Editor’s note: Eric Pendleton is project training manager at at Luminoso, a Cambridge, Mass., text analytics firm.
How big has the “big” in big data really become? I decided to find out during a quiet moment the other day: 2.5 quintillion bytes of data – 90 percent of which has been created in the last two years (according to IBM). Of that, a further 90 percent is estimated to be unstructured, i.e., text-based and in the form of open-ended survey responses, contact-center transcripts and social media content. That’s a heck of a lot of Facebook posts.
While companies are still becoming aware of the massive amounts of unstructured data they have, they are not necessarily aware of the risks they face or the untapped value they’re missing out on. But experience shows that they should make this a priority. For the average Fortune 1000 company, increasing data accessibility by just 10 percent results in more than $65 million in additional net income.
So, how do you make the most of all those transcripts, reviews, tweets and open-ended survey responses? Having worked to implement or improve unstructured data analytics programs, I’ve come up with five actions you can take to unlock the value in your big data.
1. Determine your business objectives. What is the primary goal you want to accomplish or problem you want to solve? What information do you need to move forward? What data will help you get that information? Your answers to those questions will impact which data you should focus your time on as well as how you’ll approach your analysis. It is important to determine your business objectives before diving into an analysis.
2. Remove data silos. Data is useless if it's inaccessible. At too many companies, data is collected and managed by different departments – leading to situations where the data exists within the organization but cannot be pulled and used across departments. Removing your data silos requires lowering both technological and cultural barriers to sharing information. As the Aberdeen Group states, data silos are “often caused by a failure of technology, but [they] can also be caused by database administrators’ detrimental sense of proprietorship.” Companies must adjust the way data is stored and formalize the process of who grants permission to access data so that analysts can pull what they need, regardless of who collected it or where it’s stored.
3. Empower the data analysts. Companies need to improve operational data practices by making it easier to access unstructured data from their information management systems and by giving data analysts a broader mandate to pull and use the data. Companies no longer have the luxury to wait weeks to pull and analyze data and gauge business results; decisions must be made more quickly in today’s fast-paced markets. Traditional business intelligence is too slow for short attention spans and fickle trends.
The question of how to acquire the data from different sources in the first place is one that needs to be addressed at the highest levels within any organization. Data analysts alone cannot perform the necessary operations without a commitment from the top to address ownership and compliance issues.
Start by building a business case for analytics. Whether you use a past success story or a hypothetical (but realistic) example, it’s important to showcase how analyzing your company’s data is tied to clear business benefits. Outline a challenge your company is facing, the questions that need to be answered to overcome that challenge and how leveraging your data can resolve those questions and help your company move forward. By showing that data analytics has a business impact, you’re much more likely to gain the support of one of your executives – who can then advocate for appropriate resources and data access to empower your analysts.
4. Analyze unstructured and structured data together. Companies that combine unstructured, text-based data (open-ended feedback, reviews) with structured, numbers-based data (demographic information, consumer purchase histories) and analyze them together can gain significantly deeper and more actionable insights than looking at either data type alone. While structured data is often easier to process and analyze, it can only reveal overall trends – not the reason behind those changes. Unstructured data can reveal a deep understanding of the why behind the data, it’s just more difficult to track and may be dismissed by skeptical executives who reason that “it’s just what a few people say.” Illuminating structured data (the what) with unstructured data (the how and why) is a sound tactic to gain an understanding of your customers and create long-term loyalty.
The first step to doing this is the simplest one: make sure you’re collecting and analyzing your unstructured data. Because unstructured data can be a bit more unwieldy, too many companies simply ignore it in favor of their structured data, missing out on key insights. One challenge is that many companies simply have too much unstructured data to efficiently analyze using traditional approaches. Artificial intelligence and natural language processing have made it feasible to automatically categorize and track unstructured data. This quantifies your unstructured data, enabling you to understand the why behind your data and track if root drivers are becoming more or less prevalent.
5. Measure and track the ROI of your analytics program. There is a scarcity of literature about the ROI of analytics tools. However, according to Boston-based Nucleus Research, in 2014 every dollar invested in analytics returned $13.01 (an increase from just $10.66 in 2011). Tracking ROI can be challenging but it is key to proving the value of data analytics and gaining support from others in your organization (especially at the executive level). As discussed earlier, identify a business challenge or objective that your company dealt with and how data was used to solve that problem. What knowledge was gained because of your analysis? How might the outcome have been different had that knowledge not been discovered? While quantifying such what-if scenarios requires some estimation, it’s important to do so to prove your value. Keep in mind that ROI isn’t always measured by revenue or profit; it could be an increase in customer/employee satisfaction, a reduction in attrition or some other measure.
The future of unstructured data
The future belongs to unstructured data and the valuable business insights it contains. Companies need to evolve and update their business intelligence processes to include unstructured data and unlock its value. By implementing the five steps above and developing an appropriate business strategy – combined with the right data practices and analytics tools – you can discover how to tackle your current business challenges and chart your company’s future.