Some good reasons to get involved in social media data
A few weeks ago I was invited to speak at the Social Media Analytics Seminar of IBM/SPSS in Utrecht. I had the privilege to speak after Damiaan Zwietering, Sales Engineer Predictive Analytics at IBM. The goal of both out stories was to give a realistic view on the possibilities on the use of textual data from social media. Damiaan focused on the fact that it is truly doing research without asking questions, and went into depth on different tools and techniques that can be used. The aim of my story was really to put theory to practice and explain how we use social media analytics at InSites Consulting, to give an overview of the do’s and don’ts and more importantly to give insights in what is possible and what is not possible. Social media are there to stay when it comes to marketing and there are some good reasons to get involved in social media data.
- Reduced interviewing bias: When performing social media nethnography we are not asking questions to participants, we collect answers which we try to relate to questions. The social media activities, which generate the information, occur in a natural context. It is part of people’s daily activities and thus in that sense less biased. The observation is unbiased as there is no interpretation or deliberation from the participant’s end.
- New consumer and patient insights: Within the context of a commercial business operation, the only answers that are business critical are those to new and relevant questions – all the rest is just ‘nice to know’. The ‘new and relevant’ bits are where social media represent breakthroughs for companies. Until now, one never knew if the questions asked in a questionnaire were indeed the most relevant and important in the consumer’s (here patient’s) mind. We only get answers to those questions that are explicitly asked. As we are sitting on a mountain of data generated by users we need to ask ourselves what to get out of it. When doing so we will discover topics and patterns which we did not think of naturally.
- Contextual information: The web has a reference frame (like a truth thermometer) built-in which allows identifying relevant and important (new) answers and to qualify and quantify them in context. The input from patients is in their own natural language and they report what they find important from their personal perspective. The information also contains more contextual information, as people report their thoughts and feelings in the “heat of the moment”, not when they are probed to recall it.
- Emotional insights: As opposed to traditional interview based research, the emotional component of an answer is captured and connected to the rational response. An emotionally charged answer (regardless of the type of emotion) is always more important and actionable.
- An additional advantage from a methodological perspective is that one can do research over long periods of time, as the researcher can go back in time as far as the social medium holds relevant data from user posts.
To put the theory into practice, I talked about 3 different cases (check out the above presentation on SlideShare) where we applied social media analytics, and also discussed the possible next steps in research. The first application of is idea generation. Social media data can be used to find “the golden nugget”, to look for ideas that can lead to new product development. Annelies Verhaeghe, Head of Research Innovation at InSites, has look into more than 80 000 English conversations talking about the impact of getting older on the quality of life. One of the conclusions of that study is that the topic ‘eating’ was talked about rather ambiguously. On the one hand elderly people like to eat and enjoy food a great deal. On the other hand, elderly have much more difficulties when eating than before. For the first time in their life they have to watch what they eat and they no longer like the food they loved before.
It might be an idea to start a supermarket for elderly, or a shop-in-the shop, where food can be bought that specifically targets this group and they can buy the food safely without having to think about their diet.
A second application is to look at what consumers are saying about your brand. What are the most popular brands online in your category and what drives these branded conversations? If people are positive, what are they talking about, and if the conversations are negative, what are the reasons for this? We analyzed data from a specific review site in the US, where consumers could give their opinion about ice cream. We not only analyzed the text, but also compared the sentiment with the star rating people where able to give their ice-cream.
Analysis showed that the brands that were talked about more negatively also received a lower star rating. Qualitative deep-dive into specific conversations than gave us in depth information into what people disliked about specific ice-cream. The question then remains whether the consumers that are discussing this topic only are the few, or these opinions and views are more widely spread. Are these conversations representative for the broader population?
InSites Consulting has developed a methodology called ‘conversation mapping’, where we ask consumers to report on their branded offline conversations. This validated quantitative survey gives an answer to questions such as “Who are my brand fans?”, “What can we do for them?”, “What are consumer talking about?” and “How is this influencing my brand’s performance?” In that respect, conversation mapping is the perfect tool to further understand these online conversations and the start of managing your conversations.
The third case where we successfully applied social media analytics was a product launch. Through text analysis we bring structure into these conversations to use them for research. Almost 2.500 conversations were analysed when Telenet launched their product Yelo. Through text analytics, we discovered several themes in the data & for each theme, we determine the size of the cluster and the sentiment of the cluster
1: Act: themes in the market that are often mentioned with a negative sentiment
2: Develop: themes in the market that are often mentioned with a positive sentiment
3: Threats: themes with negative sentiment that are currently not often mentioned but that are explicitly negative for certain market niches
4: Potential: themes with mixed sentiment that are often mentioned. In the future, we can try to influence the sentiment of those themes positively
After the social media study we even co-create further with these early adoptors. We invited Yelo users to share their opinion and co-create Yelo in a prelaunch community. We received more than 1800 submissions and selected 100 users, based on their device usage, social demographics and being influential on new media and technologies.
A prelaunch community can typically take place before launch or in bèta launch of a product and will provide insights into the experience, improvements based on testing, and co-creation around positioning and communication. Based on their experience we collected improvements, input for webcare, guideline for the content to provide and inspiration for the go-to-market strategy in a 3 step process:
Exploring their current and future usage of Yelo, review the different functionalities in detail and by consecutive learning, this was the basis for the co-creation phase where users came up with new features. As fitting in a prelaunch community, experience, web care, improvements, web care are tackled… Of course the Yelo-developers also had some possible improvements in mind, therefore it was interesting to see how they fitted together with the features the users requested. It was key to provide some structure in all these suggestions to further define the development roadmap. Both Damiaan and myself were able to show the many applications of social media analytics, but also gave some insights in what is not possible, what the limitations are and what complementary research methodologies are.
Questions? Contact me!