Jeremy Griffiths Jeremy has been with Maritz Research since 1998 and is Director of Marketing Sciences for Europe. He is involved in projects across all sectors and methodologies and brings this broad expertise to provide analytical and statistical consultancy to recommend the most appropriate methodology for all the company's research projects. Read the full biography here.

Non compensatory choices:

...or do as I say, not as I do

By Jeremy Griffiths - 21st March, 2011

Introduction

Understanding the process people go through when choosing one brand over another is at the heart of what we do as researchers and marketers. But some of the most widely used approaches may be too simplistic to be really accurate. In this article I would like to review different choice models and suggests new ways for understanding how people chose.

Customers choose brands that meet their needs and satisfy their wants. Marketers who better know how their customers trade off their “needs and wants” to make choices will have a competitive advantage over those who do not. As researchers we support marketers with brand studies to assess the importance of tangible and intangible brand attributes. We design conjoint studies to test the appeal of new features or price points and so on. These studies are commonplace and so is the analysis that underpins the modelling of brand choices. The vast majority of brand choice models use a method called Multinomial Logistic modelling (MNL). This powerful model won its inventors (James J. Heckman, Daniel L. McFadden) a Nobel Prize for economics in 2000.

There are a number of differing opinions on MNL. Some colleagues and clients praise the method of modelling a choice rather than simply a preference rating. Others are more concerned by a rationalisation of what might be considered a largely emotional choice. Considerable consumer psychology literature exists on the way people use information about attributes to make choices among brands. Paul Green and Yoram Wind provided a famous discussion on this topic in their book, Multi-attribute Decisions in Marketing (Dryden Press, 1973). They describe two classes of models for customer decision-making.

		Compensatory choice models
		Non-compensatory choice model

In compensatory choice models, the overall value (utility) of a brand is a weighted average of the brand’s position on some set of attributes, where the weights measure attribute importances. Consumers, seeking to optimise their choices, select the brand with the highest overall utility. The compensatory aspect of these models derives directly from the weighted averaging that occurs. A shortfall on any one attribute can be made up for, that is to say compensated, by advantages on other attributes.

For example, a consumer may have a preferred brand of hair shampoo, but if a competitor brand is on a price offer he or she may compensate what they believe to be the better product of their preferred brand for the lower price of the competitor brand. This is not indicative of how everybody selects brands and indeed individuals may alter their choice processes across different products and sectors. The key point however is that the MNL model, of which most brand research is based, assumes all consumer decisions are compensatory in nature.

Green and Wind also describe non-compensatory choice models. Consumers making non-compensatory choices consider attributes sequentially and benefits on some attributes may not overbalance shortfalls on others. For example a mother wants to buy freshly squeezed orange juice, but her children dislike the feel of pulp in their mouths and refuse to drink it. She will therefore not consider any brand of orange juice containing pulp. Adding antioxidants, vitamins or cutting the price won’t sway her pulp averse children; to her children there are none of the other benefits that can compensate for pulp!

Of course, hybrid approaches may be used where a non-compensatory first stage limits the selection, followed by a compensatory second stage to make the final choice. For example, if I wanted to buy a smart phone. I could choose between a Blackberry, an Apple or an Android based handset. My non-compensatory choice could be towards an Android handset because I believe it offers me a superior operating system. The benefits offered to me by Apple or Blackberry do not outweigh my desire for an Android. I may have a next layer of non compensatory choices to make around which network my handset operates under. Since I believe one network to be superior for network coverage and customer service my choice cannot be swayed away from that network. After that, precisely which Android handset I ultimately buy may then be determined by more compensatory choices as I make trade-offs between designs, functions and price.

In a recent survey, 266 members of the American Marketing Association (AMA) reviewed nine employment compensation plans and chose among them. Asked to indicate which of four brief descriptions best captured their choice process,

		3 percent selected a description of a compensatory choice process
		28 percent chose a process that combined compensatory and non-compensatory elements
		58 percent selected a non-compensatory process called the lexicographic model
		11 percent went with a close relative called elimination by aspects

All in all, 97 percent of respondents, all professional marketers, described their decision process as wholly or partly non-compensatory. This simple example shows that we all make some level of non compensatory choices between options yet we rely on research that assumes that our customers are different and use purely compensatory choices.

Time to do something different?

A straight forward MNL choice model does therefore have some limitations. An alternative method, and the most popular method of choosing by the AMA members in the employment compensation plans example above, is more closely related to the Lexicographic Choice Model.

The Lexicographic Choice Model suggests that a customer has a hierarchy of attributes, which are considered in sequence. The customer first determines how the brands compare on his most important attribute and, if one brand dominates the others, he chooses that brand. If not, he looks at those brands performing best on his most important attribute and compares them with respect to his second most important attribute. If one winner results, he chooses that brand; if ties persist, our customer goes down the list to the third attribute and so on. The process concludes when he finds a winner or runs out of determinant attributes, in which case he chooses randomly among the equivalent remaining brands.

We tested a number of consumer and business-to-business studies in a variety of categories including packaged goods, financial services, telecom, hospitality and retailing to compare the Lexicographic Choice Model against the MNL Model. MNL and Lexicographic models could be run on the same data sets by simply adding a few simple extra questions to a standard brand choice study in order to collect the extra information needed about non-compensatory effects.

The results showed the percentage of correct predictions is significantly higher for the Lexicographic model than for MNL in all 10 studies.

This series of tests demonstrate that there are advantages to using the Lexicographic Choice Model over the more widely used MNL approach. It provides marketers with an alternative to test how our customers make brand choices, rather than merely assuming they use compensatory procedures.

Jeremy Griffiths

Comments on this article

Want to share your thoughts...?

Want to share your thoughts?

NOTE: Please note that this board is moderated, and comments are published at the discretion of the site owner.