The proliferation of social media listening and monitoring platforms is a sure indication that business has embraced social media as a valuable communication channel for brands and their customers to exchange information, obtain and deliver customer service, and build relationships.
With so many socially-active consumers willing to publicly share their feelings and experiences, social media has emerged as an important source of business intelligence data that drives customer insights, product feedback, reputation management, influence and reach tracking, sales forecasting…and the list goes on.
However, a quick survey of social intelligence solutions suggests that solution providers and the big brands haven’t yet effectively leveraged social media to address one of the biggest pain points in customer management; namely, customer churn. Making the connection between churn and social media seems like a logical and natural next step in the evolution of social analytics. After all, millions of complaints, rants, criticisms, and threats are broadcast each day on social channels, many of them overt signals letting brands know that they should be expecting a cancelation call very soon. Yet, despite the obvious ocean of churn signals just waiting to be harvested from social media, you would be hard pressed to find a social listening/monitoring solution that does anything really useful in predicting churn. Why?
The reason lies the way that the current crop of monitoring solutions are configured along two important search axes:
- Using explicit rather than non-explicit search parameters
- Most social media listening is explicit – tuned in to catch specific brand mentions (e.g. “Verizon”, or “Rogers”), specific product names (e.g. “FIOS” or “U-verse”), or specific hashtags and keywords (e.g. “#sprintsucks”).
- Non-explicit content (everything else) gets filtered out as noise.
- Monitoring the marketplace rather than monitoring a defined user group
- If a company wants to understand what the marketplace is saying about their brand, product, and reputation, or if they want to be able to respond to all socially-generated complaints or inquiries, it makes sense to monitor the entire social universe. That’s easy enough to do.
- However, companies that prefer to focus what their customersare sharing on social channels first need to make the connection between each customer and their public social profiles, and then monitor only those profiles. Making that first-time connection between the offline and online realms is not as easy, especially when multiple users are associated with one account (marriage, family, business).
Nearly all of the currently available social media monitoring technologies are configured to scour the entire social universe and capture only explicit mentions. In order to prevent users from being swept away by a tsunami of social data, conventional social listening tools are obliged to filter out non-explicit messages as being irrelevant. So if you were Comcast, your conventional listening tools would pick up the following explicit tweet: “Comcast connection down again. #nothappyaboutit”, and plunk it into your dashboard or social CRM for follow up. Mission accomplished.
But if you stop there, a lot gets left on the table. The non-explicit signals, the ones that are passed over by the social listening tools, are often a richer source of customer insights, and many of them are related to churn. The tweet: “I am so done with cable being out, time to start looking at options” is non-explicit, and would not be caught by most social listening solutions. However, if you are the cable company serving that user, it’s an enormous churn clue, as long as the tweet can be attributed to a specific account holder in your customer database. That is the very approach that ThinkCX has taken. By focusing on your customers and including non-explicit content, we’re betting that you will significantly increase the ROI of your social monitoring solution. Instead of limiting your social listening to a read and react exercise, you’ll be adding a new predictive dimension and gain incredible insights about what your customers plan to do, based on their non-explicit, nuanced posts. The ROI really starts to kick in when you start to reach out pre-emptively to those customers, well before friction becomes flame.
Any technology that claims to leverage social media signals to predict churn must possess three important characteristics, namely;
- Attribution. Attribution in this context refers to the technology’s ability to associate a company’s individual customers with their social profiles in such a way that the past and future messages they post are automatically connected back to that specific customer. Without attribution, it is nearly impossible to predict individual churn situations. Ideally, attribution also includes the capacity to associate social content not only back to primary account holders, but also to the relatives or companions who may be sharing their account – which is very difficult to do.
- Cost efficiency. The methods of obtaining and analyzing social messages at scale can be prohibitively expensive if not done efficiently, and so the technology used to extract churn signals must be carefully designed and built with cost efficiencies in mind.
- Nuanced interpretation. It’s easy enough to build a social listening solution that traps brand, product, keyword and hashtag mentions, so many offer it. Fewer are able to offer a semantic interpretation of posts, and fewer still are capable of the nuanced interpretation necessary for effective churn prediction, such as discerning sentiment vs. intent, or past vs. future activity. The algorithms and machine learning processes required to do this represent the leading edge of innovation in predictive analytics.
At ThinkCX, we believe that social solutions providers will eventually move beyond reactive listening into the predictive realm, and deliver even greater value to their brand clients by serving up potential churn notifications (and upsell opportunities too – but that is a subject for a future post!) Our team of researchers and developers are pushing back the limits of attribution and semantic interpretation in order to give subscription-based service providers access to solutions that seamlessly and inexpensively pull social data into their churn prediction machines. After all, with as many as 60-70% of a brand’s customers active on public social channels, the volume and relevance of social media-driven churn data available to retention teams is just too significant to ignore any longer.