Why doesn’t the best always win?

Markets are not fair. Big brands enjoy a disproportionate advantage over smaller brands. And, not only do they have more customers, but their customers use the brand more frequently. Kyle Findlay looks to network theory to explain how this advantage works

In order to understand how market share forms, we must agree that markets are inherently social constructs. We either define ourselves in terms of society (for example by the clothes we wear and the music we listen to) or against it (perhaps by joining a counter-culture group such as punks or goths, who, ironically, are often an even more homogeneous group than the general populace). Either way, we cannot fully escape the sculpting effect that society has on us. Indeed, it has been found that even our risk of becoming obese is influenced by what our friends weigh.

Society defines us, even as we define society’s structure and norms in return. Social influence is inescapable. In addition, we are faced with more social influence than ever before in the digital age thanks to social media platforms such as Facebook and Twitter, search providers such as Google and Yahoo and online retailers such as Amazon. This makes it difficult to make decisions in isolation.

Background theory Network theory helps us to understand the ebb and flow of social influence. In addition to understanding who the people are that act as information gatekeepers or exert influence in a social group, network theory helps us to understand how people group together around specific ideas.

When we consider that a brand is really just an idea – a bundle of perceptions, associations and experiences wrapped around a functional product – we realise that network theory can help us understand how people group together around conceptual focal points, linked by a common idea that we call a “brand”. In conventional marketing parlance, we call the group of people tied together by a common usage experience, a brand’s ‘market share’.

The funny thing about market share is that it often doesn’t behave in the way we expect it to. One might rationally assume that a brand’s market share is a function of its quality and how well it delivers its product or service. One might be forgiven for assuming that the best brand always wins and that customers will always have the prescience and good judgement to spot a ground-breaking new market entrant, ensuring that it eventually comes to dominate the market with its own large market share.

Although these assumptions might seem reasonable, we know that they are not 100% correct. In reality, the market share that a brand enjoys is as much a function of timing and luck as it is a function of quality and delivery. The best in the market doesn’t always win, and many new products fail to make a dent in the market regardless of how good they really are.

For example, the Sony PlayStation 3 might be the most technically advanced home-console with the most features out of the box such as Blu-ray and wi-fi support, but the Nintendo Wii, with its dearth of extras and its previous-generation graphical powers, still managed to hit a sweet spot between price and broad appeal that saw customers flocking to stores for units.

 The ‘double-whammy’ effect

Despite brands’ best intentions and superior products, it is often very difficult to predict which ones will do well in a market and, more often than not, new products fail. One might be forgiven for imagining that incumbent brands are surrounded by unassailable fortifications that make it difficult for newcomers to chip away at their market share.

Marketing wisdom does recognise the ‘fortifications’ that form around big brands’ market share. For example, marketing scholars such as Andrew Ehrenberg have long shown that the phenomenon of ‘double jeopardy’ applies in most markets. Double jeopardy describes the phenomenon whereby big brands benefit from a double benefit at the expense of smaller brands: not only do big brands have more customers, but their customers use them more often than small brand customers use their respective brands. This positive feedback means that big brands tend to get bigger and small brands tend to get smaller.

In other words, big brands are self-sustaining entities with an internal consistency that lends itself to continued growth in the future at the expense of smaller brands. This feedback loop represents the fortifications that new and small brands need to overcome in order to survive and grow if they hope to reach a stronger position in the future.

Thus, any marketing initiative that wants to enact change needs to understand the opposing forces at play that seek to reinforce the status quo. Any would-be competitor brand needs to overcome these forces by putting more energy into finding a crack in existing defences or undermining them completely through the process known as ‘disruptive innovation’ which may involve product re-formulation and/or new strategic departure in advertising/ communication.

 Skewed markets

Network theory gives us a formalised name for this phenomenon whereby big brands tend to grow bigger over time, crowding out their smaller rivals in the process. Indeed, there is a generic term for the class of mechanisms that can produce such a crowding-out effect. They are collectively referred to as ‘generative mechanisms’, and one of the most well-known examples is ‘preferential attachment’, which we will focus on for the remainder of this article.

Preferential attachment is a simple concept and it works like this: imagine a store shelf with five brands on it (see Figure 1). The store shelf has ten facings. To begin with, let us imagine that five brands start on equal footing with two facings each. All other things being equal, each brand has a 20% chance of being chosen according to their number of shelf facings.

Now imagine that, as luck would have it, more customers buy brand 1 than any other brand (perhaps because it is in the best position, has great packaging or offers the best value proposition). Regardless of the reason, when it comes to restocking the shelves (Round 2), the retailer sees that brand 1 is their best seller and assigns the brand an extra facing. Brand 1 now has three shelf facings. However, the shelf still has only ten facings in total, which means that brand 1’s extra facing comes at the expense of one of the other brands. In our example, brand 5 loses a facing as a result.

This shake-up on the store shelf changes our brands’ purchase probabilities. Brand 1 now has an advantage with a 30% chance of being chosen, while the remaining brands have a 20% chance, except for brand 5 which has a 10% chance. This small initial difference can quickly spiral downwards owing to a feedback loop that ensures that as brand 1 sells more than its competitors, word of mouth surrounding brand 1 spreads faster than its competitors and that increases its chances of being sold even more. Thus, it expands across the shelf, crowding out the other brands.

 Natural parallel

Evolutionary psychologist and author, Steven Pinker, neatly describes the same process in a biological context: ‘Natural selection works like compound interest: a gene [brand] with even a 1% advantage in the number of surviving offspring it yields [facings] will expand geometrically over a few hundred generations, and quickly crowd out its less fecund alternatives [competitors]. Why didn’t this winnowing leave […] us with the best version of every gene [product]? … The world would be a duller place, but evolution doesn’t go out of its way to keep us entertained.’

People who use brand 1 may enjoy the product just as much as people who use brand 2, but there are more people talking to their friends about brand 1 than brand 2. Consequently, word of brand 1 spreads faster, again reinforcing the feedback loop that will ensure that brand 1 comes to dominate in the long run. This is the role of social influence in markets – a role that is amplified in a world with ubiquitous connectivity and social media platforms that make it easier than ever to share one’s thoughts about brands.

If we were to follow this feedback loop for a long enough period of time during which the shelves are consistently restocked, we would find that the distribution of shelf facings could be described by a class of skewed functions called ‘power laws’. Power law distributions are characterised by a few large observations at the head of the distribution and many smaller observations in the long tail of the distribution (in this case, a few brands would have many shelf facings, while most brands would have few facings).

Power law distributions stand in stark contrast to the more familiar normal distribution, or bell curve, that is an underlying assumption of many traditional statistical approaches. What a normal distribution might treat as an outlier actually becomes the most important observation in a power law distribution.

So, the reason why markets aren’t always fair and why big brands tend to get bigger is because generative mechanisms such as preferential attachment and social influence bias our behaviour as customers, thus shoring up big brands, making them bigger still at the expense of smaller competitors.

 Theory v observations

There is one more twist in the tale though. Once we’ve arrived at the conclusions described above, the next logical question is whether or not our market share data display this characteristic power law pattern.

The short answer to this is, no, they do not show clear-cut power law distributions in terms of market share. However, most markets do show clear inequalities in terms of the distribution of market share, with most markets containing a few large brands and many small brands, and these distributions tend to show the characteristic power-law-like drop-off from the market leaders to the next biggest brands.

In markets that do not initially show clear inequalities between the market leaders and other brands, the issue is usually one of market and category definition. We often found in such cases that we were looking at categories that have not been sufficiently tightly defined such that brands that do not compete directly in customers’ minds as supplementary products have been lumped together. We found that no market perfectly reflected a power-law distribution, despite most markets clearly displaying some form of inequality between the biggest brands in the market and the smaller brands.

Perfect power laws form only when the cost of distributing a quantity is low. And we know that in most markets, there is usually some form of ‘cost’ or trade-off involved when customers make a choice between brands, either in terms of the actual product price or in terms of the time and effort required to find a product that is poorly distributed.

This means that power laws form in frictionless markets and, considering that most markets have some form of friction, all we are left with is a general biasing tendency towards market share inequalities that rarely blooms into full power law distributions. However, this does not diminish the value associated with understanding how these inequalities form over time in the first place. This helps us to understand why the best brands don’t always win and why it is so difficult to introduce new products into a market.

In addition, it gives us some indication of the inertia or gravity surrounding a big brand. This is useful to know as it gives large brands an indication of the ‘buffer zone’ they have to work with – how often can they afford to disappoint their customers before they are abandoned? And, if I am a competitor brand, it is useful for me to know how high and thick the walls are that I need to besiege.

What marketer can really afford not to understand the basic mechanics that drive their market? Understanding market-share formation from a network perspective allows us to understand why customers don’t always appear to act rationally and why business isn’t always fair.

 Acknowledgements

A massive thanks to Anna Retief for her power law curve-fitting efforts across these 46 datasets. In addition, a big thank you to Aaron Clauset of the Santa Fe Institute and Michel Goldstein of Yahoo Research for their guidance in applying the most appropriate method for identifying power laws.

Kyle Findlay is a senior R&D executive at the TNS Global Brand Equity Centre. kyle.fi [email protected]

This is an edited version of a winning paper in the WPP Atticus Awards. The full paper and references are available at www.esomar.org/web/publication/paper. php?id=2171

 like ‘natural selection’, a gene [brand] with even a 1% advantage in its number of surviving offspring can quickly crowd out alternatives [competitors]

The market share that a brand enjoys is as much a function of timing and luck as it is a function of quality and delivery. Many products fail to make a dent in the market regardless of how good they really are

figure 1: a thought experiment showing an imaginary store shelf. as brand 1 expands across the shelf in three consecutive ‘rounds’ it creates a skewed distribution in terms of market share for the five brands that will eventually form a power law