(I wrote this article as a contributed guest piece for DataSift’s blog. DataSift didn’t pay me to write it, but they did sponsor my October 30, 2012 Sentiment Analysis Symposium.)
Given: Your social listening, engagement, marketing, and customer support initiatives — and beyond them, a range of emerging, social-infused business processes (that is, every process that involves people) — require clean, complete, well-structured, and reliable social data. If you’re already a DataSift customer, you surely agree, but if you’re not yet convinced, poke around the DataSift site a bit — or contact me and let me convince you — and I’m sure you will come around.
Yet good data is only a starting point. Keep your eye on the finish line and on the path from start to finish.
The finish line is crazy-great products and services, happy customers, and, yes, profit, however measured. The path is paved with opportunities. Good data enables you to recognize and seize them and also to manage risk. The challenge is in the recognition, action, mitigation, and measurement. For those functions, and to understand and prepare for inevitable changes in demand and market conditions, you need to see beyond the numbers. You need insight and foresight, which can be obtained systematically and reliably only via appropriate and timely analytics.
Given the abundance of useful data available, the aim is to do more with more. Analytics and semantics will guide you.
Analytics is a set of algorithms, software, and practices that transform data to produce insights that support decision-making. Don’t be fooled by the e-metrics types who would narrowly define “analytics” as counting site visits, referrers, social connections, and retweets. Appropriate analytics — for the social-online era and its “unstructured,” geospatial, network, and time-based data — extends to handling of information extracted from text and other content. That extracted information may include names (people, places, companies, brands), events, and relationships.
And foresight: That comes from predictive modeling that fuses data from disparate sources, of disparate types. You need to understand how all that data is interrelated.
Semantic computing is about data relationships — about “this is a that” — and about attributes. (The common definition, that “semantics” means “meaning,” is hopelessly shallow.) Examples help: “Joe Biden is vice president of the United States” captures multiple relationships, just as the statement that “India is a country… in Asia… of over 1.2 billion people” classifies India as a county, hierarchically within a continent, with a population-size attribute, and just as “I prefer Android phones to the iPhone” conveys relationships, namely my relative preferences.
Remember that we’re dealing with social sources, replete with opinions, attitudes, and emotions, namely sentiment. Statements like the Android-iPhone one above are a social-network norm although they are typically expressed ungrammatically, with slang, abbreviations, misspellings, and sarcasm and in dozens of human languages, distributed across thousands of social and online platforms. Meaning-finding, via analytics and semantics, can get complicated, especially if you don’t have the right set of tools and processes.
And here’s the pitch: You need automated analyses to make effective use of sentiment. One way to learn how is to join me and DataSift at my up-coming Sentiment Analysis Symposium, October 30 in San Francisco. DataSift is a sponsor.
You might also take advantage of the optional, exceptional learning opportunities that precede the symposium the afternoon of October 29, a Practical Sentiment Analysis tutorial, developed and taught by Dr. Diana Maynard of the Univ. of Sheffield, and a new Introduction to Social Media Listening class, taught by former IBM Distinguished Engineer Mike Moran. Learn how to do more with more.
But there are many ways to learn. One is simply by doing. You know your own business better than anyone else. Trace back a sampling of focused business goals, from decision-support insights to source data, via a set of analyses and visualizations. Definitely do look to the many great solution providers out there, for consulting and tool help, if you need it.
What’s an example? Maybe you want to study the comparative effectiveness, measured by both social/media response and sales increase, of a set of trial, limited-duration, regional marketing campaigns. Arrange a data feed from the online and social platforms where people are posting campaign reactions or general product comments. Set up filters to eliminate data that’s not immediately pertinent. Find and configure analysis tools to study posting and message-diffusion patterns and to extract sentiment related to the products/services, classified by location and demographics. Look for trends, shifts, and outliers across the campaign — often change will provide the greatest insights — and investigate root causes by drilling through to individual postings and aggregate opinions. Link to related data sources, to customer profiles, transaction records, general market data, and so on. Implement changes and measure their impact relative to predictions.
I have sketched a process that will take you from social data to business insight. There are many variants that suit the range of business conditions. The data is available. It’s up to you to learn and gain the rewards.