5-Point Response Scale & Sentiment Analysis
  • 20 Aug 2019
  • 1 Minute to read
  • Contributors
  • Dark
    Light
  • PDF

5-Point Response Scale & Sentiment Analysis

  • Dark
    Light
  • PDF

The 5-point response scale

This is an old but ongoing methodological discussion on whether we should use a scale with neutral middle category (e.g., 3 or 5 categories) or "force" the respondent to give a certain tendency of "agree/disagree (AD)" (e.g., 4 or 6 categories). The optimum number of alternatives is between 4 and 7. With fewer than 4 alternatives, the reliability and validity decrease; and from seven alternatives onwards, the psychometric properties of the scale scarcely increase further.

We deter from using a 4-point scale with force categorization since it affects our PTM (People To Meet) algorithm. Research also suggests that an AD type scale should offer 5 answer categories rather than 7 or 11 (essentially higher) because the latter yield data of lower quality. It’s better to use a 7-point or higher response scale only with respondents having higher cognitive ability while a 5-point scale works best for the general population.

Sentiment Analysis in Amber chat

A key element in Amber's workflow is our proprietary sentiment analytics engine which infers the sentiment of each response shared by the employees while retaining domain context. This ensures an error rate significantly lower than some of the most popular NLP engines in the market. This also helps Amber speak to employees empathetically as the conversation flows.