Social Care Source Analysis
By Edwin Margulies
Social Engagement for Customer Care starts with identifying viable social sources for content. This means organizing content from Twitter, Facebook, LinkedIn, articles and Blogs. Source Analysis is the discipline of understanding and adapting to relevancy trends in each source, and doing it well means an uptick in productivity for social care professionals.
What Sources are Meaningful?
Social Networking channels are burgeoning with data. But to filter that data into meaningful and actionable information is the hard part. Many social engagement tools have been developed to transform this raw data into actionable information. Take, for example, Natural Language Processing / Understanding (NLP/NLU) engines, and keyword search engines.
Depending on the charter of your customer care group, you may focus more on Tweets than blogs. Blogs offer rich data and detailed descriptions of news items, trends and opinion. But blogs do not always contain actionable information from a customer service view. Tweets on the other hand, are by nature very succinct, and therefore get right to the point when it comes to a cry for help or a complaint about your product or service.
You can also argue that tweets can carry a lot of noise. Especially re-tweets. Re-tweets are kind of like echoes in the forest after an "original bird" makes a statement. In terms of action, it makes no sense to drill down into a re-tweet because it is usually only the sentiment of the original author that you can take action on. Clearly, it is interesting to see trends in re-tweets if you are doing brand and trend analysis, but for a customer care agent, re-tweets are just spam.
Making sense out of Sources
First, get familiar with your social analytics tools. Great social analytics tools allow you to not only choose intervals, but also source, sentiment, relevancy scores and outreach progress. If you don't have a good analytics package, consider using an all-in-one social engagement platform that rolls analytics, NLP, and outreach together.
Second, use automated clustering to characterize the persistent business issues and trending topics associated with each source. For example, your analysis may show that 80% of the blogs your search criteria is tuned-in to are news-related or contain technical review information. Further, your analysis may show that tweets on the other hand are mostly service or complaint-related. There are no hard and fast rules. Each enterprise has its own way of profiling sources, so you will have to establish a methodology for profiling based on initial and ongoing analysis of clusters.
Third, focus on sentiment. It's a good idea to see which sources are running "hot" or "cold" sentiment-wise. You can characterize sentiment easily by putting posts in three bins: Happy, Neutral, and Not Happy. Use your analytics package to display each source and the break-down of Happy, Neutral and Not Happy posts. This will give you insights into prioritization. For example, if you discover that tweets in a certain cluster are running "hot" but the same cluster in blogs is "cold" or neutral, you can use your rules engine to prioritize tweets for "first outreach." This is a great way to mobilize your social care team around the most important authors.
Lastly, you can prioritize and filter sources based on influence. A common approach here is to use your analytics package to list the top authors by the number of posts and their influence. But be careful, just because a tweet is authored by someone influential does not make it a candidate for action. Influence is an attribute that helps you to focus and personality and also to anticipate the way an author may respond to a general announcement or even a generic outreach meant for a broader audience.
It's a good idea to follow influencers and try to have direct message conversations with them. This establishes trust and in turn, any open communications you may post on a certain subject are less likely to be put down in public by that author. Nothing beats sincerity though, so you must be sure to personalize and carefully craft responses to influencers because your brand's credibility is on the line.
Don't forget the Outliers
There are always exceptions. So the danger in source profiling is the fact that there are outliers in each category that are worth exploring. A business analyst or "social quarterback" can do a great service to the rest of your social engagement team by periodically reviewing outliers and cherry-picking or re-routing posts that are actionable.
For example, let's say you've set up your filters and rules engine to put all neutral sentiment blogs on the "news" cluster in a "to be reviewed" bin. Here, your social quarterback or analyst can use influence and other attributes to eyeball posts for content that may have fallen through the cracks in your first pass filtering.
This quarterbacking function is essential for two reasons: 1) It saves the bulk of your social care team from eyeballing each post and taking away from outreach activities; and 2) your analyst or social quarterback can make course corrections by re-tagging clusters, sentiment and relevancy of outlier posts - thus providing "training data" back to the NLP engine. This practice of re-characterizing some outliers is therefore a great way to save your overall team time and to fine-tune the NLP at the same time. This ensures your automation efforts perfect over time.
You can establish meaningful approaches to Source Analysis for your social engagement for customer care efforts in a few easy steps. By developing a methodology for source profiling, and doing standard analysis on your sources, you can quickly put together strategies for each of your social sources. The use of an analyst or social quarterback role can further enhance your outreach performance.