interactions of the underlying financial institutions and tasks/resources/suppliers.

Principle 3: The more complicated the structure the greater the effect of complexity

The more interconnected, varied and dynamic a network is, the greater the effect of complexity and the greater the opportunity for improvement.

In complex networks the levels of interconnectedness, variety and dynamism are controlled by variables originating from either the individual nodes and links or the properties of the network. Understanding these is the key to understanding and controlling complexity. A non-exhaustive list of individual node and link variables are as follows:

• Node type (e.g. people, tasks, locations, risks, etc.)
• Link type (e.g. strong/weak, periodic/constant, directional/two-way, etc.)
• Variation of the above over time

A non-exhaustive list of network property variables are as follows:

• Node volume (e.g. 100 people or 1,000 people, etc.)
• Link volume (e.g. 100 links or 1,000 links, etc.)
• Node distribution (e.g. clustered or random, grouped or scattered, etc.)
• Link distribution (e.g. even or heavy-tailed, grouped or scattered, etc.)
• Variation of all the above over time

Some examples of major, modern-day, complex networks that contain both sets of variables are:

• Complex Projects: A ￡1Bn complex project will typically have over 1,000,000 unique assets, 100,000 unique tasks, 10,000 unique people and 1,000 unique suppliers and operate over multiple years, with design change, rework, fluctuations and personnel turnover happening on a continuous basis. Average baseline performance: only 7.8% of projects are delivered on budget and on time.
• UK Rail Network: This delivers 4.5 million passengers per day over 900,000 miles of track to 220,000 station stops using 24,000 trains, at close to 100% capacity, with maintenance and upgrades required continuously. Baseline performance: 63.8% of trains arrived on time at 76,247,953 stops between April 2018 to April 2019.
• Global Social Network: The estimated population of the planet in 2019 is 7.7 billion. Facebook has unique profiles for 31% of this (i.e. 2.4Bn profiles as of 2019). The average degree of separation between individuals on Facebook is 4.6, and the ‘effective separation’ (i.e. the average number of steps needed for the average person to navigate the network) is between 5 and 7.
• The World Wide Web: This consists of four ‘website continents’ and is estimated to consist of 134 billion web pages in 2019. The amount of the web that Google has indexed = 4%; rate of web page growth per year = 8%.

Principle 4: Local actions drive global behaviour

To control the evolution and/or behaviour of a network, actions and policies must be directed at individual nodes and links.

This is because ‘networks’ don’t actually behave; it is their underlying components that do. For example, it is not the ‘economy’ that grows, but the underlying revenue of the businesses that sit within the networked economy that grow.

This subtle concept has significant ramifications: to gain maximum control over a complex system policy makers and management must set simple, local rules and allow the system to evolve naturally over time. This is how the tech giants of today effectively design virality into their business models and how complex projects can be protected against domino-like failures that lead to large time and cost overruns.

Examples of these types of local rules include:

1. The UK Government’s Apprenticeship Policy introduced in 2017. In 2017 the UK Government introduced a policy aimed at increasing the number of apprenticeships in the UK. This policy targeted the individual ‘nodes’ of businesses by taxing every UK employer with pay bills in excess of ￡3M per year. It then allowed them to claim this tax back through the provision of apprenticeship training. By introducing targeted taxes and incentives for individual businesses (i.e. node type 1) it was able to increase the number of employees (i.e. node type 2) entering apprenticeships training (i.e. the links between 1 and 2) across the whole UK.
2. Google’s first search algorithm in 1998. In 1998 Google released its first search algorithm designed to map the World Wide Web. This algorithm was special in that it ranked websites based on their position in the web’s network rather than ranking them based only on their nodal characteristics (e.g. total website traffic), as competitors did. By more effectively leveraging the power of the web’s network, Google was able to return more appropriate search results than other engines and rapidly grow its market share.
3. PayPal’s Referral Programme in the early 2000s. In the early 2000s PayPal accelerated its growth by offering $20 to people opening an account (i.e. a new node) and $20 if the referred anyone (i.e. a new link). By incentivising the creation of nodes and links on their platform PayPal was able to grow their user base by hundreds of thousands and in an exponential way. Off the back of this the Wall Street Journal ran a story ‘lauding’ their viral growth and valued the company at $500M. This underpinned their next funding round.

Principle 5: Complete control cannot be achieved

Complete control of a complex network can never be achieved because its behaviour cannot be absolutely predicted (i.e. it is stochastic).

This is because the individual behaviour of the nodes and links and the emergent properties of the network cannot themselves be absolutely predicted (i.e. they are stochastic). This means that:

1. Single predictions of system-wide performance, such as the GDP of the economy, or fixed costs or end dates for Complex Projects, are misrepresentations of reality.
2. Models that assume linear effects, such as the Critical Path Method in Project Management (and derivative methods like PERT), when applied to complex networks, provide incomplete representations of reality.

To clarify (1): Single predictions of system-wide performance can be taken when required by social and political factors, such as the setting of budgets and the communication of risk. Complexity science simply states that it is physically impossible to assure this single prediction and that any belief to the contrary is a misunderstanding of reality.

To clarify (2): In small, simple and static networks where emergent properties do not significantly influence performance, models that assume linear effects can produce accurate results (i.e. where predictions closely equal reality). In large, interconnected and highly dynamic networks where emergent properties significantly influence performance, models that assume linear effects do not produce accurate results. This fact has significant implications for fields such as estimating and risk management and directs our efforts regarding which networks offer the greatest opportunity for improvement (re. Principle 3).

Pulling it all together

Complexity works by allowing the probabilistic network (Principle 5) of a system (Principle 1), created by local rules (Principle 4), described by a multitude of variables (Principle 3), to create emergent properties (Principle 2) that influence performance. The larger, more interconnected and more dynamic a network is the more complexity acts and the greater the opportunity for improvement and control.

The implications of ‘How Complexity Works’

The modern world contains many complex systems that can be depicted as networks, from the global economy to power grids, and from ecosystems to complex projects. Given the size of the opportunity for improvement Venture Capital investors and Strategists are targeting these for economic gain using two main approaches:

1. Constructing platforms that map the network of the system (Principle 1); and,
2. Deploying complexity science to control the behaviour and evolution of such systems (Principles 2–5).

Examples of investments made following this thesis are:

• Illumio, who are mapping networks in the IT sector to protect companies against Cyber-attack.
• Crowdvision, who map the movement of people to increase the efficiency and profitability of organisations that operate large public places.
• Nodes & Links, who are mapping networks in the Projects sector to increase delivery performance and protect businesses against failure and claims (Disclosure: I am the CEO and co-founder of Nodes & Links).

Some less well known examples include:

• Facebook, who have mapped the social network of 30% of the human population and now use this to deliver targeted marketing for others.
• LinkedIn, who have mapped the professional network of 7.4% of the human population and now use this to deliver targeted search capabilities for companies.
• Google, who have mapped close to 100% of the index-able web, and 4% of the total web, and now use this to deliver targeted marketing for others.
• Twitter, who has mapped parts of the daily conversation of 1.6% of the world’s population and now use this to deliver targeted marketing for others.
• Uber, who have mapped the demand and supply networks between taxis (node type 1) and customers (node type 2) in 600 cities to create the links (i.e. rides) and take a share of the revenue.
• AirBnb, who have mapped the demand and supply networks between home owners (node type 1) and guests (node type 2) across 191 countries to create the links (i.e. stays) and take a share of the revenue.

In the introduction I stated that ‘complexity offers the biggest opportunity of our generation’. This is the economic opportunity. Just think, how much of your business/project/or industry can be modelled as a network of interactions and how crucial are these interactions to the performance of the whole?

In Summary

In this article I have explained the five core principles that underpin how complexity works and explained why its impact is so powerful and widespread. I have explained that:

• Networks are graphical representations of systems that capture components and interactions as ‘nodes and links’.
• Networks have ‘emergent properties’ that dictate the performance of the system they represent and limit our ability to understand and control them.
• The more interconnected, varied and dynamic a network is, the greater the effect of complexity and the greater the opportunity for improvement.
• To control the evolution and/or behaviour of a network actions and policies must be directed at individual nodes and links.
• Complete control of a complex network can never be achieved because its behaviour cannot be absolutely predicted (i.e. it is stochastic).

Next time…

In my next article I will give evidence of the power of complexity by revealing the game-changing results in performance and control that we have achieved in the world of Complex Projects.

Contact

If you have an interest in Complexity or Projects and would like to get in touch feel free to contact me at [email protected].

Greg is the CEO of Nodes & Links, a venture-backed technology company that is world-leading in the application of complexity science to business and project management. Based in London, the company provides cutting-edge research to the world’s leading journals and builds technology that controls complexity to make the world a better place. Aegis — the company’s primary hybrid-collective intelligence platform — will be released for general use in early 2020 and will assure the world’s most complex projects of their performance.