Emotions, how people feel, are key to determining how we behave and respond to stimuli. Many, if not most, companies still are asleep at the wheel and aren’t doing anything on this front. Eventually they will wake up and realize that no effort at measuring and understanding virtually any domain of consumer/human behavior – from customer experience, employee engagement, user experience and brand to message, product and package testing – is complete without at least some effort at capturing the emotional dimension.
Here’s the challenge: measuring emotions isn’t easy. Our emotions are not immediately accessible to our thinking selves. So when we ask a question about feelings, we are pushing people to look into something that is inherently murky and obscure, even to ourselves. Once we consciously reflect on a question and give a “reasoned” answer we have left the realm of subconscious or visceral emotions and feelings for the world of the thinking. While people might be able to engage in introspection, explore and eventually identify their emotions, this is not easy or quick – and not a very practical approach for time-constrained data collection.
While Freud’s psychoanalytical model may no longer be considered good science, 100 years after Freud the problem is largely unchanged. It’s a classic Catch 22:
• We can’t even begin to understand what motivates customer or employee behavior without a sense of their underlying emotions.
• As soon as we ask people to rate or assess their feelings we force people to undertake a cognitive activity that immerses them in the world of conscious thought.
• But our feelings dwell in the land of the subconscious (or quasi-conscious at best), a world that is difficult to penetrate by our thinking selves.
That said, we still need to at least try to measure feelings. What’s a poor marketer/researcher/strategist to do?
The Traditional Approach
“On a scale from . . . please rate how you feel about your most recent experience/this advertisement/this brand . . . ” The traditional survey approach is to directly ask respondents to rate or score their feelings on some type of scale. This can be a numerical or categorical scale or a scale using symbols or emoticons.
This approach is easy to implement, conventional, inexpensive and practical. It parallels how we ask customers and employees to rate a company on other attributes. The results can be readily integrated and processed with any rational assessment of performance.
Given that we don’t consciously know our feelings, however, this approach is far from perfect, as it violates the fundamental premise of the separation between thinking and feeling by asking people to think about their feelings. It certainly is better than ignoring emotions, but clearly is a blunt instrument. At the very least (or best) it tells us how we think we feel, which is better than ignoring feelings altogether.
Neuro and Physiological Measurement
We know that feelings elicit physical reactions. Why not directly measure neurological and physiological responses that can inform us how people are feeling? We can wire people up and measure how their eyes move or how their heartbeat, breathing, perspiration or brain waves change in response to stimuli. Measuring autonomic reactions is not meant to totally circumvent the need to ask people to try to articulate their feelings; rather, such measures are designed to complement what people say.
If you are (or can rent) a neuroscientist and the technology, these approaches are great for small scale tests of how people react to stimuli ranging from ideas, concepts, symbols, imagery and colors to messages, experiences (both digital and physical), ads, packaging and products. How does someone feel when they are using your website? Sitting in a car? Trying on clothing? Watching/listening to your ad? There are countless applications where up-close-and-personal physiological measurement can be immensely informative.
These approaches, however, tend to be somewhat complex and not readily scalable. Wiring folks up means they need to be in proximity and in small-ish groups. (Efforts at using webcams for large samples have failed.) You definitely need some specialized (albeit available) skills and tools to make this work.
If you want a representative sample or need data at scale, associative measurement might be a better choice than physiological. Implicit Association Tests (IATs) can measure respondents’ visceral feelings or the strength of their association or attachment to a belief through inferential measurement that is less intrusive than traditional survey questions and more compatible with subconscious measurement. Based on social psychology work regarding bias, IATs generally rely upon the speed with which people answer questions about the association between two topics to measure the strength of someone’s attachments. (Yes, it turns out that the speed with which you answer a question is associated with how strongly you feel about your answer.)
IATs do require new data collection, whether embedded in an existing survey or a new survey. The format of questions is somewhat gamified and far quicker than the mind-numbing ratings of long lists of attributes. IATs can be completed via laptop or mobile device, so there is no issue of scale. Design, analysis and interpretation are not straight-forward, however, so this is not a DIY approach, but help is available at reasonable costs.
Emotion or Empathy Analytics
Unstructured text is a goldmine for exploring feelings. Instead of asking someone to rate or categorize their feelings, we can take their comments – whether in response to a question or unprompted comments they might post or offer – and analyze their words to determine their underlying feelings. Does the person express frustration or perhaps anger? Are they surprised or maybe delighted? As fellow humans, we can do this manually, much like coding any open-end. Human decoding of emotions, of course, is not scalable. But we now have the tools to automate this function to scale, including the availability of complex algorithms to root out the underlying emotions embedded in what someone has said or written.
This is not simple positive/negative sentiment analysis or scoring favorable/negative comments. Rather, it’s developing an algorithm, building a lexicon, curating for a particular company or industry and using machine learning tools to unearth the underlying emotions embedded in our words.
This can be accomplished in virtually real-time, at scale, using any unstructured text, from survey open-ends to social media, from call center transcripts to talk-to-chat. So depending on what you want to know, you might not need any new data collection. Want to know how people feel about any of the estimated six million apps available through Google and Apple Play combined? Just scrape the data and run it through this type of empathy analytics engine.
OK: sarcasm still (and always may) be a problem. And people still need to express themselves in words. But you can turn this laser to analyze the feelings of customers, employees or any group, or even evaluate outbound communications.
Exploring the emotions that motivate customer or employee behavior (or the emotions which you and your firm may be projecting) is a must-do. But how?
Of course, there is no one-size-fits-all answer: it depends, and using multiple tools to capture emotions from different angles always is advised. Are you looking for more qualitative or quantitative applications? Do you need a representative sample at scale or will a small population suffice? Do you have a survey you can use and with how much open real estate? Or are you looking to analyze existing data? Do you need real-time results? And, of course, what are you trying to accomplish?
Now you might be feeling a bit frustrated about now . . . in which case, just let me know.