The Ultimate Guide to Human Emotion-Driven Market Research
Let’s learn more about Looxid Labs’ technology and where we are heading to
What problem we are solving: People buy emotions not things
According to an article “Ask your customers for predictions, not preferences”, in Harvard Business Review, big companies invest an enormous amount of time and money on market research, in order to reach and convert as many consumers as they can in the shortest time possible. Although market research is often viewed as interesting, it could also be useless in terms of limitations that surveys have. To illustrate the questionnaires, consumers are asked to rate their preference(e.g. product appeal, unique selling point, and purchase intention) on a five or seven point scale. And when we analyze the results, it is common that most of the responses score 3.5(in a scale of 5), no matter how well designed the questionnaire and measurement scales are. Measuring purchase intention accurately is especially difficult, since the answers to the relevant questions are notoriously overstated compared to the actual sales performance. Well it’s for sure that this isn’t the surveyees’ fault; the results are drawn from rationally driven responses, whereas consumers tend to be heavily triggered by desires and emotions when making purchases. Needless to say, as long as we stick to the conventional surveys, it’s quite unlikely to obtain a useful insight into consumers’ future behavior and to figure out the reason behind consumers buying decisions.
How others are dealing with the problem: What should be the ideal way of foreseeing consumers’ future behavior?
As described above, despite the inherent difficulties in assessing consumers’ preference using the traditional methods, billions of dollars are spent per every year to anticipate what consumers will want. Businesses and brands make this kind of decisions, since they need to generate awareness for their products, services or brands, as well as differentiate themselves from the competitors in the consumer market. Moreover, when it comes to defining your target audience, it’s hard to find anything as effective as consumer surveys. Indeed, that’s the underlying cause behind the growing popularity of big data analytics and the emergence of preference markets, with the intention to complement the limitation of conventional market research techniques. Big data is a promising technical breakthrough, providing intelligent analysis for tons of consumer data, that could answer the most knotty problem for companies: who buys what, when, and at what price? To answer this question, well-known data-driven tech giants, such as Google, Netflix, and Amazon, have been anticipating the likely success of products, contents, or services with feature learning based preference model.
In-depth technology inside the data-driven innovation: How does Netflix’s recommendation algorithm work?
Then, how does big data analytics enable data-driven companies to better understand users? Based on a recent article released by Netflix, the company knows what users want to watch before they abandon the service and move on to something else within just 90 seconds. And according to Netflix Technology Blog, they provide personalized recommendations for over 100 million members by creating new ranking algorithms and evaluating their performance offline. Following this process, they conduct online measurement of core evaluation metrics including month-to-month subscription retention and member streaming hours by leveraging A/B testing. There are numerous recommendation algorithms at Netflix, but if we focus on the ones used in Netflix’s rating prediction, there are two algorithms: Restricted Boltzmann Machine(RBM) and SVD++ — a form of matrix factorization.
Just like Netflix and other data-driven companies, many research institutions and consumer goods companies are also devoted to developing models that forecast consumer behavior. In particular, a recent design preference modeling study from Design Science Laboratory in University of Michigan, the researchers developed quantitative preference models that predict customer choices among design alternatives by collecting purchase histories or survey results. They applied this idea to automobile purchase decisions using three feature learning methods, which includes principal component analysis(PCA) and low-rank + sparse matrix decomposition. (We plan to further review the detailed principle of the machine leaning algorithms for quantitative preference models mentioned in our article in the future.)
Can new market research tools predict what consumers will like?
Nevertheless, the benefit of the traditional market research method should not be underappeciated. Successful development of preference or forecasting models on products and services can only be achieved by acquiring data, such as consumers thought on the service or product. Other than in concept testing, market research can be used to narrow the new product development funnel, prioritize features, and test multiple aspects of user experience. Conversely, exploring new research methodologies will be a way to deal with the challenges mentioned above.
Indeed, application of neuroscience is considered as an option in some visionary companies. Neuroscientific measures are based on the assumption that people normally can’t successfully identify or articulate one’s internal states or preferences, when asked to do so in verbal/written self-reports.
Several decades of research confirm that consumers’ brains carry hidden information on their true preference. More specifically, a neuroscientific evidence suggests that the ventromedial area of the prefrontal cortex(vmPFC) provide an integrated representation of implicit, subjective valuation and preferences, independent of conscious awareness. Furthermore, implicit preference proved to be predictive of the actual choices people make. And That’s where Looxid Labs comes in.
What should be the next step?
Looxid Labs’ mission is to better understand user’s unspoken emotions using emotion-aware AI connected to human eye and brain. In an attempt to introduce our core capabilities, we are launching Looxid Labs’ Technology Blog. You will have access to more details on, for instance, how our products can be the key to overcome the limitations of existing market research methods by directly measuring human emotional status based on neuroscientific data. Looxid Labs’ technology is mainly powered by blending a large number of different machine learning techniques. In our medium blog, we will cover the full details of our user testing platform, developed algorithms for consumer preference prediction, validation process of our technology, etc. every other week. Please stay tuned for our articles and latest updates to better understand consumers’ unspoken emotions!
Reference
- Ask Your Customers for Predictions, Not Preferences
https://hbr.org/2015/01/ask-your-customers-for-predictions-not-preferences - Improving Design Preference Prediction Accuracy Using Feature Learning
http://mechanicaldesign.asmedigitalcollection.asme.org/article.aspx?articleid=2516869#Abstract - Innovating Faster on Personalization Algorithms at Netflix Using Interleaving
https://medium.com/netflix-techblog/interleaving-in-online-experiments-at-netflix-a04ee392ec55 - Amazon Knows What You Want Before You Buy It
https://www.predictiveanalyticsworld.com/patimes/amazon-knows-what-you-want-before-you-buy-it/3185/
