Social Media Research: From Buzz to Biz
Posted by Jeffrey Henning on Mon, Jun 21, 2010
At the MRA Annual Conference in Boston, Annie Pettit, PhD, with Conversition Strategies, presented "Social Media Research for Results: Turning buzz into sweet, sweet honey". Annie's thesis:
Social media research is market research based on a different data set. It's not a foreign strange thing at all! It's one more method, one more tool that you have at your disposal to reach out and listen to another group of people. As with any method, there are standards processes, like the survey process we know inside and out. Surveys have pros and cons, social media research has pros and cons.
Some of the limitations of social media research include:
- You cannot identify the demographics or geographic location for everyone. "While some data sources come with that information, it is so tiny that you can't generalize from that data."
- You cannot measure incidence or awareness. Just because people aren't talking about a brand doesn't mean they don't use it.
- Not every large company has a large presence in social media. Little known large companies with little social media content include Sysco, ThyssenKrupp, Gazprom and Sinopec. Other companies aren't talked about much for their size: for instance, Chevron, Unilever, ConocoPhillips.
- Not every consumer-facing brand is unique enough to be easily researched. Unique branding is a plus when it comes to social media research.
If your brand is not a high-volume, top-of-mind consumer brand, maybe the brand won't work for social media research: instead, monitor the category and your large competitors.
To demystify social media research, Annie used survey research as an extended analogy.
- Data Quality - "Surveys begin with quality questions. Social media research begins with quality data extraction." You need to carefully frame your extraction: you can't simply take all the data that is returned from a simple keyword search. For instance, taking everything mentioning "BP" will include blood pressure medications, and BP has dozens of other meanings.
- Sampling - "Identifying the relevant sources." While some argue that sampling isn't necessary for social media research, Annie said, "You do not have enough servers in the world to gather all the data. If you aren't sampling, you aren't going to be able to create quality resources. Every crawling vendor does things in different ways: your results will be a reflection of the source of your data."
- Weighting - "Proportioning the relevant sources." Because the crawling vendors return different mixes of data, by source, it is important to weight for consistency. "Weight your sources so that you are not at the whim of the volume of data your tool has crawled and collected." While less than 10% of Internet users are on Twitter, it produces a disproportionate amount of data, up to 60% of the data set from some vendors. You typically need to weight it down. "Microblogs represent the extreme responder: hate it, love it, whatever is top of the mind. Blogs you think through: you write complete sentences, you take your time, you take care with what you are doing. The sentiment depends on where you gather the data."
- Sentiment Analysis - "Scoring the responses." Human and machine scoring doesn't get it right all the time. "There is no perfect method: choose the method that is right for what you are doing. Obviously, humans cannot keep up with the volume of data."
- Standardize Scales - "Ensuring comparability within and between test groups." You can map derived sentiment scores to whatever you prefer for presenting results: a 1-5 scale or even a 0-100 scale.
- Content Analysis - "Categorizing verbatims into meaningful buckets." You can apply thousands of constructs, derived from survey research, to your collected verbatims: purchase intent, awareness, switching, product, placement, promotion.
- Category Norms - "Comparing to benchmarks." For instance, contrast BP sentiment with that of its competitors, to the oil and gas sector or to the energy sector: BP sentiment "took a nosedive but it didn't take the other oil companies down with it."
As Annie said, "Shockingly, social media is not the ultimate of perfection! At the same time, survey research is not perfect, but you know the methods so well that you can brush the negatives under the rug... Sentiment analysis plus content analysis plus validation provides ‘research gold'."