MKT360 Chapter 10: Marketing Scales

Study Guide for Exam 3

Attitudes

relatively enduring predispositions to respond to an object

Scale Development

how we create scales 

Scaling

the process of assigning number to abstract concepts

Unidimensional scales

measuring only one attribute or dimension 

Multidimensional scales

measures multiple dimensions or facets of a concept, idea, or attitude 

Correlation analysis

the degree to which one variable changes with another 

Factor analysis

reduces a larger number of items into a smaller subset of factors based on similarity

Characteristics of a Good Scale

• Relatively easy for respondents to understand 

• Clear and concise 

• Provides useful data 

• Discriminates well (the ability to tell two groups of people apart) 

• Limited response bias 

• Reliable and valid 

Comparative Scales

• Rank-order 

• Q-sort 

• Paired comparison 

• Constant sum 

Non-comparative Scales

• Graphical rating 

• Itemized rating 

Either Scales

• Semantic differential 

• Likert 

Rank-Order Scales

• Comparative scale 

• Tend to mimic reality 

• Consumers must be familiar with items being ranked 

• Works best for 5 or fewer items 

• Graphing results: 

  • Ranking of one of the items 

  • Ranking top choice 
    

Disadvantages: Rank-Order Scale

• List may not be categorically exhaustive 

• Respondent may not have knowledge of all items listed 

• Respondents don't follow instruction 

• Difficult to rank middle items in a long list 

• Criteria used in the ranking may not be clear 

• Produces ordinal data, not interval 

Q-Sort Scales

• Comparative scale to rank large sets of items 

• Sort items into groups based on some criteria 

• Quota size for each group may be determined by researcher resulting in normal distribution 

• Ordinal data is produced by this technique 

Disadvantages: Q-Sort Scale

• List may not be categorically exhaustive 

• Respondent may not have knowledge of all items listed 

• Scale is relatively difficult for researcher to set up 

• Sorting requires more effort on respondent's behalf 

• Produces ordinal data, not interval 

Paired Comparison Scales

• Respondents choose between two objects 

• Easier for respondents than ranking a series of items 

• Overcomes order bias associates with rank-order scales 

• All possible combinations must be listed 

• Limited number of items can be compared 

  • Combinations [(n)*(n-1)/2] 

Disadvantages: Paired Comparison Scale

• Reporting results is challenging 

• Requires a small set of items, or respondents will become fatigued 

• Produces ordinal data, not interval 

Constant Sum Scales

• Respondents allocated points among various attributes 

• 10 or fewer items should be ranked 

• Relative distance between ratings can be assessed 

• Produces ratio level data 

• Used to assess brand preference or importance of attributes, benefits, or other characteristics 

Disadvantages: Constant Sum Scales

• Must put faith in respondent's math skills 

• The value of the last item the respondent rates is completely determined by the way the respondent scores the previous items 

• Too complex for long lists of items 

Graphical Rating Scales

• Places a response anywhere on a continuous line 

• Non-comparative scales 

• Scales are normally anchored with antonyms 

• Produces interval level 

• Means and standard deviations can be reported 

Disadvantages: Graphical Rating Scale

• How to convert the response to a number? 

• Sliding scales are more prone to non-response 

• Sliding scales add time to surveys 

Itemized Rating Scales

• Respondents choose a response from a select number of items or categories 

• Non-comparative scale 

• Scales may use words or pictures for categories 

• Easy for respondents to understand and use 

• Most produce interval level data when it can be assumed that there is equal distance between category responses 

• Some produce ordinal data if equal distance cannot be assumed 

Disadvantages: Itemized Rating Scale

• Number of response options must be sufficient for respondents to answer accurately 

• Category labels must match respondent's attitude 

• If there is not an equal distance between categories, the scale produces ordinal data 

Net Promoter Score (NPS)

• Calculated an individual's willingness to recommend a given product or service 

• Customers asked on 11-point scale ranging from 0 to 10 

• Although NPS looks like a typical itemized rating scale, its scoring system is unique 

Disadvantages: NPS

• Businesses frequently misuse NPS 

• Although NPS is correlated with sales growth, existing customer satisfaction and loyalty metrics have higher correlations with market share, likelihood to repurchase, and profitability 

• The NPS calculation ignored measures of central tendency 

Semantic Differential Scales

• Finite number of choices 

• Anchored by dichotomous words or phrases 

• Can be comparative or non-comparative scale 

• Scales usually have 5 or 7 points 

• Produces interval data 

• Scale anchors must be bipolar opposites (i.e., true antonyms) 

• Easily answered when proper anchors are chosen 

• Often used to assess brand image or personality 

• Have high reliability 

Disadvantages: Semantic Differential Scale

• Halo effect may occur 

• Scale only produces interval data when poles are true antonyms 

Likert Scales

• Respondents indicate level of agreement or disagreement iwht a series of statements 

• Very popular in marketing research 

• East to create 

• Easy for respondents to understand and answer 

• Can be comparative or non-comparative 

• Use 5 or 7 points 

• Produces interval level data 

Disadvantages: Likert Scale

• The grid/matrix format does not work on cell phones 

• Acquiescence bias is common of some items are not reversed 

Scale Selection Considerations

• Research objectives 

• Information needs 

• Research participants 

• Mode of administration 

Correlation analysis

• Values range from –1 to +1 

• Positive relationship (+) 

• Inverse relationship (-) 

• No relationship (0) 

Factor analysis

• Determine underlying constructs 

• Factors measuring same construct 

Reliability and Validity

• Correlation analysis

• Factor analysis

Cronbach's Alpha

a key measure of reliability 

How is Cronbach's Alpha related to correlation analysis?

It’s based on number of items and their correlations