Based on this chosen method, the primary variable I would be working with to measure customer loyalty would be at the Ordinal level.

Here’s a breakdown of why and how this affects data evaluation:

Why Ordinal is the Primary Variable for Customer Loyalty (in this context):

To directly gauge loyalty, I would likely employ a customer loyalty scale within a survey. This scale would present customers with a series of statements or questions about their likelihood to:

• Repurchase: “How likely are you to purchase from us again in the next [timeframe]?”
• Recommend: “How likely are you to recommend our company/products/services to a friend or colleague?” (Net Promoter Score – NPS style question)
• Resist Competitive Offers: “How likely are you to switch to a competitor if they offered a slightly better price?”
• Overall Loyalty: “Overall, how loyal do you feel towards our company?”

For each of these questions, I would use a Likert scale or a similar rating scale. For example:

• 1 = Not at all likely / Strongly disagree
• 2 = Slightly likely / Disagree
• 3 = Moderately likely / Neutral
• 4 = Very likely / Agree
• 5 = Extremely likely / Strongly agree

The responses to these scaled questions represent ordinal data.

Describing Why This Variable is Ordinal:

• Ordered Categories: The response options on the Likert scale have a clear order. “Extremely likely” represents a higher degree of loyalty than “Very likely,” and so on.
• Unequal Intervals: While there is an order, we cannot definitively say that the difference between “Slightly likely” and “Moderately likely” is the same magnitude as the difference between “Very likely” and “Extremely likely.” The psychological distance between these points is subjective and not necessarily equal.
• No True Zero Point: There isn’t a true zero point indicating an absolute absence of loyalty. A rating of “1” still represents some level of (negative) sentiment or low likelihood, not a complete lack of any feeling.

How the Ordinal Nature of the Loyalty Variable Affects Data Evaluation:

The fact that our primary loyalty variable is ordinal has significant implications for how we can analyze the collected data:

• Appropriate Statistical Analyses:

  • Descriptive Statistics: We can calculate frequencies (counts and percentages) for each category and determine the mode (most frequent response) and the median (the middle value when responses are ordered).
  • Non-Parametric Tests: Because the data doesn’t meet the assumptions of interval or ratio scales (like equal intervals and a true zero), we must use non-parametric statistical tests for comparing groups or examining relationships. Examples include:
    • Chi-Square Tests: To see if there are significant differences in the distribution of loyalty levels between different groups (e.g., customers who interacted with the new protocol vs. those who didn’t, or different customer segments).
    • Mann-Whitney U Test (Wilcoxon Rank-Sum Test): To compare the median loyalty scores of two independent groups.
    • Kruskal-Wallis Test: To compare the median loyalty scores of three or more independent groups.
    • Spearman’s Rank Correlation Coefficient: To assess the strength and direction of the monotonic relationship between loyalty scores and other ordinal or interval/ratio variables (e.g., satisfaction with the specific service interaction).

• Limitations on Arithmetic Operations: We cannot perform arithmetic operations like calculating a meaningful mean (average) or standard deviation on ordinal data. While some researchers might calculate a mean for Likert scale data for ease of interpretation, it’s important to remember that this assumes equal intervals, which isn’t strictly true for ordinal data. Therefore, interpretations based on means should be cautious and accompanied by non-parametric analyses.

• Visualizations: Appropriate visualizations for ordinal data include:

  • Bar Charts: To display the frequency distribution of loyalty levels.
  • Stacked Bar Charts: To compare the distribution of loyalty levels across different groups.
  • Box Plots: To compare the median and spread of loyalty scores across groups.

• Interpretation: When interpreting the results, we should focus on the direction and magnitude of the differences or relationships based on the ordered categories. For example, we might say that a significantly higher percentage of customers who experienced the new protocol reported being “Extremely likely” to recommend the company compared to the control group.

Supplementary Qualitative Data:

To enrich our understanding and provide context to the quantitative findings, I would also collect qualitative data through open-ended survey questions or focus groups. This data, being nominal (categories without inherent order, e.g., reasons for their loyalty rating) or ordinal (e.g., ranking reasons for satisfaction), would be analyzed using thematic analysis to identify recurring themes and provide deeper insights into the “why” behind the loyalty scores. This qualitative data can help explain the quantitative results and uncover nuances that a purely quantitative approach might miss.

In summary, while we are quantifying a subjective concept like loyalty, the most direct and practical way to measure it in this context yields ordinal data. This necessitates the use of appropriate statistical methods and careful interpretation that respects the ordered but not necessarily equal nature of the response categories. The supplementary qualitative data will provide valuable context and depth to the quantitative findings.