In common corporate discourse it is taken as a given that redundancy is bad. It’s bad, people think, for a company to have two people perform the same job function or to have two similar products undifferentiated in the market place. Redundancy, so this line of thought goes, is economic waste that must be eliminated in the form of lay offs or budget cuts. This way of thinking reflects an expectation, bolstered by the industrial age, that organizations should behave like machines. Each employee is thought of like an interchangeable part. Yes, you might need some spare parts floating around. But for the most part, when someone leaves, a new worker can be be plugged in with minimal interruption to the system as a whole. When the environment around an organization is static, the machine-like elements of a company pop to the foreground. When interactions with the outside world are predictable, corporations can create formulaic routines for optimizing their internal behavior.
The machine metaphor, however, breaks down when considering digital companies operating in the age of the internet. Unlike the relatively static environments of many industrial-age companies, the environment surrounding many of today’s organizations is fraught with compounding uncertainty:
- The uncertainty inherent in how people interact with software. When you are designing a piece of software, whether it is a consumer web site, app or an enterprise platform, you don’t know in advance which design directions will fail or succeed. Instead, it takes a core group of people close to users to “discover” a working product. You don’t really know what works until you see real people using it.
- The uncertainty inherent in how internal changes propagate through networks. When you do create a piece of software that clicks with its users, business success is dependent on the ability for the user base to scale to a sufficient number of paying customers. Successful growth tactics are difficult to predict in advance. When developing marketing avenues or viral mechanics, extensive tinkering is necessary to discover messages that stick and spread. As explored in Malcolm Gladwell’s The Tipping Point, seemingly trivial changes can be the difference between virality and invisibility.
- The uncertainty inherent in external changes. After discovering a product that resonates and scales, maintaining product-market fit requires constant adaptation. The networked system of the internet is ripe for rapid change. Phenomena can emerge and swiftly capture the attention of a large portion of the network. Competitors can quickly appear, yanking away mind share. A search engine algorithm change can destroy a crucial user acquisition channel.
Organizations aspiring to operate like machines involve centralized management creating fixed recipes to dictate the behavior of front-line employees. Due to the environment of uncertainty inherent in the internet, such an approach is becoming increasingly flawed. A centralized management team, removed from the front lines, does not know in detail what will make a sustainably successful web site, app, or marketing approach. Instead, the creators on the front-lines must be valued, not as interchangeable parts, but as sources of knowledge and ideas.
Consequently, we’re seeing today’s organizations emphasize learning as a critical element of their processes. When looking for a metaphor for how companies should aspire to learn, the richest place to turn is the human brain itself. In his book Images of Organization , Gareth Morgan explores how the brain can used as a model for how organizations learn and self-organize. While Morgan identifies many patterns of brain function that are applicable to organizations, “redundancy,” in particular caught my attention due to the diversity in how it’s manifested in today’s companies. Morgan writes:
Any system with an ability to self-organize must have a degree of “redundancy,” a kind of excess capacity that can create room for innovation and development to occur. Without redundancy, systems are fixed and completely static.
In the human brain we find this redundancy in the vast networks of connectivity through which each neuron, or nerve cell, is connected with thousands upon thousands of others. This enormous capacity generates considerable evolutionary potential. It all vast amounts of information process from which thousands of potential patterns of development can emerge, contributing to the brain’s constantly evolving structure, refinement, and intelligence.
A clear example is pair programming, that is, having two engineers simultaneously work on the same task. On the surface, spending two resources on a task instead of one might seem economically wasteful. It’s unclear whether it makes task completion take shorter or longer. But teams have found that pairs, compared to individuals, produce cleaner, more maintainable code with less defects. Furthermore, pair programming prevents critical knowledge from getting isolated within one person’s mind. When only one team member has deep knowledge of an area of the code base, the team becomes vulnerable when the business requires more changes in that area. The one person with specialized knowledge may be absent or stretched thin between other responsibilities. Pairing is a key ingredient in keeping knowledge flowing through an engineering team. Flexibility, quality, creativity, and efficiency are maximized when all engnieers are capable of working in all areas of the code base.
Pair programming is an example of how product development teams create redundancy in their internal process. Split testing (or A/B testing), in contrast, is an example of how teams learn through creating redundancy in their external outputs. With split testing, teams produce redundant versions of the same page and distribute web traffic such that some portion goes to each version. While the redundant versions of the page serve the same overall purpose, each site instance contains variance along the lines of content or design. This allows the team to discover which page elements perform the best according to company objectives. For examples, product development teams can use A/B testing to determine which home page layout leads to the highest click through rate or which labeling of the buy button (e.g., “buy now” vs “check price”) leads to the highest conversion rate.
Both pair programming and split testing mirror patterns of parallel processing prevalent in the brain. Morgan writes:
A lot of the brain’s activity seems to be completely random and characterized by a massive amount of distributed and parallel information processing. At any one time many parts of the brain may be involved with the same activity or information. This redundancy allows initiatives to be generated from many locations at once, reducing dependence on the activities of any single location. The process generates the multiple, competing “drafts” of intelligence from which an evolving pattern eventually emerges. The redundancy reflected in this system of parallel process is vital in generating a range of potential outcomes, in coping with error, and in contributing to the brain’s flexibility, creativity, and adaptiveness.
Teams that employ redundancy show a respect for the uncertainty of their external environment. They recognize that their assumptions about how to reach objectives are only hypotheses that need to be examined and tested. Creating redundancy is a key technique in discovering surprising, counter-intuitive knowledge about what works and what doesn’t. These “secrets,” as Peter Thiel argues, are instrumental in achieving explosive business victory.