Planning for ROI in Text Analytics
Want good ROI from text analytics? You’ve got to plan for it! In any field, there are always two major ways to get returns on investment – increase revenue or decrease costs. The key to good returns is shortening the path from the analysis to the payoff. That takes planning.
Many text analytics offerings focus on “insight.” Everybody wants insight, of course. Nobody wants to feel clueless about the mechanisms of a business or what’s on the customer’s mind. But it may be a long and complex path from insight to cash. In Text/Content Analytics 2011: User Perspectives on Solutions and Providers, a text analytics market study by Altaplana’s Seth Grimes, respondents were asked about returns on their investments in text analytics. The response wasn’t pretty – most did not reach positive return on investment. My take on that – the heart of the problem was lack of planning.
Rob Cooley, CTO of Optimine, a keyword bid optimization firm, speaking at a recent panel discussion, made a simple observation – the most successful analytics programs begin with the end in mind, then work backwards to create a plan for achieving that end. If you want to see good returns for your analytics investment, you should listen to Rob!
Let’s think that through. The first thing you must have is an end point that either brings in revenue, or reduces costs. For an easy start – look for places where you are spending money right now, and might reduce costs. For example, do you collect open ended responses on surveys? Do you use a service for coding those open ended responses? Many clients report that these services are expensive, slow and provide inconsistent results. If that’s your story, then you have a fine opportunity to benefit from text analytics.
Start with the desired results – good, consistent coding of open-ended survey responses, delivered quickly. You’re affected by several kinds of costs: the price of your coding service, opportunity cost of waiting for the results of the survey, and cost of poor quality in the work. Estimate those costs, and you’ll know something about how much alternatives are worth to you. As you search for a solution, you’ll be able to make rational choices about the value of those solutions and potential returns.
What about the revenue side of the equation? Would you like to retain more customers? What’s it worth to you to keep a customer? How could you identify customers about to leave? Han-Sheong Lai, Director of Consulting, Operational Excellence & Customer Advocacy at Paypal, uses a novel text analytics application to review customer comments and identify clear indicators of intention to close accounts. Armed with lists of customers ready to fly the coop, Paypal can take action to work with those customers and retain as many of them as possible.
A clear understanding of your own goals helps you avoid investments that won’t provide good returns. Quantifying potential returns helps make the case for the expenditures necessary to meet those goals. Defining and prioritizing requirements helps you to select the tools and processes that are most appropriate for your needs.
The common-sense practice of defining a goal, then making plans to back it up, steers you clear of unproductive activity and expense. At the same time, it may open new doors. What new things would you try if you had a clear understanding of the benefits?
Sales bring in revenue, so what’s something that leads to sales? Ad clicks are one of many possibilities. What makes people click on ads? Relevant products, compelling copy, appealing images… but what do any of those things have to do with text analytics? A lot! The ads that consumers see online are often tied to keyword searches, so text is always an element of the process. Could you go further? What if you could use more than just keywords from consumer posts? If you also knew something about the consumer’s mood and personality, and had more context around keywords, would that help you to create more compelling ads?
How would you put your new ideas into practice? You’d need clever ways to serve ads based on this information, and a solid testing program to learn what ads work for whom. This process could be challenging, but with the right planning process you could assess what it’s worth to you, then seek out the building blocks to make it work.
Whether you are just getting your toes wet in text analytics, or digging deep, you surely hope for good returns for your efforts. In addition to hoping, try planning. You’ll get much better results.
Meta Brown is author of "Data Mining for Dummies" (forthcoming from John Wiley and Sons). She has introduced and expanded the use of analytics in offices and factories across the US and beyond. Got a question about promoting analytics? Or on using analytics? Just want to say hello? Email Meta at firstname.lastname@example.org, tweet her @metabrown312 or visit http://www.metabrown.com
Other Posts by Meta S. Brown
The moderated business community for business intelligence, predictive analytics, and data professionals.