Wednesday, 28 October 2009

Behavioural economics for marketers

I had a very interesting conversation yesterday with the head of a branding agency who shares our vision of developing a scientific foundation for the practice of marketing. My view (and that of Rory Sutherland of the IPA and Ogilvy) is that behavioural economics can provide this foundation. But as a discipline, behavioural economics is not yet mature. It has a hole at its centre which needs filled before we can genuinely say that there is a science of marketing. Filling this hole is the goal of my research and this article outlines how it can be done.

The key step in the development of behavioural economics is to close the gap in our understanding between neuromarketing and phenomenological descriptions of behaviour.

The practice and underlying science of marketing can be represented as a multi-layered set of disciplines, from highest to lowest level:
  • The craft and folk wisdom of sales and marketing
  • Experimental phenomena from behavioural economics
  • ???
  • Neuromarketing and the application of brain science
  • Biology of human needs
  • Organic chemistry and physics
This gap is where cognitive models provide the missing link in both explaining and predicting human behaviour.

Beneath this level, neuroscientists are discovering the specific physiology of the brain and how it responds to certain stimuli. There is a lot of useful work on how the limbic and motor systems control certain responses below our conscious awareness.

Above this level, there are experimental results which show statistical patterns of behaviour: experiments for example on self-control, hyperbolic discounting, valuation heuristics and relative price judgments. However, these experiments generally come with no underlying model to explain the results. The behaviour of the subjects can be shown to be inconsistent with the conventional economic model of rational preference, but to date there has been no alternative model.

Neuroscience gives some indication of the basic underlying mechanisms, but no neurological theory can explain hyperbolic discounting or give any insight into the actual discount rates that people apply in different circumstances. Neither can neurology explain why people who are primed with words or phrases about age and frailty walk more slowly than those who are cued with words about vibrancy and youth. Nor can it explain how people judge value by relative price instead of absolute numbers.

The reason for this is that we are intelligent, thinking beings; we are not driven purely by dumb emotions and drives; our brains are powerful tools for translating those emotions into actions via knowledge, and for learning new kinds of actions over time. To understand these processes, and close that gap in the understanding of marketing, cognitive models are required.

Cognitive models provide insight into how people think, learn and know; show us how people make decisions; clarify what motivates people and what they value; and shows how much of one value they are willing to trade off for another, and under what circumstances.

How do we build a cognitive model?

A cognitive model needs to have a plausible basis in the neurological capabilities of the brain and simultaneously to explain the observed behaviour of real people.

Neurological constraints provide some limitations on the processing capacity of our cognition, and on the degree of attention and focus we have. Observational constraints show that we are capable of powerful high-level thought and decision making, while highlighting specific consistent biases or flaws in our reasoning, which shed light on the mechanisms that provide that capacity for thought.

No model gives a perfect prediction of behaviour - that is in the nature of a model. But can we build a model which is sophisticated enough to successfully predict a high proportion of experimental results, while still being simple enough for us to use in practice?

The way to do this is to make two simplifying steps. First, move one level back from the experimental phenomena to understand their proximate causes - that is, the direct reasons that people take the actions they do in experiments. These reasons or motivations can be modelled, and their relative strengths can be estimated either via lab experiments, or by placing individuals in specific simulated decision contexts and recording their answers.

Second, move up from the neurological level to a scale that is more understandable on a human level: the scale of individual concepts and facts. The relationships between these contexts and facts can be determined for a given individual partly via the language they use, and partly by examining their problem solving mechanisms (by experiment or by posing puzzles).

Once we are dealing at the level of motivations and concepts, the relationships between them start to become clearer and it is possible to derive functions relating the two. Individual behaviour can then be represented as a stimulus-response cycle; and we can understand how to influence those responses by analysing the communication processe with information theory and the adoption of new ideas via the psychology of learning.

A mathematical interlude: feel free to skip this

I will briefly outline here a mathematical model of the stimulus-response process. This is not the only possible model, but it has proven to be a highly practical and manageable way to represent and understand cognition and behaviour.

First we develop a value model whose purpose is to represent a relevant subset of human motivations. The scope of this model depends on the application: if you are a law firm the relevant motivations are to do with reducing risk and winning court cases, while the relevant motivations for a branding agency are the desire for innovation, growth and a positive social or self-image.

The internal structure of this model is a hierarchy of motivations which ultimately are based on underlying neurological processes in the brain, pain-avoidance, pleasure-seeking and the satisfaction of physical and material needs. Because the brain is a highly malleable construct and very efficient at developing heuristics, people are not generally conscious of pain-avoidance and pleasure-seeking in their day to day decisions, but use proxy motivations. Maslow's hierarchy of needs is a good example of how to understand the way we build up multiple levels of motivation which become more abstract and gain a broader scope at higher levels.

In an economic context, one of the key motivations is money (with a critically important distinction between the individual's own money and the money they control on behalf of an organisation). Our goal as marketers, ultimately, is to get people to trade off one value against another: specifically, money in return for some other value that we can satisfy with our product or service.

Having constructed this network of motivations, we transform it into a multi-dimensional vector space, whose base is a set of vectors v = v1, v2, ..., vn, each representing a single motivation or value.

At any given time the individual is considered to be at a single point within this space, and we define a function u = u(v1, v2, ..., vn) representing their utility at that point.

[In fact, utility is not an invariant function over time but has a path dependence. However, for purposes of this model it is easier to represent it as a non-path-dependent function but incorporate a time variable instead: u = u(v1, v2, ..., vn, t). This does not affect the current analysis. Similarly, we can use a model with no concept of utility but where individuals attempt to independently maximise each value. This is appropriate for simple computational models, but in all biological contexts agents ultimately face tradeoffs between multiple goals. A separately differentiable utility function is a good way to model these tradeoffs, but it is possible, equivalently, to specify a matrix of "exchange rates": x(m, n) = δvm/δvn.]

An individual's goal is to seek maximum utility in this space by achieving the highest possible value of u. Classical economics assumes that they have visibility of the entire space and how it will change over time, and can choose the path which provides highest total utility, integrated from the current time to infinity. This is where cognitive/behavioural economics diverges from the classical model in three ways.

First, individuals cannot move at will within this space. There are only a limited number of actions available to them: these represent the economic transactions or other decisions that are open to them at a given time. Each of these represents a tradeoff: a reduction in value on one dimension and an increase in value on another (the tradeoff is present in all non-trivial situations, because we assume that if people can make a gain with no corresponding tradeoff they will immediately do so until that opportunity is exhausted. Decisions of economic interest take place once the "free lunches" have already been eaten).

Second, individuals only have local visibility over this utility function. They do not have certainty about the utility that will be available to them at some distant point. People do have a reasonable idea of how their utility will increase or decrease if they make a change in certain values (such as "amount of food consumed", "money in wallet" or "prestige of brand image"). Any long-term goal can only be evaluated by extrapolating its effect from local changes in these values. This is one reason why we often take actions which make us happy in the short term, but from which we suffer in the long run.

Third, we have limited attention. We cannot simultaneously calculate the effect of changing dozens of variables all at the same time. So in general it is rarely practical for us to make tradeoffs between three or more values; mostly we will only focus on two at a time. It seems that the brain has an internal mechanism for shifting our focus from one point to another periodically (the boredom reflex) so that we are encouraged to regularly re-evaluate the tradeoffs that are available. When we shift focus, it is likely that we have learned which tradeoffs are particularly attractive to us: for a tradeoff (x<->y), that's where the value of δU/δx is substantially greater than δU/δy. Importantly, our focus and our awareness of which actions are available is strongly impacted by sensory input - this has major importance for the advertising industry.

The coefficients of these tradeoffs (the derivatives of the utility function) are dependent on several factors, some of which marketers can influence:
  • where the individual is within the value space
  • the individual's cognitive state and the magnitude of the different weights in their motivation hierarchy
  • which components of the motivation hierarchy they are currently present to

So, what should marketers do?

On the basis of this model marketers need to take the following five types of action to encourage buyers to purchase their products or services or to increase the prices they can achieve:
  1. create additional actions (available tradeoffs) in the buyer's value space
  2. prime the buyer so that your actions are cognitively present
  3. influence the "exchange rates" between different value dimensions - for instance to persuade consumers that the exchange rate between safety and money is high
  4. build "bridges" between different parts of the value space so that people can be persuaded to undergo short-term pain (e.g. spending money) in return for long-term benefit (e.g. a comfortable retirement)
  5. in the longer term, inventing and communicating new values which can add another dimension to the value space, and ensuring that their brand is the individual's key association with that value

And behavioural economics? It fits into this model very well. Each of the key experimental phenomena of behavioural economics is an expression of one of those actions:
  • priming (2)
  • anchoring effects on pricing (2) and (3)
  • branding in general (5)
  • interest-free credit and other hyperbolic discounting phenomena (4)
  • memory effects and availability (1)
Applying the insights from this theory can increase both the quantity of sales you make and the prices you can achieve.

All that remains to complete your sales and marketing plan is to model the values, measure the exchange rates and invent the right techniques and actions to apply this theory. I'll give some examples of those processes over the coming weeks.

No comments: