The data model behind Rate My Meeting

How can we bring in the proven power of ratings to work? Uber shows that ratings drive a transparent feedback culture. Our vision is clear: we want to make work better. Today I'd like to explain how we think 5-star ratings and thumbs-up or down will help us do so.

Photo by Kaleidico / Unsplash

Back to school

Remember statistics in university? When we learned about independent and dependent variables? And the mathematical methods to allow us to measure the significance of relationships among and between these? About paired t-tests? When to use ANOVA or ANCOVA? The difference between linear and logistic regression analysis? Fun! It even became part of my job as a data & analytics consultant. And as I ideated about the long-term value of Rate My Meeting (RMM), it came in handy again. Because if I want to allow you to get actionable insights from the data you collect. What information will you need RMM help you collect? What insights could help you become a better meeting leader?

The dependent variable

For sure, we need a dependent variable—some consistent form of measurements that we can compare over time across all dimensions. We need to be able to average it, and thus should be a continuous numerical value (not discrete). This conclusion allows for multiple options. Most common? A 3, 5, or 7-point Likert scale. We turned to our users to decide. Their verdict? Use 5-star ratings. Respondents will easily understand and relate to this qualifier. Ratings are universal, and thus ambiguity will likely be low.

The independent variables

Selecting independent variables is trickier. Statistically speaking, we first want them to tell something about our independent variable. How strongly did X influence my 4.6 average rating? Second, we want to learn about their significance among each other. Did either X or Y significantly affect my rating more? And then, finally, can we derive patterns or constructs of variables that determine meeting success?

From an actionability point of view, X and Y in the previous example need to be things you can act upon. The prevailing theory distinguishes between two types of competences. First, there are easily observable skills, knowledge, and behaviors. Gained through education and training, these dimensions can relatively easily be adapted (Boyatzis, 1982). At a deeper level are the intrapersonal and interpersonal perceptions of an individual. Formed through social behavior, these competencies can be changed only over a more extended period through conscious intention (Spencer & Spencer, 1993).

Examples of observable competences

  • Prepared agenda
  • Time management
  • Chairmanship

Examples of personality traits

  • Empathy
  • Integrity
  • Charisma

We want to be able to measure both. And as we train and mature our data model, we have decided to keep our independent variables, for now, discrete and categorical (0 or 1). Trying to nuance the later across a more extensive scale we feel will prove especially challenging. Fewer choices increase response rates and make it easier to derive statistical power.

Given this binary choice, we went back to our users and learned that they appreciate a thumb-reference most. They said they are used to throw "likes" around on social media using them today, making this a small, logical step. Thumbs seem universal, but at the same time, we realize that different cultures prefer different options more. Make sure you let us know if that is the case.

Thanks for reading, as always: be sure to share your thoughts or feedback. Always happy to connect.