Different social media monitoring platforms have different strengths and weaknesses. Having clear objectives and a clear development roadmap provides the basis for the most cost-effective investment in a platform which will support the desired level of maturity of use.
While the application of advanced data science capabilities is not widespread in use of social media monitoring platforms it is certainly gaining more attention as customers seek to optimise their investments in Influencer marketing.
When it comes to maturity in the interest and use of social media monitoring tools, customers typically fall into one of three broad categories. Maturity starts with (1) measuring activity, moves through (2) context analysis and then on to (3) Influencer identification.
Level 1 – Volumetrics
These customers are often at an early stage of monitoring and are interested in how many times they are getting mentioned in social media (or how often their competitors are mentioned). They will be interested in the most popular tweets about their brand / topics, the reach and how volumes vary over time.
Twitter Volumetrics Social Media Monitoring
The key technology applied to support this level of interest is the data acquisition capability of the social monitoring platform. One thing to watch for here is the quality of data acquisition and the range of sources. In particular not all social media monitoring tools have access to the full Twitter dataset because access to the full “Twitter Firehose” is expensive – it’s a key income stream for Twitter.
Free tools only get a subset of the Twitter stream often as low as a sample of 20% and with long delays. Conversely, fully paid access to the “Firehose” means that you get every single tweet in real time, and the top tier tools provide this level of access.
Customers at this level of maturity will also be interested in the basic identification of those people who are tweeting their brand / topics. This basic identification is generally at a global level as opposed to the data science-driven Influencer identification discussed below.
At this level of discussion the topic of “authority” emerges. Someone talking about the brand / topic is often identified by a ranking score relative to others and this is often deemed to be “authority”.
Typically an authority score is a measure between zero to 10 i.e. the “value” of twitter user’s network. An “authority score” is generally a combination of the number of followers and those following a user, the ratio of followers to following, tweet frequency, and possibly account longevity (as in Sysomos) – all normalised to a number between 1 and 10. Knowing this allows the identification of the most influential users talking about a brand / topic.
Authority scoring can also be applied to blogs with the most online influence for your brand or specified search terms. The entities identified as “most authoritative” are those leading conversations and experience high engagement with the content they distribute online.
Level 2 – Context
At this level of maturity customers are wanting to understand not just the volumetrics but also the content and context of the activity.
This requires effective algorithms for machines to make sense of the content data and to present it in various forms and visualisations, including sentiment analysis, language translation, short document summaries, keyword word clouds, visual buzzgraphs, popular phrases, semantic analysis with entities, and key conversations. The usual entry points to this level of usage are (1) sentiment analysis and (2) wordclouds and related text analysis. Trends and volumes of topic conversations are also of interest. The key technology applied to this component of social media monitoring is natural language processing – algorithmic text analysis.
The key presentations for context understanding are most frequently:
- Sentiment analysis;
- buzzgraphs (proprietary to Sysomos); and,
- “Entity Analysis”
The buzzgraph which displays the leading keywords and, as important, the relationships between them. Different words are connected by dashed, thin or bold lines to show weak or strong correlation between them (see buzzgraph above).
Entity Analysis goes beyond words and phrases to analyze the text semantically. By using natural language processing, conceptual entities are extracted and categorized as being a person, name, city, company, etc. You are no longer limited to the entire universe of words and phrases, but more meaningful entities categorized automatically by the system.
Below are the extracted entities from about the Playbook tablet (remember that?) and a zoomed in version of just Industry Terms.
Sysomos Entity Analysis MAP KINSHIP Digital
Sysomos NLP Entity Analysis Words Entities KINSHIP digitalHere, the text is analyzed semantically to categorize “New York Times” as one entity meaning a publication, whereas “New York” is another under cities.
This form of analysis seeks the most frequent whole phrases and entities such as people, brands, countries, sports and is useful in particular for understanding which topics are resonating in long form social media e.g. news, blogs, forums.
Typically you can also narrow the search results by geography, language, gender, source, date range and can drill down into conversations by jumping into advanced forms of text analytics, demographics and influencer biographies and follower analysis.
The combination of context analytics (level 2 maturity) with volumetrics (level 1 maturity) enables the association of brands / topics with sentiment and volumes of activity.
By discovering people who can drive awareness, affinity and scalability around topics customers can focus their resources on engaging with the most influential people and can achieve better ROI than with a more random approach.
This obviously feeds into building marketing and engagement messages and content which resonates with those with high authority and wide engagement potential.
Level 3 – Influencers
Customers at the highest level of maturity in social media monitoring are looking to very accurately identify relevant Influencers, rather than those more general Influencers typically identified at the pervious level. The key technology applied to this level of research is data analytics.
The objective is to be able to identify Influencers who are as specific as possible in a relation to specific topics (including brands). Social media monitoring platforms can readily identify the most authoritative accounts in a set of Twitter conversations (as above). The shortcoming is that these “most authoritative” are frequently on a global basis across a very broad range of topics and are not efficient targets as effective brand influencers.
Determining specific topic-influencers is only practically done for Twitter at the present time.
This is done by applying clustering and page-rank type algorithms to the set of accounts which embrace the topic conversation. That is, by looking at all the tweets containing relevant mentions, and then analysing the accounts from which those mentions came, data science can cluster the accounts into “communities”. Sysomos does this particularly well and this is a specific reason that we are a strong Sysomos advocates.
Cluster Analysis Communities Sysomos MAP
Community analysis helps you understand the Influencer dynamics in more detail. Rather than simply monitor for mentions (as at the Volumetrics level of maturity) you can work out which different groups of people are discussing the topic, who is influential within those groups, and the different viewpoints those groups are focused around.
Put another way, this type of algorithmic analysis of communities allows you to identify who the catalysts are inside your influencer population. Diving into the data set lets you find the specific individuals who are influencing the influencers.
In short you can analyse who is driving the conversation and the topics which are most relevant to which communities and is the beginning of investment analysis for influencer marketing. This is a skilled process, and this level of analysis is necessary in order for you to optimise your investment in an Influencer Marketing program.
This then allows for the development of very specific engagement plans which will resonate most with the key communities of interest, and also be as unobtrusive and “organic” as possible. Customers working at this level are mature social media monitoring users.