Insight From Social Media Data
That word is segmentation. In my previous post I highlighted the various sources of intelligence achievable through social media analytics and possible ways to measure them. Here I take a step back and discuss how those sources of intelligence are obtained through the careful segmentation of social media data.
The focus will be on the following media:
- Earned social media (PR and customer generated)
- Owned social media (both your channels and that of competitors)
- Paid social media (e.g. using social listening to inform paid promotions)
- Unclaimed social media (e.g. related topics that may not mention your brand)
Below are some of the sources of intelligence that were addressed previously but segmented by business function to accomplish varying business needs. Segmenting social media data will allow you to slice and dice information to enable greater insight for multiple stakeholders. It can therefore accommodate for different time and response needs – crisis management for marketing, to trend analysis for business intelligence. Please note that they are very much interchangeable – sentiment data could easily be used for product development, marketing or business intelligence. Similarly competitor research could aid sales as well as marketing and so on. As always feel free to feedback additional areas that you think of or already use.
- Sentiment (e.g. are people upset with the environmental impact of your brand or simply the industry as a whole, if you’re in the car industry for example)
- Products (e.g. understand how various Fiat 500 models are being discussed)
- Complaints / service / faults / bugs (e.g. improve customer service with quick response times and avoiding unnecessary escalations – with subsequent increased exposure)
- Reviews (e.g. crowd-source feedback for product development and marketing)
- Time / trend (e.g. how are emerging topics, trends and issues critical to strategic business decisions?)
- Business unit (e.g. split out journalists tweeting corporate updates versus general consumer content)
- Competitors (e.g. in the short-term, what are your competitors doing and what tactics can you deploy? Longer-term, perhaps consider mitigating competitive threats, taking advantage of white space in the market or areas to lead, share of voice within business categories etc.)
- Subject matter (e.g. if you’re Nike you may be interested in what is being said about running in general)
- Categories (e.g. Wal-Mart might want to know how content is split amongst food & drink, entertainment, merchandise and clothing – what drives intent to purchase? Recommendations?)
- Attributes (e.g. are consumers talking about the improved camera on your phone, or are they loving a completely overlooked feature?)
- Brands (e.g. compare the buzz generated for Coke, Fanta and Sprite – around key events in the calendar?)
- Interests (e.g. group Twitter users into clusters of interest groups as defined by their profiles)
- Campaigns (e.g. how are your campaigns resonating with various audiences – is there an unexpected fan base somewhere?)
- Crisis management (e.g. get real-time alerts for key topics)
- Communities / networks (e.g. where are your customers talking)
- Spokespeople / influencers (e.g. cut through the millions of mentions – the noise – and just listen to key voices)
- Consumer journey stage (e.g. segment the “I want one of those” consumers from the “I’ve got one and love it” and then use programmatic paid campaigns or promotions to target the former)
Below is a model that visualises where each of the above segmentation options could sit across different business functions. It also indicates which data sets could assist the longer-term strategic planning of a business, versus providing short-term tactical intelligence. Competitor research, for example, could sit under tactical as a business is not usually privy to another organisations longer-term business plan. Here Samsung could employ real-time marketing to react to an announcement Apple made. However, competitor data could equally sit under strategic if the data is used to map out longer-term increases of share of voice. In fact competitor data could also come under customer service, product development, marketing and sales. The model illustrates a natural place for certain data to sit, but as with all models it is merely a possible reflection of reality. An additional dimension is added along the x-axis to show which functions and corresponding data sets tend to be more reactive as opposed to proactive.
Here some examples of how segmented social media data for the UK’s largest supermarkets could look using Brandwatch:
As with campaigns and measurement always consider the wider business objectives before you embark on a segmentation adventure. Determine how that social media data is going to be transferred to other business functions in the form of actionable insight – it needs to answer the question “so what”? Lastly, seek to uncover unanticipated intelligence that can drive change, innovation and make the most of opportunities.
- Wider business objectives
- So what?
- Search for the unexpected
For a great real life case study Kevin Ashton’s piece is a must read, in his post he perfectly highlights his use of social media data to develop products. UM Wave is also worth a look as they have taken segmentation globally and offer great insight which you can freely interrogate. For the true data scientists out there, you can visualise various social graphs by using a fantastic open-source package called Gephi. How do you currently segment your social media data?