There has been limited research carried out via social media and so very little evidence that it offers proven business value to the industry at large. This year’s experimental and exploratory Political Buzz study, undertaken by Confirmit and SSI during the UK General Election in 2015, sought to change that. It compared traditional survey driven online panel data with social media data with a view to understanding the differences, correlation and/or complementary aspect of each ethodology on the subject of the UK General Election. The aim of the study was never to predict the winner of the UK General Election – there were plenty of people doing that already. Rather we focused attention on analysing the data differences, data similarities and developing best practices that can help to guide researchers through the social media maze.
Best practice concepts
1. Preparation is everything
Taxonomies form the framework into which social media data is captured and analysed. They are effectively acting as a dictionary and filter with which to understand freeform data and extract relevant buzz from social media noise. Creating the right taxonomy that incorporates the relevant expressions for your study requires domain expertise and is fundamental to effectively categorising the content you capture. Clean taxonomies and expressions are crucial and quality assurance must be built into the development process to ensure accuracy. The rewards here can be significant: with social media research, expressions can be adjusted at any point in the study and retrospectively applied. Access to data is also more flexible than with traditional sources. Think of this like the questions you ask of respondents in traditional research, except that instead you are asking questions of the data.
2. Understand the social media landscape
Analysis of social media data is less a separate step then a continuation of above. Social media is vast. The wrong sources will result in too much noise and little relevant buzz. You must invest time to find the right influencers, aggregators, Twitter handles and blogs that impact your project. This will depend on topic, geography, and a myriad other factors that require subject matter expertise. Going beyond the key sources and digging into sub-sources also provides a rich seam of relevant social media buzz – a further investment in time that can reap dividends in terms of the quality of the insights you gather. The traditional research equivalent is your sample, and just as it is important to have the right sample, it is important to have the right social and online media sources.
3. Fit for purpose
Social media analysis is powered by advanced technology. It combines different statistical approaches, and involves the extraction of concepts into models, the training and re-training of software engines, and the delivery of sentiment analysis. A successful social media research project requires a unique combination of people skills and technology to cater for the sheer volume of data involved. To achieve accurate results, the technology must be set up and managed effectively. Tuning the system is critical to ensuring that you get the best from the previous stages. It’s here that a social media programme needs the skills of experienced social media analysts: tuning the performance of the sentiment engine requires a thorough knowledge of statistical techniques, as well as familiarity with the tuning parameters within the software itself.
4. Commitment and skill
While the project was exploratory in nature, real data was gathered using validated sources, and the results clearly demonstrated that social media research is best used as a complementary approach to traditional research. Quantitative sources are great at extracting the key issues that need to be addressed on a subject. Meanwhile social media research help us to dig deeper on the critical issues that Quant highlights in a much broader way than traditional qualitative methods currently allow. Topics of a social or political nature are ideal for combining social media and traditional research modes – each mode complements the other.
A compelling addition to the research process
A key takeaway from this experimental and exploratory study is that richer insights can be gained through the combination of social media data with traditional sources of research data, than can be gained through the adoption of just one of the sources. The three key political issues were identified through the online panel data, while the share of buzz and sentiment scores added depth to these political issues, and allowed us to examine what people were thinking about. And howed us how strongly they felt about each of these issues. Of course, not every research project or topic will benefit from the addition of social media research. Topics of a social political nature, such as the election used here, are ideal topics to run the combination of social media and traditional research modes. The different modes complement each other, and the richer insights gained from their combination, may not have been so readily achieved through the adoption of just one of the modes on its own.