What to consider starting with social listening projects
Social Listening, often referred to as Social Engagement, has matured from its beginnings several years ago as a new marketing tool. Originally conceived to help identify trends more quickly and avoid possible screw-ups, it is now also used to identify influencers and target groups. Furthermore, it is an essential component of marketing automation, customer communication and product development.
The initial hopes placed in this new approach were high, so disappointments were not long in coming. The reasons for this were many and varied: immature software, misconception regarding its use, insufficient integration into business processes and a lack of integration into the existing IT system.
Social Listening tools have become more sophisticated and user-friendly thanks to technical progress and continuous customer pressure. The connection of more and more new social networks has progressed thanks to a better API, and the processing of large amounts of data has become easier and faster. In addition, the steep learning curve at publishers has led to Social Listening no longer being a sales channel among many. It has become the focus of sales and marketing activities. The business processes are now better coordinated – not only with regard to social listening, but across all departments and tasks.
In short: the development to date is promising.
However, two problems have so far proved to be relatively persistent. First, the known problem of inadequate integration into existing IT, and secondly, that lesser-known problem of qualitative analysis.
Both problems can be solved.
Social Listening not only consists of data coming from social networks, but also has to be connected with information from many other data pools. CRM must not only provide campaign data, but also deliver customer history and information. Sales, contribution margins and other cost information must flow from the commercial software, and product information and metadata are needed. The more fragmented these data pools are, the more complex it is to prepare the data and exchange it between the different applications. Publishers with a purely best-of-breed approach therefore find it more difficult than those with more homogeneous IT landscapes, as the complexity of their IT is usually greater. Nowadays, however, a publisher can hardly avoid the homogenization of IT if it wants to keep administrative and organizational costs within limits.
The second problem with social listening is a tougher nut. Quantitative analyses, i. e. counting of likes, shares, re-tweets, comments etc., are relatively simple. On the other hand, it is highly problematic to obtain qualitative information from social networks. This is due to the fact that considerable know-how is required with regard to the operation of these tools and, even more so, expert knowledge about the respective sub-market, target group and product is required. The combination of technical expertise, market and product knowledge and behavioural economics is rarely found in one person. To date, companies have tended to have young, technology-oriented employees using social listening tools rather than involving colleagues with a broader horizon of experience. Their experience is decisive in setting up the analysis parameters and analysing the results.
This starts with the generation of search terms, keywords or tags that are used by the software to browse social networks. If the term or combination of terms is too broad, the hit rate increases considerably, but contains a lot of irrelevant results. If it is too specific, the result set is too small for good analysis. This is where the success or failure of Social Listening is decided.
Another aspect to consider concerns two widespread phenomena in social media: irony and sarcasm are widespread and difficult to identify from algorithms. This distorts the analysis considerably. But it is precisely the qualitative analysis that makes social engagement tools so attractive. The problem remains to be solved.
The most promising approach to this is the use of machine learning, i. e. self-learning algorithms, which are becoming increasingly effective in detecting irony and sarcasm through targeted training and feedback. More about this in another part of this series.
Spotted any mistakes, misconceptions or misunderstandings? Please leave me a note in the comments or contact me via Skype (cck. knk), mail (firstname.lastname@example.org) or mobile (+33 7 86 51 62 37).