Three application levels of AI in research and practical advice on how to get started
Across all industries, Artificial Intelligence (AI) is currently en vogue. AI further boosts the key central technologies of recent years, e.g. the Internet of Things (IoT), virtual reality, self-driving vehicles, intelligent agents or smart robots. It depicts the journey from automated to autonomous systems. The difference lies in the fact that automation follows predefined processes while autonomous systems can independently achieve defined goals.
Andrew Ng, the founder of Coursera and former AI manager at Google and Baidu, describes AI as the new electricity. This is quite an appropriate analogy when you consider that electricity has revolutionised mobility, industrialisation, agriculture, medicine and the everyday life of people around the world. If you look back in time, human brainpower was available to meet the challenges of the 19th century. In the 20th century, we programmed computers to solve our problems and, in the 21st century, we train AI so it can solve problems itself without following strict rules.
What does AI have to do with market research?
Today’s AI applications involve chatbots that answer user questions, learn from them, and can derive more in-depth information about consumer behaviors. AI-powered applications make use of people’s motion profiles for tailored services and location-based marketing or classify the probability of illnesses such as depressions based on images posted and filter functions used on Instagram. Ultimately, all these examples are about collecting, processing, analysing, and interpreting data to support decision-making. The link to market research thus becomes obvious.
In the following, we want to shed light on three application levels of AI in market research.
The first level is about improving the efficiency of existent workflows or certain tasks within a process. Just think of the mining methods used to analyse the enormous volume of data on opinion platforms and social media sources. So far, sentiment analyses, buzz analyses and network analyses have been widely used for market research. Machine Learning (ML) methods introduce the next evolutionary step of social media analysis to classify the masses of data more accurately into categories and subcategories, and to highlight their complex relationships. One example is the use of ML to analyse huge amounts of crowdfunding data. In a study aiming to reveal new technological trends, the text descriptions of innovation projects submitted on the Kickstarter platform were explored and analysed using Latent Dirichlet Allocation. Subsequently, the identified clusters could be linked to the funding success on the platform and the DNA of successful ideas could be isolated.
The second application level of AI in market research is the IoT, which promises real-time research and real-time response. In the future, intelligent sensors will generate the data that currently takes us a year to generate, in only a fraction of a second. The main driver of this phenomenon is the immense number of networked devices that will exceed the 50 billion range in a few years. These devices collect consumer usage data ranging from morning dental care, driving to work, wearing fitness wristbands, using food processors, and setting the thermostat. All these provide market researchers with a huge playground to explore customer behaviors in longitudinal studies, whereas currently most market research projects still only offer snapshots of consumers’ lives. Furthermore, through the interaction of IoT and AI, forecasting models will become an increasingly important part of market research. Predictive analytics used to minimise damage to household appliances or reduce accidents in the workplace are just two examples of this trend.
A third application level of AI in market research relies on the emerging innovation field of affective computing. Affective computing deals with the research and development of systems that can recognise and interpret human emotions. Its aim is to enable machines to take into account the emotional state of the user and to adapt their behavior in order to provide an appropriate reaction to the user’s emotions.
The research project and startup TAWNY.ai is an example of affective computing. Biometric data such as heart rate variability or skin resistance of the skin are collected with the help of wristbands or cameras in order to classify the human emotions and affective states. Based on this information networked devices become empathic and can significantly improve the user experience. For instance, future smart homes know which settings make their residents feel most comfortable. Netflix might soon provide programme recommendations depending on the mood of the viewer. And the workplace might soon adapt to the affective states of employees, too. In the future, market research will be able to merge physiological data with facial expression analysis, voice analysis, text and sentiment analysis into a multimodal input to derive a complete emotional profile of consumers.
Joining forces with your client
So what does it take to commence the AI journey with your clients? While building AI capabilities within your organisation and defining a service offering utilising AI are crucial aspects to get started, we want to share some experiences of how to team up with your client and take some first steps.
First, before discussing a potential project, we like to give clients a primer on AI depending on their current level of knowledge. Especially since AI has become the magic bullet for any kind of business problem, a sound understanding is essential more than ever. The primer provides a common basis of knowledge allowing for constructive discussions and helping to manage expectations of what AI can do. For instance, we introduce key principles of machine learning such as supervised and unsupervised learning, feature engineering as well as the prerequisites of training data and its interplay with the development of algorithms.
Second, brief ‘data audits’ proved to be very helpful to fuel opportunity exploration. In this regard, the knowledge market researchers have about their clients’ data (e.g. knowledge you have gained in previous projects) is priceless. With a brief survey of data sources, accessibility and quality, you can highlight potentials and trigger the creative process. For instance, the mere overview of data sources oftentimes serves as a stimulus for ideation. Linking this ‘data push’ with existent examples of how similar data has been leveraged by other companies can further empower clients to match their needs with their data treasure. As a side effect, this initial audit helps to unveil and address urgent issues regarding data management.
Third, we found that workshops used to start a data-driven innovation process should be specifically designed to embrace and consider AI as a potential solution technology. This includes the structure of the workshop and the materials used. The goal is to sufficiently embed the data perspective in the methodology and complement the user, the technology and the business perspective. Creating ideas for data-driven products, for example, requires a different set of aids (e.g. creative stimulus, frameworks or sheets) than ideation in domains without a focus on data does.
A typical phenomenon we observe in workshops focusing on creating AI-based ideas is the vague nature of these ideas. AI serves as the sparkling placeholder in the idea description, which stifles any profounder elaboration or discussion of the idea. To overcome this ‘black-boxing’, we try to encourage teams to flesh out the ideas by adding the data layer. Along a guided process and structured sheets, the teams learn how to link ideas to the data introduced in the audit and create richer concepts.
Lastly, the team constellation for these sessions deserves a closer look. Irrespective of their affiliation (i.e. consultancy vs. client), we create teams consisting of the following personas: design thinkers that have the right mindset and processual know-how; data ‘owners’, who have in-depth knowledge about data availability and who could even access data during the session in a pragmatic fashion; data scientists and AI experts bringing the know-how to the table to empower the team to uncover and realise the potential of the data; product owners who have the user insights and a good understanding of the product strategy. Facilitating a meeting of data and product owners stimulates the creative process and establishes fruitful new relationships in your client’s organization.
In times when everyone is talking about AI, it is crucial to bring the topic down-to-earth. While de-mystifying AI may be disappointing for teams at times, it nevertheless is rewarding to empower teams to leverage this versatile technology.
