Friday, 25 March 2011
Improving social media monitoring…
As Social Media Monitoring travels through the technology hype cycle it’s clear that many early adopters are experiencing disappointing results. Having heard the stories of start-up businesses like Threadless or GiffGaff harnessing social insight to create real, tangible value (see my post on GiffGaff – a case study of customers in control), many large organizations are often disappointed at the results they are getting from SMM. If you find yourself in the “trough of disillusionment, below are some of the common pitfalls along with some suggestions to improve results and drive value from your investment.
A common complaint about SMM is accuracy. Many SMM tools do not have access to the full Twitter fire hose and so only get a fraction of the full Twitter feed. Most tools also struggle to automate the classification of content (e.g. “positive, negative, neutral, mixed”) – accuracy rates can be as low as 20-30%.
To supplement this lack of accuracy, many organizations (or the agencies they employ) turn to people to manually read through and classify content. Whilst this might improve accuracy, this is hardly a scalable approach. Dell, for example has around 22,000 mentions per day - a figure that is rapidly rising. Add competitor mentions to that number and there becomes a danger of simply setting up another monolithic contact center to trawl through vast volumes of data and attempt to respond 1:1 to every mention.
So how do you address some of the issues above and start to drive some value from SMM? Firstly, it’s important to accept that you will never get 100% accuracy. The 80:20 rule applies. The SMM tool you purchased is likely to be only a small part of the solution. Human intervention is almost certainly required to “train” the tool – e.g. refining search terms to filter out the noise, interpreting slang and sarcasm etc. In addition, text analytics / natural language processing can be used to improve automation and accuracy levels. In fact Social Media has provided a shot in the arm for the Text Analytics industry and given a relatively old technology a new less of life.
But improving the accuracy of your social media monitoring and classification is not enough. Fundamentally, social media monitoring should not be looked at in a silo. If you are going to use text analytics on social verbatim then why not also use it on other unstructured verbatim sources such as surveys, e-mails, call centre notes etc? Furthermore, when you then start to gain insights from these unstructured data sources it's vital to combine with structured data to analyse and understand the real impact on the business. A spike in social media complaints for example could either be an early warning signal or it might be indicative of an issue that is already causing repeat calls, product returns etc. Again human intervention is required to spot these trends and interrogate your business information sources effectively. The speed at which social media storms or viral campaigns can travel means that humans must spot trends quickly and have flexibility to interrogate multiple and disparate data sources. Rather than thinking of a reactive 1:1 contact centre it might be more helpful to think of an air traffic control room with big screens visually displaying and blending huge volumes of data to allow controllers to make smart decisions.
Finally, make sure you turn insight into action. If customers are Tweeting and blogging about your products and services they are usually doing so for a reason. The reasons can represent opportunities to improve your products or services, protect your brand reputation or create new business opportunities. Either way, it's critical to get to the root cause and address whatever it is that's causing the spike. In other words, to extend the analogy above, you must link your air traffic controllers with the people on the front line, empowering them to make decisions and facilitating the necessary integration into the rest of the organisation e.g. Customer service, product development, sales, legal etc.
One large organization that has achieved success with SMM is Dell. Their Social Media command centre blends technology with human intervention. I won't repeat the case study which has been nicely written up by Matt Wilson here but I will pull out one quote which I liked from Adam Brown, executive director of social media for Dell. Describing his vision for Dell’s Social Media hub he said:
“The goal of the command center is to help decide how the SOS team’s time and energy can best be spent. Help us understand which of these are relevant, which of these are shown as trending information, which of these are things we really, really need to respond to. There’s no way we can respond to 22,000 conversations a day.”
Adam Brown nails the key issues with Social Media Monitoring here – accuracy, relevancy and focus. It will be interesting to monitor the success of Dell’s approach but they seem to be building with scale and integration in mind rather than trying to reactively fire-fight every social mention in a silo.