An Unstructured ApproachBy Samuel Greengard | Posted 02-01-2013
An Unstructured Approach
By Samuel Greengard
Getting closer to your customer is the mantra of every retailer. But when Gary King, CIO of Chico's, addresses the concept, he takes a highly unstructured view. Although the clothing chain diligently stores records of customer transactions in a conventional database and maintains profiles of customer preferences, it also taps into a growing stream of unstructured data, including social media, clickstream information and online behavior analysis. "We are achieving an entirely different view of our customers and using it to build better relationships and greater loyalty," King explains.
Make no mistake, Chico's is at the forefront of a rapidly evolving data ecosystem. Although structured data isn't going away, the role of unstructured data--and the way organizations combine different types of data sets--is fundamentally altering the face of IT and business. "Technology is providing new and powerful ways to use data and extract greater business value," says Nick Millman, digital, data and analytics lead for Accenture Europe, Africa and Latin America.
Of course, gaining deeper insight into patterns, trends and behavior represents a formidable challenge. There are business issues to address, technical challenges to solve, and a need to think about business problems and data in entirely different ways. In addition, enterprises must keep an eye on mobility, cloud computing, social media, big data and analytics trends so that it's possible to feed relevant information to software, systems and experts that are able to transform it into real-world results.
This data-driven era is just beginning to take shape. "Management's reflexes are built around navigating, managing and extracting value from structured data,” says Andy Rusnak, a partner in the finance and performance management practice at Ernst & Young. “Unstructured data represents an entirely different premise. It often comes from outside the four walls of the corporation. It comes from sources that you have little, if any, control over."
Beyond the Database
Although structured data has served the enterprise well--it has created remarkable opportunities to manage a mind-bending array of business requirements--rows and columns represent somewhat of a flat-earth approach to analytics. IBM estimates that about 80 percent of enterprise data now falls into the unstructured and semi-structured categories. This includes email, Microsoft Office files, text and IM streams, video, audio, photos, PDFs, social media exchanges, metadata, stored web pages, health records and much more.
In fact, industry estimates peg the growth of unstructured data at 10X to 50X annually, depending on the specific company and the industry. But it's not only the volume of data that CIOs struggle with, it's the complexity of combining and rearranging myriad elements in a way that produces tangible results. “The challenge isn't to capture, store and process all the data," Rusnak says. "It's to understand which pools of data to tap into and how to combine them in a way that helps the enterprise achieve a competitive edge."
An Unstructured Approach
The pool of unstructured data can seem infinite and overwhelming. A 2011 Unisphere Research/MarkLogic survey found that 86 percent of executives describe unstructured data as important to their enterprise but only 11 percent have clearly defined procedures and policies for handling it. In addition, 40 percent of the respondents admitted that they're not fully aware of the extent of unstructured data within their enterprise and only 45 percent say they're moderately or strongly committed to using it.
Rusnak says today's data environment requires an entirely different CIO mindset. "The primary focus cannot be on data scientists and statistical methods although both may play an important role. It's about the organization achieving competencies that aren't always clearly defined. It's necessary to find people who have a deep understanding of business processes and a high comfort level with manipulating data and combining it in new and unusual ways. They must understand how to draw conclusions and correlations from chaos and disorder."
It's often a journey into uncharted territory. For example, one agribusiness company now uses unstructured data collected through social media and combines it with a database of farmers to develop one-to-one marketing campaigns that focus on the specific needs of growers in a region. Meanwhile, telecos are turning to anonymized usage and location data to create new types of service packages and product offerings. And retailers are deploying next-generation analytics to understand spending patterns within demographic segments, when purchases occur and what triggers these transactions. "The right systems help organizations monetize their data," Millman says.
Chico's is a perfect example of a company that is diving headfirst into unstructured data and using it to transform the business. The retailer--which operates 1,350 stores within four major brands, including Chico's, White House Black Market, Soma and Boston Proper--is looking for ways to more effectively interact with its customers. The company has operated a loyalty program for years. Chico's acquires data about spending patterns and which promotions work more effectively with different customer segments. King says the company has a 90 percent success rate on connecting the person to the purchase.
But now Chico's is taking data collection and analysis to a new level. It is plugging in unstructured data from a number of sources--clickstream behavior at its site and on the web, customer-related social media posts, customer reviews, and sentiment data about its brands and products--in order to build a far more robust analytics model. The company relies on several SAS tools, including SAS Social Media Analysis, SAS Sentiment Analysis and SAS Text Analytics. "We consolidate all the data in a customer hub, conduct the analysis, and examine correlations between browsing behavior and purchasing behavior," King says.
By combining structured and unstructured data in unique ways, Chico's is blazing a path toward greater personalization of ads and promotions. It is also adapting its website to personalize products and offerings based on a combination of past buying habits, social media posts and browsing patterns. "By knowing how a customer reacts when we send an email or what type of web messages generates a visit to a store and leads to a purchase, we're able to engage in far more effective marketing," says King. Getting to this point, he says, has required "a new curiosity about information-driven decision-making" and a much more collaborative and free thinking process that cuts across organizational silos.
An Unstructured Approach
The data has also yielded some unexpected benefits. Increasingly, Chico's can gauge how a marketing strategy is working in real-time across channels and make tweaks and adjustments on the fly. The retailer is also better able to identify the type of person that's best suited to work as a sales associate in its stores and what type of dialog with customers maximizes the odds of a sale. Ultimately, "the technology is taking us closer to the customer,” says King. “It is transforming the business in ways that we couldn't have anticipated a few years ago."
Developing a data strategy and putting the systems, software and tools in place to execute it is critical. Accenture's Millman says there's a need to identify an organization's touch points and where it can most benefit from unstructured data. This requires conversations across departmental lines and an ability to see the business in 3D. It also means considering tools that fit today's analytics environment, including non-relational databases that fall into the NoSQL category and open-source distributed computing models like Apache Hadoop that can significantly ratchet up computing power.
It's also necessary to address an array of practical and technical issues, ranging from how best to analyze raw data and combine it with metadata to understanding where it can be plugged into an analytics strategy. Too often, Rusnak says, organizations fail to translate initiatives into real-world results because they lack an adequate master data management plan or there's simply too much complexity resulting from multiple data formats, cloud and SaaS environments, and distributed storage. In fact, about two-thirds of IT executives can't say where all their enterprise data is stored, according to industry research.
Finally, it's vital to cultivate the right thinking and skills within the business and in IT, Rusnak says. In many cases, CIOs and other executives must define new positions that defy traditional job titles and responsibilities. They also must work with CMOs, COOs and other senior executives to break down silos and connect the enterprise dots more effectively. And they must have a strong IT governance model in place to establish internal data ownership and sharing requirements. In some cases, these issues can extend to business partners and third-party providers, as well as data residing on systems outside the enterprise.
In the end, it's wise for CIOs to approach unstructured data as a treasure chest but one that can be difficult to unlock. Building the right infrastructure and pathways requires different thinking, plus entirely new skills. It requires unprecedented cooperation and next-generation tools and policies that put the data fully into play--and connect it to conventional databases and search engines. "Analytics is evolving and a new era of big data is emerging,” concludes Millman. “Unstructured data is a key piece of the puzzle."