The Three Layers of Social Data Analysis that Rob Sullivan Runs at Social Chain
How do you decide on the best way to analyse social data?
There are different ways you can look at social data and use social media to generate customer insights. We’ve previously shared the concept of ‘social asking’ from Join the Dots, and shared our process for mining insights from social data.
In this article Rob Sullivan, data analyst at Social Chain, tells us about the three layers of analysis he uses. He also offers you his tips for making your social data work harder.
Analysing Social Data is tough. But it doesn’t have to be...
Like everyone analysing social data at the start of their career, Rob found the unstructured nature of text, image, audio, and video tough to analyse. But, sticking at it and putting the time in, he started to get results beyond the basics. It’s important to note here that Rob’s go-to social data tool is Crimson Hexagon. He says that:
“You really need to have to know at least some basics about text mining and machine learning to get to any reasonable standard beyond the basics”.
And Rob believes this work pays off…
“the insights that can be derived from social data can bring real benefit to understanding customers that may not be found from using other research sources”.
He says that although we are relatively far down the road with social media (10+ years), people are still hesitant to look at unstructured data types. For those business who do go in for social media analysis it can offer an advantage.
However, he does stress that to get a deeper understanding of customers and their behaviour, it is important to distinguish the different classes of social analysis as there are a few layers and ways of doing it.
Quickly turning around surveys distributed via Facebook ads or organically on big community pages. Using social to distribute a survey is a method that Social Chain has adopted for research prior to big pitches and in other reports.
In fact, Rob tells us that they would send out a survey and within 48 hours or so they would have 2,000 responses - with minimal spend or prize incentive. He believes this to be disruptive to traditional survey companies, with more of an ‘always on’ method rather than a survey carried out once per quarter [or equivalent].
Posts and Comments
The analysis of text, image and audio from posts and comments.
Like us, Rob believes that there has never really been anything equivalent to this at scale in the consumer and market research industry. He continues:
“Millions of posts will tell you something if you ask the right specific questions. If you don’t you will get noise. But, if you can find out ‘x’ about your customers that higher management were unaware of or had overlooked with the traditional methods of research, that can be very useful.”
Sharing and Engagement
Analysing what people are sharing, liking and commenting on.
Rob warns that this method of analysis is less useful given that these are becoming Pavlovian behaviours.
He does say, however, that the shareability of content when not competition related will tell you a lot about who your customers are at their core and their wider lifestyles and goals.
Social Data Analysis Maturity
Rob’s social data analysis approach is more than measuring engagement or brand mentions. He also moves beyond just taking whatever data comes back from a general social media search query.
His approach focuses on the questions to ask, how people communicate in social networks, working out the best data sources and the different types of analyses to run. This is something that takes time to work towards.
Starting Slowly and Using Your Imagination
To level up your social data analysis Rob advises to start slowly. Start with one data source, for example Twitter, and get to know the source and the types of analysis you can run with it.
Once you’re happy with what you’ve retrieved from the first source move onto the next - such as Instagram and image recognition.
For more seasoned analysts, Rob advises that:
“it’s not so much technical ability or the software that we use, but our own imagination that limits our ability to analyse social data.”
His advice is to think broader, think outside of your working life.
If you’re coming from an analyst’s background, the chances are you could be quite left brained. Rob suggests opening yourself up to other creative possibilities with the data.
He gives the example, instead of reporting on hashtags, share of voice or simple word clouds, you could look at how you can parse the way people talk online or take images or speak in unique ways.
One way of doing this might be to look at co-associations people make between brands on a forum as this might have implications for a company’s marketing activities.
At The SI Lab, we’ve always found the integration of behavioural science into social data a compelling and effective way to get more from the data.
Do you have a unique method of social data analysis?