The Boulder BI Brain Trust


Recently in social ROI Category

Analyzing Social Analytics

| | TrackBacks (0)

Social (media, technologies, applications) continues to penetrate the corporate world.  To date the penetration has primarily been driven by the demand to take advantage of Facebook’s massive reach and enable companies to get closer to their customers.  More recently, companies started taking similar advantage of Twitter, LinkedIn and Google+.  The quick adoption of social by companies of every size, including the enterprise, has led to the rapid success of companies like Buddy Media, Vitrue, and Wildflower, whose tools and services enable companies to establish their social presence, primarily on Facebook.

In the course of investing in social application companies over the past 3 years, we have considered several different uses of these technologies in business processes, from product development to marketing to customer service and job-applicant tracking.  In a report published by the McKinsey Global Institute, the authors identify 10 different ways social technologies can add value to enterprise functions.  Along with this broader usage, companies are beginning to ask more pointed questions about the value of social, and the quantifiable benefits they receive from the use of such media, technologies and applications, as well as for actions that will enable them to increase this value.  Analytics can play an important role in answering these questions and be used to provide insights and the corresponding actions, i.e., Insight as a Service, to increase the value of social.  While the initial use of social analytics has been by the CMO, such analytics can also be used by the HR, customer support and other executives in the corporate suite.

Each use of social generates and leverages different types of data, from simple structured data about the characteristics of a brand’s fan base, to unstructured (and often messy and noisy) postings in activity streams, to more complex data captured in a social graph including Facebooks Open Graph.  Some of this data (e.g., tweets) is very fast with a short “shelf life,” while other data is more enduring, such as a brand’s social graph.  All this is indeed big data and must be approached as such. 

Collecting, organizing and preparing this data for analysis can also be challenging.  In surveys conducted in 2011 by Gleanster, 89% of the surveyed executives found the tracking and measurement of results from social campaigns to be very challenging, while 94% found it difficult to generate actionable insights from the social campaign data they collect, which suggests that they may not even be collecting the right data.  Some of this data can be useful even without being rigorously analyzed.  For example, in a recently published white paper, Bazaarvoice states social data alone is starting to impact CMOs’ decisions.  It is also becoming evident to these executives that social data needs to be integrated with other corporate and third-party data in order to create the metrics and analytics to determine ROI.  This points to the realization that while social data, including social graph data, is a new and less tested data type, early adopters of social are still able to recognize that this data by itself will not be sufficient for defining the value-analysis metrics and KPIs.

We are already starting to see that the analysis of social data, when done correctly, can yield significant results, allow business users to determine “what moves the needle” and establish the ROI of using social media and applications.  But to determine which of the collected social data needs to be analyzed, one must first define what will constitute success in a particular process where social is used.  For example, if social is used in marketing, will success be an increase in customer loyalty, higher brand engagement, or sales growth?  What about in customer support?  Will success be the optimization of the customer experience, or the reduction of the customer support costs?  Once the objective is established, then one can define the metrics and KPIs that will be appropriate for determining its success.  For example, to establish social ROI around customer loyalty, a CMO may want to understand whether prospects that are acquired through social media have a higher probability of becoming customers (i.e., convert) than prospects acquired through an email campaign, as well as understand the individual characteristics of the prospects that actually converted.   Or the CMO may be interested in determining whether a particular social channel, such as Facebook, is leading to more conversions/sales than another such channel, such as YouTube.  In its report on social analytics, Altimeter explains the various business objectives that can be achieved through the use of social and provides different metrics for understanding whether the objective has been successfully achieved.  Obviously each metric is driven by different data.  It is the analysis of these metrics and KPIs that provides the insights and actions which enable the determination of ROI.

Social Analytics for Marketing

While the insights and actions derived from the analysis of social data can produce significant ROI in a variety of business functions and processes, early attempts to develop social analytics have centered around the marketing department’s needs because CMOs have been some of the earliest adopters of social media for branding and customer acquisition.  But while CMOs have seen their brands successfully acquire many fans, they have had trouble determining which of these fans to try to convert to customers, and establishing the lifetime value of each such converted customer.  We have seen three types of social analytics for the CMO: sentiment analysisattribution analysis, and social graph analysis

Sentiment analysis refers to the measurement of a consumer’s attitude towards a brand and its competitors.  The attitude may be towards the overall brand or towards a particular product or service offered by the brand.  Sentiment analysis can be used to identify certain trends (e.g., a brand’s loyalty), but more importantly, it can be used to make predictions, such as predicting the broad sales performance of a product (and if combined with location-based and channel data, it can predict the performance around a particular geographic location, or predict the sales performance of the social channel in comparison to the sales performance of other channels, such as email).  This type of analysis may be based on simple keyword-counting or on full-blown Natural Language Processing (NLP) of the social streams and other social content such as blogs and wikis.  There are already tens of, rather undifferentiated, companies using keyword-counting to provide sentiment analysis solutions and a few that use more sophisticated NLP approaches. 

Attribution analysis is the process of assessing the effectiveness of advertising campaigns by channel and each channel’s contribution to sales.  As social has become an important channel, marketers have started using it for both direct response and brand advertising campaigns.  Through attribution analytics the marketer is trying to determine what percentage of each social ad’s audience has been exposed to the ad, how many times each ad was seen, whether the use of social media “drove” prospects to properties owned by the brand, (e.g., the brand’s web site, or a store), and whether social media is more effective than paid advertising.  Given the dynamic nature of social media and the very large number of variables involved with each campaign, isolating the elements that made a campaign successful might be tough, and thus attribution is very difficult.

Social graph analysis refers to the examination of the graph’s structure to reach certain conclusions.  A graph’s structure consists of the data associated with each node and the connections/links among each two nodes.  In a social graph a node may represent an individual, (e.g., Evangelos Simoudis, his place of residence, his birthday, etc.) or an organization (e.g., the Coca-Cola Company, headquartered in Atlanta).  Connections may represent certain relations about the nodes (e.g., Evangelos Simoudis is a fan of Coca-Cola, and John Doe is a friend of Evangelos Simoudis and reads his blog, which he publishes through a particular URL).  This is a more recent form of social analytics and it is being made possible by the API-based accessibility of the various social graphs, such as Facebook’s Open Graph.  Graph analytics are being used to identify a brand’s advocates and establish which of these advocates are realinfluencers.  For example, one way of using the Open Graph to establish whether a fan is a brand’s advocate, is by determining if the fan is publishing sponsored stories about the brand.  By subsequently analyzing how these stories propagate along the social graph of each social channel (e.g., a sponsored story published in Facebook may be re-tweeted, thus moving from one social channel, and one social graph, to another) among the advocate’s social sphere, and the social sphere of each of his friends, a brand can determine which of these advocates are influencers.  Finally, by analyzing the characteristics of these influencers (e.g., determine each influencer’s lifetime value), marketers can treat them differently than other advocates and fans.  For example, the brand may decide to extend them special discounts to buy a product or service.  Three of my own portfolio companies (Extole8thbridge andThisMoment) have developed extensive social graph analytics that are now being used by their customers.  Extole and 8thbridge analyze Facebook’s Open Graph, whereas ThisMoment analyzes Google’s social graph.  A very good example of Extole social graph analytics can be found here.  According to eConsultancy research, 88% of surveyed companies indicated that social graph personalization, which is the result of social graph analytics, generates results.


  1. As the use of social tools expands in a variety of business processes, there is increasing need to analyze the generated big data to extract insights and drive the right action(s), which improve the performance of each such process.
  2. Social data is the key ingredient in social analytics, but often it must be combined with other data to produce the necessary high-quality results, very much like in any other area where analytics is being used.
  3. Though it is considered the most widely used form of social analytics, sentiment analysis is not the only type of analysis to be applied to social data.  Attribution analysis and graph analysis are important forms of social analytics.
  4. Marketing departments in general and the CMO in particular have been early adopters of social analytics.  By using social analytics they are starting to demonstrate the benefit and associated ROI of applying social media, technologies and applications to a diverse set of tasks, including advertising effectiveness, and brand engagement.


This page is a archive of recent entries in the social ROI category.

Social Media is the previous category.

Vendor Briefing is the next category.

Find recent content on the main index or look in the archives to find all content.

social ROI: Monthly Archives