Emotionality Predicts Better Then Ratings
We rely on signals sent by others to make all manner of decisions. From whether to invest in stocks, where to buy a home, how to dress, and what to eat, we care deeply about what we think others are doing. So it stands to reason that customers would seek out signals from others when making decisions digitally. On this assumption, businesses have offered quick and easy systems for customers to provide ratings about their experiences, or product experiences. Whether the Net Promoter Score©, or a 5-point rating system, these quantitative ratings systems have spread to every corner of the world, and into every conceivable language. Unfortunately, these ratings systems have not been the best predictors of future business performance. One reason that should be familiar – I have been asked to give 5-star ratings by last-mile employees in over a dozen countries. This undoubtedly helps the employee, but limits the predicative ability of future behaviour for the business.
While this anecdote might be familiar, it’s not only manipulation by employees that limits predictive ability, there is a bigger challenge referred to as the ‘positivity problem.’ Let me explain, Amazon’s average star rating is 4.2 out of 5, and many others from Uber to Grab, to Lazada and Flipkart will doubtless have similar scores.1 This problem challenges data science and business leaders limiting their ability to accurately predict behaviour and future success, it has even led some multinational companies, like Netflix to abandon this type of rating system altogether.
So How Do We Make Them Better?
The most common approach has been to apply sentiment analysis, a technique that assigns a positivity score (or valence) to the written feedback provided by customers. This techniques provides a positivity score to words such as “Like” and “Dislike” with the former scoring higher then the latter. The challenge here with the current approach to sentiment analysis is that social psychologists have long known that focusing on positivity alone is a weak predictor of future behaviour. So an enhancement was necessary. Enter, Emotionality. Let me explain, look at the images in Plutchik’s Wheel of Emotions below. The emotions at the centre are the most intense, and become less so as you radiate outward.
Emotionality measurements focus on the written feedback from the customer, and a scoring system is used to differentiate words like Enjoyable and Impeccable, with Enjoyable being rated the more emotional word. This approach is more likely to get at the true attitude of the user then conventional sentiment analysis or quantitative ratings systems, and most importantly are more likely to be predictive of future behaviour.
For most of us, we don’t remember the entirety of an experience, whether it’s a shopping decision, or something more important like a holiday. As our experiences become memories, they are ‘compressed’ using a mental heuristic referred to as Peak-End Rule2. This process compresses our memories to two points, the most intense moment (the peak) and the end of that experience. While we can unpack an experience to retrieve the whole experience with some effort, on casual remembrance, these two points are what we will remember. Emotionality understood simplistically serves as a type of file-management system, the more intense those two experiences, the easier it will to retrieve those memories in the future. This has practical implications for how we make decisions. The easier an experience is to recall, the bigger the impact it makes on habit formation (or avoidance).
Consider a home owner exploring new plants to put in the garden at home, for most of their life this experience has meant going to the store, browsing the aisles, asking questions to their spouse, friend or employee, enjoying the smells and the sounds, paying for, and loading the plants into the car. This is a positive experience, one that will be triggered easily when planting season comes around next time, and experiences that are triggered by easily accessible memories are very helpful motivations that lead to habits.
Now, 2021 is different, many home owners, particularly older one’s like my mother, are concerned about going to the store, to engage in this habit as she’s done for many years. Imagine after some deliberation, she elects to purchase plants online from the store, and that the experience is good. She finds the plans she wants, they’re delivered outside her house at the correct date and time, they were in good condition, the soil/fertilizers were also delivered at the same time, a very efficient tidy experience. She was asked to provide a rating and some feedback. She was very happy, and described the whole experience as Terrific! Now, one thing to note, my mother will never give a 5 out of 5, that’s just not how she is, and so the company might think there’s improvements to be made, but that word Terrific if the business tracked it would tell a more complete story. A 4 out of 5 combined with that word should be strongly predictive of future plant buying online behaviour. My mother’s emotion is now more vivid, easier to recall, and because she has made a public statement about it, her testimonial will be more likely to motivate her to be consistent and trigger her to shop online.
Putting this into practice and using emotionality on your sites will require some configuration of the standard sentiment analysis tools provided by the technology companies. Google for example has a tool that allows a business to add custom sentiment analysis. Doing this customization and adding emotionality should measurably improve the accuracy of behavioural predictions, as presented in an April 2021 in Nature Human Behaviour by Matthew D. Rocklage, Derek D. Rucker, and Loren F. Nordgren.3
1Woolf, M. Playing with 80 million Amazon product review ratings using Apache Spark. minimaxir http://minimaxir.com/2017/01/amazon-spark/ (2017)
2 Choices, Values, and Frames: Experienced Utility and Objective Happiness