Too Much Pork for Just One Fork

Article Featured Image
Innovative marketers questing to make marketing smarter need clean data. Without a heaping helping of text and numbers to satisfy campaign-choice hunger and customer initiatives, however, marketers' efforts are guesswork. In the so-called age of information, finding what is most relevant and usable for a marketing department is an intimidating task. With customer information everywhere online and data warehouses bursting at the seams, marketers must have the data and text mining tools to glean, manage, and understand marketing data. In the past few years data and text mining solutions have greatly expanded in functionality, and many companies are taking these tools to new levels. Structured data, which most often powers data mining, is being found in new places outside of marketing organizations. On the other hand, unstructured data--text mining's lifeblood--is being created daily by an onslaught of savvy online consumers. Together, the hard and soft data mutually make each other more relevant, providing the perfect mix for a marketing meal. Here, a look at how to get the freshest ingredients for a data mining banquet. Data mining software has been around for more than a decade, but only in the past few years has the technology gained widespread attention from marketing departments. When it was developed in the 1990s, according to Jill Dyche, an analyst at Baseline Consulting, data quality difficulties retarded data mining's acceptance. "I think seven or 10 years ago, when companies were first looking at this stuff, they were doing it against unsummarized data--not a lot of data, incomplete data, bad quality data--so their results weren't very good," Dyche says. Now that warehouses have matured into commoditized platforms, companies have many years' worth of customer data stored and ready to be leveraged. The number and type of channels from which data can be mined have also increased. With developments like speech-to-text technology in call centers and the explosion of e-commerce, companies can find, in a number of places, data relevant to marketing. Quantitative, structured data that has been accumulating in data warehouses is attainable through marketing services, but unstructured data has been accumulating by the bushel on the Web--from emails to phone calls to demographic studies and marketing surveys. "Five years ago there was a tremendous amount of effort just to have enough input," says Ron Halverson, president of marketing services company Halverson Group Inc. Today, he says, "The amount of information coming into most organizations is to a degree that's almost unmanageable for most companies." The potential value of this information has improved as well, most notably in text mining, the automated process of digging out specific information from a large body of text. A few years ago text mining was being used by companies primarily to scour competitive information online, to look at competitors' Web sites to get a better picture of the competition. Colin Shearer, senior vice president of market strategy at SPSS, says, "The big change we've seen in the last three years is this move toward using [text mining] for CRM, trying to look at the text data that describes the customer." As more customers voice opinions and place personal information about themselves on the Web (similar to what companies have done via corporate sites in the past), text mining efforts for CRM have become more possible. Food, Glorious Food Although data is more bountiful now than ever, information will not find you. It is important to know where to look for quality marketing data that is pertinent and fresh. For many companies it makes sense to begin the search for customer data at home. Marketers can find a great deal of valuable information by looking only as far as the data stored in other internal departments, most notably in customer service. Sid Banerjee, cofounder and CEO of text mining software company Clarabridge, says, "We're finding that most companies nowadays have created pretty sophisticated customer support organizations through email, through call centers, and now through interactive chat." The textual data stored in these documents can inform marketers on customer complaints, reactions to marketing messages, as well as thoughts and attitudes taken from both customer outreaches and company surveying. A further advantage of mining customer support documents is that as this information is constantly being collected, it is always up to date. Web 2.0 has presented another way for companies to mine unstructured, attitudinal data. When the blogosphere first emerged a few years ago, marketers began to troll its content to run a count of product or company name messages. Today, however, text mining vendors have started to build RSS feeds into their solutions to target the blogs and wikis to find rich analysis rather than a simple number. Shearer of SPSS, a company that has developed this functionality, says that when text mining is applied to the blogosphere, a company can find out not only if it is being talked about, but also what is being said. "You can measure how much is positive and how much is negative. This is really becoming quite hot compared to measurements taking the pulse of the blogosphere." Data from blogs, wikis, and various social networking sources helps companies get attitudinal feedback on their brands similar to what they would want from a focus group, but without the extra expense. Additionally, some believe Web 2.0 data to be less biased than surveys or focus groups. Banerjee says, "On the Internet you tend to get true opinion, but in a study people will often give the surveyor the opinion they think the surveyor wants." There is the possibility that consumers using the Web in this way do not accurately represent a sample set of your average customer. As these practices become more widespread, however, this disadvantage becomes less problematic. Although it is important to pay attention to emerging trends, such as employing customer and Web data, it is crucial to have quantitative, structured data in place first. Banerjee explains of structured and unstructured data, "You're definitely seeing the one plus one equals three here." When used together, the two data types will inform each other to give a fuller, more rounded view of your individual customers and your large demographic. No matter where the data is coming from, it is crucial to make sure that it is clean before analysis. As Dyche indicates, dirty data will make data mining efforts useless. Cleansing data--ensuring that customers are not stored multiple times and checking that all numbers and text information are accurate--is a daunting but necessary step in the mining process. Hewlett-Packard has been using SAS's enterprise mining solutions for almost nine years; however, the company still reports that one of the primary problems it runs into is dirty data. Randy Collica, business and data mining analyst at HP, says, "Old or inaccurate or missing data is always an issue when you're dealing with CRM data." He explains that although data improvement projects can be costly, they are crucial. "You have to prove [to the company] that investing in it is an important function," he says. HP uses SSA Software in conjunction with Siebel to promote data quality. Mine to Make it Yours
When fresh, reliable data has been secured, there arises the question of what to do with it. In measuring customer attitudes, demographics, and behavior, the end result of these calculations should be constant if the analysis is correctly performed. Customers are not affected (in most cases) by data mining efforts. Baseline's Dyche says that the change in data mining technology has come out of not what it does, but who can use it in which ways. She cites the most notable change: "[Users are] not just back in the dark, windowless room with statisticians anymore. They're emerging out to the business-user community." Data mining today is something that marketers can complete independently, no longer needing to rely on statisticians. Ease of use has become a new mantra for many data mining vendors. With user-friendly software marketers without high-level analyst skills can uncover trends and tips, such as who the most valuable customers are and which ones are most likely to churn. Marketers going ahead with solo data mining efforts are enabled through developments by traditional data mining vendors to make their products softer. Data mining capabilities are being embedded into other software packages, and more GUI-based products also provide smaller businesses analytics might to power their marketing decisions. SAS and SPSS have both developed more point-and-click, graphic-enabled user interfaces in the past few years for their data mining solutions. SPSS's Shearer says that companies "are going from having 10 to 100 analysts just working away on their models to actually automating it, making it more efficient." On the other end, other software companies that do not traditionally play in the data mining market have been baking this technology into their solutions. In February Microsoft released the new SQL Server 2005 SP2, an update to its data management and analysis program, which includes data mining add-ins for the 2007 Microsoft Office system. Dyche notes Microsoft's embracing of data mining as part of its SQL Server platform as a big step toward the wider proliferation of data mining technology in larger software offerings. She says of the release, "It's just a huge sea change for the rest of the vendors in the industry. That more than anything else is going to invite companies to look at data mining as part of their existing capability set." Automated, easy-to-use tools are making data mining for marketing quicker to use, but analysts and vendors assert that the new software does not replace the need for statisticians. Companies should try to employ both analysts and marketers in data mining efforts. Numbers and statistics can be delivered via data mining analysis solutions, but it still takes a high-level knowledge of the algorithms behind the software to fully understand how they should be interpreted. Teresa Jones, senior research analyst at the Butler Group, says, "If you put [statistics] in the hands of somebody who doesn't understand [them], it could risk expensive campaigns that don't deliver what you expect them to deliver because you're working off a false hypothesis." Jones says that one of the most common mistakes marketers make in reading analysis is judging the significance of a trend or casual relationship. Testing hypotheses and campaigns on a small scale before rolling them out can prevent possible marketing missteps. Act Like You Know Good data mining makes marketing smarter, but is pointless unless marketing acts on what it has learned. Shearer says, "You can keep on analyzing [data] till you're blue in the face, but until you take some different action with it, it's not going to give you any benefit." One key part of this plan is marketing's interaction with sales--marketing is still responsible for handing off customer information that helps sales to sell. Salespeople need to be aware of the trends that marketers are uncovering to know where and how to sell. Dyche says, "Usually, data mining knowledge-worker capabilities are outside the sales organization whereas they're inside of marketing." Because salespeople are customer-facing, it often falls on marketing to do much of the back-office research. As sales is often looking for the end result of that research, the explanation of how it translates into a lead and not the numbers themselves, marketers must hand off more than graphs, but foster a dialogue with sales as well. Marketing must also work with sales to understand how and why its campaigns and investments succeeded or failed and therefore inform its data and hypotheses. Marketers now must measure the effects of their efforts through response to campaigns, as well as the final return on investment after leads have been passed through sale's hands. By mining data from sales and customer service and then sharing findings back to these departments, data mining can weave in on itself to more closely integrate companies. Although data mining efforts may seem daunting, analysts agree that when done intelligently and carefully the payoff is guaranteed. The benefits include smarter marketing, more efficient targeting, and enterprisewide understanding of the customer. Michael Schiff, principal consultant for MAS Strategies, explains the bottom line: "If you're not using data or text mining today you need to look at it more closely, because your competitors probably are." Contact Editorial Assistant Jessica Sebor at jsebor@destinationCRM.com. Surveys Serve Up Service Success Cablecom, a leading Swiss cable network company, has proven that a fresh approach to data mining can mean serious euros. Although Cablecom had completed data mining products in the past through using standard models to predict customer behavior, the company continued to see high customer attrition rates. To help keep subscribers on board, Cablecom implemented SPSS's Dimension software to run specific online surveys for individuals at various points in their customer life cycles. The open-ended answers were then collected and mined with a text mining tool to target indications of dissatisfaction, potential upsells, and complaints. Cablecom has seen an improvement in churn rate from 14 percent to 3 or 4 percent from these efforts. The company is now looking to advance not only through online and email surveys, but also through text gathered from customer phone calls. Federico Cesconi, director of CRM at Cablecom, says that after the implementation, "what has changed has really been to combine this approach that comes from the classical school of data mining with that possibility to listen to the voice of your customer." --J.S. Click No More Dying by Inches to read the companion feature.
CRM Covers
for qualified subscribers
Subscribe Now Current Issue Past Issues

Related Articles

Text Links May Make the Best Marketing

Overall, 25 percent of online visitors will click a text link first; younger consumers, however, may be more graphically inclined.

¡Viva la Revolucion!

It's been a long and wasteful battle, but are we nearing the end of the war between marketing and sales?

SAS and SPSS Hold the Lead in Gartner Magic Quadrant on Data Mining

Magic Quadrant for Data Mining '08: Packaged applications are all the rage -- but as the pair atop the list remains the same, a Visionary emerges overseas.

Sales and Marketing: One Big Happy Family?

Maybe someday. In the meantime, Genius.com has unveiled a solution that strives to bridge the departmental divide.