Rob Silas on Analyzing Unstructured Data
By Juliet Stott | November 15, 2016
Rob Silas, MSPC’s director of analytics, has worked in analytics for more than 25 years. In that time, data has moved from paper to CD-ROMs and now into the cloud. But the questions asked remain the same. “The challenges of ‘where do I start with this?’ or ‘what are the right things to measure?’ or ‘how can I drive the right insights from this data?’ are just as prevalent today as they were 25 years ago,” he says.
We spoke with Silas about what unstructured data is, how businesses can use it to inform their content and business strategies, and why it’s the next frontier to conquer.
How is data being used to drive and inform content?
Data types that can inform content can be classified in a variety of ways. One of the more popular ways is to look at it as structured or unstructured.
Structured data comes from data trails left by users who view or engage with content online. It’s quantifiable and measurable and can tell you how the content is performing or resonating with the audience.
Unstructured data comes primarily from user-generated content—i.e., content consumers are posting to their own social feeds, such as photos, videos and comments.
This qualitative data can tell you how consumers are behaving toward and responding to brands. Both types are important and help us develop a better understanding of who the customers/prospects are, as well as their behaviors and emotions, which can help inform and shape our content strategy going forward.
Can you explain what unstructured data is in more detail?
This data comes from published content, such as text, videos and photos. Social media and digital marketing have created an explosion in user-generated unstructured data over the past 10 years, but unstructured data itself is not a new phenomenon.
Think about answering an open-ended question in a survey—the written words are unstructured data. Similarly, in digital marketing, the text, photos and videos written in blogs, emails, websites and social networks are examples of unstructured data.
It’s valuable because it shows what’s being said—the words and phrases being used—which can tell us more about people’s attitudes, behaviors and emotions.
How can you analyze unstructured data?
As with any data analysis, a balance of people, process, and technology must exist in order to be effective and insightful. From a technology perspective, it’s best to use a tool that will go out to lots of different social and news sites and aggregate that content.
Many tools mine the available public data on Twitter, Instagram, blogs, forums and even Facebook. Most of them allow you to enter keywords that relate to your industry or business, and then show you conversations being had around those words.
You can not only see how many times a brand has been mentioned on social but also who’s being talked about, most popular topics discussed, behaviors, sentiment, emotions and intent. The gold in the analysis is not just what’s being said but who’s saying it—the influencers, advocates and detractors.
Which tools are best analyzing unstructured data?
Social listening tools have grown significantly over the last six or seven years. We use NetBase. It does a great job of helping us identify the right conversations to analyze, as it has built-in text analytics. Salesforce Social Studio (formerly Radian6), Sysomos, BrandWatch and Crimson Hexagon are other options.
What are the challenges of analyzing unstructured data?
We’re all facing the challenges of big data—80 percent of which is unstructured. “Signal and noise” is a phrase that exists in analytics: A lot of noise exists in data, but finding the true signal allows us to communicate insights and resulting actions that help drive business results.
This is as much an art as it is a science, and it’s where people and process become critical along with having the right tool(s). It’s about finding the right conversations to analyze and understanding which audiences are taking part in those discussions.
From a content perspective, it’s about how to use the information we’ve learned from those conversations to feed into the content and digital strategy, and continually measuring changes so business can both anticipate and react to key trends.
Can you give an example of how you can use unstructured data to inform content or business decisions?
When you think about the outcomes of analyzing unstructured data, consider two buckets:
- You can understand content performance—what content is resonating, who it’s resonating with, what channels are driving the best performance, and how performance compares to competitors and/or key partners. It’s also possible to look at performance across owned, earned and shared content; benchmark against competitors; and measure the impact of content performance against key business objectives. Armed with these measurements, content marketers can determine the topics, tonality, language, content types and channels that are performing best, and optimize their efforts based on these insights.
- You can also use the data to assist with content planning and strategy. The holistic approach combines looking at historical trends in social content that has resonated with customers/prospects with analyzing data from search-engine searches, email activity and website activity. This will show you how and when different customer segments engage with different content types, and provide key inputs for planning content calendar topics, channels and timing.
We’re seeing more brands go deeper with data to drive both content measurement and strategy in a variety of industries, including CPG, healthcare, technology and education.
Can you offer tips on how to glean insights from unstructured data?
First, understand that while the technology is getting better every day and we’re now seeing advanced text analytics and artificial intelligence being used to mine this data, you still need human review and analysis and a process in place to find valuable insights and performance metrics.
From a process perspective, first identify what kind of company you are and who your customers are. You’ll be looking for different things depending on whether you’re a B2B or B2C company. If you’re in a B2C industry, a consumer conversation is probably occurring about your brand right now. Listen and learn from it.
How can what’s being said (good or bad) help you improve your sales process or buyer journey? What are customers’/prospects’ likes and dislikes? What content is resonating, and what isn’t? Are influential authors writing about the brand or the industry, and who are they?
To get a true understanding of conversations, you must also identify the audiences that are publishing content. Are they consumers, influencers, “wannabe” influencers, media, etc.? Most companies overlook he audience identification step, but it’s critically important to prevent getting “false positives” from the data.
On the B2B side, people typically talk more about their industry and products rather than brands. Combining technology with a good process across multiple data sources can help you identify who the key influencers are in your space and what they’re saying. These are the people you want to target your content toward.
Who should lead the data exploration—marketers or analysts?
When an organization is firing on all cylinders, it’s absolutely both collaborating together. But what I see most often today is marketers who have an idea of what they want to use the data for and analysts who know the data but aren’t as familiar with the business use case as the marketers are.
So, without collaboration between the two, a gap is created—marketers can feel like the data isn’t actionable, while analysts can feel like the marketers don’t understand the value of the data.
The most successful companies understand that it’s an iterative process. Marketers must educate analysts, and analysts must share data findings as they are created; the ongoing collaboration creates value for both and for the business. It needs to be a two-way street