Research studies typically seek information relating to the respondent’s behaviours, her attitudes and her classification. Questionnaires are composed of a mix of these three types of information.
Behaviours ultimately reflect the impact of the marketing mix, and serve as measures for gauging performance. Therefore the aim of behavioural questions is to obtain accurate, fact-based information, and to achieve this, the questions should be framed in a manner that elicits accurate responses.
Of relevance is the choice of these words — “usually” versus “last time”. For example, the question “What brand of soft drinks do you usually buy?” will elicit a different response from the question “What brand of soft drinks did you buy last time?” Responses to “usually buy” exhibit central tendency, i.e., the tendency to gravitate towards the few leading brands that the respondent chooses most of the time. It is appropriate for assessing the individual’s purchasing behaviour. On the other hand "buy last time" elicits a wider range of brands that more accurately reflects the behaviour of the population.
Classification information usually pertains to the respondents’ demographics, and is typically used for profile analysis. For instance the profile of consumers (gender, race, age, income, dwelling type, life cycle, etc.) who use Clinic shampoo or the profile of a company (line of business, country of incorporation, sales turnover, number of employees, average age of employee, etc.) that buys Apple computers.
Attitudes have a powerful influence on behaviours. These opinions or beliefs have three related yet distinct components — cognitive, affective and intent. Attitudes are measured in terms of both the direction (positive, negative), and the intensity.
Because they are multi-faceted, it usually takes a series of questions to capture people’s attitude about a topic. Often a Likert scale is employed to assess their opinion on different aspects of the subject.
The first step in constructing a Likert scale is to generate statements covering different aspects, both positive and negative, of the subject. For example, a bank interested in assessing customers’ attitudes to teller service, may consider the following statements to gauge their level of satisfaction:
Efficiently completed transaction/addressed query
Knowledgeable of banks’ products and services
Provided simple and understandable explanations
In cases where marketers are dealing with an unfamiliar subject, they may consider qualitative research to determine the statements or aspects that best reflect the respondent’s attitude to the subject.
Next we need to determine the direction of the statement — does it reflect a positive or negative attitude towards the teller service? Statements that are positive or negative are retained, and those that are neutral are discarded.
The question to respondents is then framed on a multi-item rating scale, such as the one shown below:
|Strongly disagree||Disagree||Neutral||Agree||Strongly Agree|
|Efficiently completed transaction/addressed query||1||2||3||4||5|
|Knowledgeable of banks’ products and services||1||2||3||4||5|
|Provided simple and understandable explanations||1||2||3||4||5|
In this example we do not have any negative statements. However if we did, those statements would be reverse scored (i.e. 5 becomes 1, 4 becomes 2 … and 1 becomes 5) during processing so that there is a uniform scale reflecting the intensity and direction.
A Likert scale is the sum of responses to all the statements or Likert items. In the previous example the scale will range from 4 to 20. It measures what these group of statements represent (i.e. teller satisfaction).
Factor Analysis and Regression
The disadvantage of the Likert scale is that it gives equal weight to all statements. Some statements are more important than others, and some statements are highly correlated to others. Ideally a measure that represents the group of statements, should take these issues into consideration.
In studies where an understanding of the importance of each statement is desirable, researchers may adopt a two-step process — firstly to reduce the statements into a smaller set of uncorrelated factors, and subsequently to determine the importance of each factor.
The statistical technique used to determine the factors is called factor analysis. It reduces the statements to a smaller set of factors — those statements that are highly correlated (i.e. fluctuate together) are grouped into the same factor, and those that exhibit low or zero correlation with each other fall into different factors. A summary measure such as "Overall satisfaction with teller service" may then be regressed on the factors to determine their relative importance. For instance in brand equity research, the equity index is regressed on the factors to derive the importance or contribution of each factor to brand equity.
Rating scales turn consumer perceptions, attitudes and preferences into something that can be measured and compared. Commonly used scales in quant include numeric (e.g. 0 to 10, 1 to 5, etc.), diagrammatic (e.g. smiley faces), continuous (e.g. slider scales used in online research), and semantic-differential scales (e.g. strongly agree … strongly disagree, very important … very unimportant).
Here is a list of things you need to watch out for, while framing questions:
Note: To find content on MarketingMind type the acronym ‘MM’ followed by your query into the search bar. For example, if you enter ‘mm consumer analytics’ into Chrome’s search bar, relevant pages from MarketingMind will appear in Google’s result pages.