The Mindset, Skillset, Dataset Approach to Social Media
A new approach called Mindset, Skillset, Dataset can help marketers make sense of complex social media data. Mindset refers to the ability to think beyond a certain frame of reference and look at the bigger ‘Why’ of solving a particular problem. Skillset refers to analytical techniques and tools that can be used to solve a particular problem. And Dataset refers to the copious amount of data generated from social media.
*This post was initially published in Adweek on Jan 29, 2015*
Social media is known to be an addictive medium and can be a performance deflator at work. But it also represents a big area of opportunity for marketers who want to target market based on social media data. The problem? Social media is comprised primarily of unstructured data, which is difficult to analyze.
A new approach called Mindset, Skillset, Dataset can help marketers make sense of complex social media data. Mindset refers to the ability to think beyond a certain frame of reference and look at the bigger ‘Why’ of solving a particular problem. Skillset refers to analytical techniques and tools that can be used to solve a particular problem. And Dataset refers to the copious amount of data generated from social media.
The availability of data and skillsets/technique is almost never an issue. Typically, it is the right mindset that’s lacking. Here are some tips for getting in the right frame of mind:
Mindset 1: Unearth Patterns – Look for deeply enmeshed patterns from an apparently innocuous dataset
Social media, due to its inherent nature of people and chatter, often contains deeply enmeshed underlying patterns that might not be immediately visible. These patterns can evolve to become treasure troves of predictive knowledge about a particular event. Google’s Flu Trends is a perfect example of identifying patterns and using them to predict the intensity of a particular event.
Flu Trends aggregates Google search queries on influenza for more than 25 countries based on user searches. As it turns out, when aggregated, individual searches can be a powerful indicator of flu occurrence in a given place.
From a skillset point of view, some of the analyses that help unearth these patterns are as follows:
Social networks analysis/mining
Tagging/links/graphs analysis and mining
Community detection and evolution
Influence, trust and privacy analysis
Social media monitoring/analysis
Mindset 2: Understand Opinions – Analyze opinions and categorize user sentiments
People used to share their secrets and sorrows with their best friends prior to the onset of social media. Now, social media listens to everything. This enormous amount of emotico-temporal data yields insights into how a particular set of users feel about a new/existing product. It can help validate and supplement the findings on problem areas for a specific product/brand done using surveys or other analytical models.
One of the leading South American personal healthcare products company adopted a new business model and wanted to understand market reaction to the change. With only a shoestring budget for analytics, they decided to study consumer sentiment on social media. It turned out to be a superior and accurate tactic, partly owing to the volume of data on the web from affected customers. The analysis helped the company understand that suppliers and middle agents were feeling very negative about this new model because it lowered margins for them, while end consumers were delighted due to the drop in overall prices.
From a skillset point of view, some of the analyses that can help understand sentiment are as follows:
Opinion extraction/classification/summarization/visualization
Temporal sentiment analysis
Cross-lingual/cross-domain sentiment analysis
Irony detection in opinion mining
Wish analysis
Product review analysis
Mindset 3: Establish Relationships – Define, measure relationships and potential connections across users
Establishing connections and similarities across user groups serves to solve two unique problems – one, how to find small, cohesive groups whose actions are of the same nature; and two, how to identify the underlying factors for these visibly unexplainable relationships. Unlike structured data analysis where similar people are associated based on a set of continuous/categorical variables, in the social media world, the data is unstructured and undefined. Establishing connections and causal linkages is an enormous challenge.
Facebook’s News Feed and Google’s Pagerank algorithms are stellar examples of creating these connections. Facebook makes friend suggestions based on a number of novel approaches including:
Social search algorithms
Social ranking
Multi-entity search
Multifaceted search
User modeling and personalization in social media
Mindset 4: Design Recommenders – Code and create socially intelligent systems
Leveraging social media and building predictive capabilities using this data enables teams to design socially aware intelligent systems that can perform a variety of tasks from understanding users’ emotional needs to delivering the right products. This capability arises solely out of the system’s power to understand each user’s state of mind, comparing compare it with a similar set of users who have undergone the same set of changes.
Social recommendation engines can strongly supplant traditional RFM (Recency, Frequency, Monetary) based propensity models, at least for a specific target demographic. These engines can help unearth the actual needs behind an emotional ask and also act as advisors by suggesting products that trusted friends or relatives have bought recently for the same purpose.
Some of the analyses that can help in this scenario are:
Social recommender systems
Semantic social media filtering
Market analysis
Cross-lingual/cross-domain social intelligent systems
Takeaways and Action Plan
One quick start idea would be to look at currently available traditional social media metrics and smartly redefine and reuse them to directionally answer the questions above. Once enough credibility is obtained in the sanctity of these metrics and definitions, tailor-made solutions can be deployed via one of the many statistical and data-mining-friendly languages like R/Python/Scala, which can enhance and enrich the entropy of information available through social media data.
A more long-term solution to this seemingly complicated set of exercises would be to have a dedicated social media listening center, manned by a small team of decision scientists. The team can start with a wide range of social media listening tools available in the market, which can later be tweaked, personalized and custom-made for an organizations’ demands – to eventually become a virtually intelligent social watch man. This man-machine interface will serve to be the singular platform for all kinds of social media extreme experimentation.
The important take away should be the fact that organizations should be careful about not drowning in the deluge of enormous social media data, but manage to stay afloat with the aid of new mindsets and novel skillsets. It is imperative to ask the right questions rather than design for wrong answers. Organizations must realize now that social media is no longer a light-sucking black hole; with enough dimensions at your disposal and a changed perspective – social media data is bursting to give us answers to questions we have not yet dared to ask.