Vendors, the trade press and even popular media are talking about the world’s rapidly expanding body of electronic data. Lots of data equals lots of value, right? The not-so-subtle message surrounding all this is that data is the new gold rush, so you’d better get your stake in the game now.
Often, the data is a by-product of everyday business or online investments in websites and social media. A lot of that data is unstructured – text, images, audio, video… , so text analytics is a natural step for those are beyond the basics in web analytics.
Text analytics isn’t easy. Nor are positive returns guaranteed. In fact, the profit story is pretty dismal. A 2014 market study by AltaPlana Corporation found that the majority of respondents using text analytics were not getting positive return on investment.
The fact that most organizations using text analytics are losing money is an embarrassment to the industry. But it doesn’t have to be that way. There are businesses getting positive returns on text analytics today. What sets them apart from the rest?
Have you looked at what text analytics vendors say about their offerings? What do they promise? Insights. I did a quick online search for text analytics software. Here are clips from the websites of each of the first four vendors I came across:
Insight is good. Nobody wants a shortage of insight. But, tell me this – what’s the value, in dollars and cents, of an insight?
Profit has a clear definition. It’s the money, the cold hard cash that your business earns. Profit is what remains after all your costs have been subtracted from revenue. Insight is not money. It might help you make money, but there are a lot of steps between an insight and profit.
Some businesses make a profit on text analytics. Yours can too. But there’s more to it than buying software and looking for insights.
Businesses who don’t break even on text analytics are consistently missing the one thing that maximizes the chance of making a profit. Here it is folks, the text analytic profit strategy that works, in just four words: YOU NEED A PLAN.
Well, maybe it should be five words. YOU NEED A DECENT PLAN.
Now, you may be thinking, “What kind of news is that? I need a decent plan? Tell me something I don’t know!”
Yet lack of adequate planning is at the heart of almost every failure to turn a profit from text analytics. If you doubt that, then the next time somebody tells you, “We tried it, but it didn’t work out.” ask what the plan was, and where it went wrong. See what you think of the response.
Here’s how you make a decent plan for text analytics, one that gives you a real fighting chance of making a profit.
Return on investment has just two main elements – costs and benefits.
Don’t start with costs. By that I mean, if costs are the first thing you think about, you’re doing it wrong. You shouldn’t be doing, or thinking about anything that costs you money until you have first determined how you are going to benefit.
Your starting point must be – what benefit do I seek, what goal can I set, that text analytics can help me reach? Remember, insight is not a goal. You must have a well-defined business goal that is measurable in dollar terms. The goal is the return part of the return on investment calculation.
One way that people get into trouble with text analytics is in not knowing what concrete things can be done with it. So, here are a few ideas for you to consider. See if some of these fit your needs, or get you thinking about other applications that are relevant for you.
Note that all these applications have clear ties to monetary benefits – reduced costs, especially labor costs, or increased revenue. Some can offer both costs savings and increased revenue potential – isn’t that nice?
Text analytics isn’t perfect, so don’t choose a goal that requires perfection.
One of the most popular text analytics applications is sentiment analysis. That is, assessing text for positive or negative attitude. Sentiment analysis is widely used for assessing social media posts and opinions of the brands mentioned in those posts. This is a great way to spend money and get no clear benefit.
Let’s say that you know the sentiment expressed in every single social media comment that mentions your brand. So what? How will you turn that information into money? That takes some thinking.
Text analytics is always imperfect, but sentiment analysis is really imperfect. So, it’s a safe bet that managers at your company, if they look at sentiment analysis at all, will look and spot some big errors. Somebody will see “That’s a bad ride.” classified as negative, and never take sentiment analysis seriously again. Bottom line, if your idea of what to do in text analytics is sentiment analysis, don’t expect to make a profit.
Choose applications with clear connections to increased revenue or reduced costs. Both would be nice.
Be realistic about the nature of text analytics. It’s about automation, not perfection. So focus on activities that are repeated frequently, where each repetition offers a chance for revenue or savings. With frequently repeated actions, you don’t have to be perfect to make a profit. You just have to be right more often than wrong.
For example, look at customer churn. You may get many communications from customers each day, more than you have resources available to read and act upon in a timely manner. Or perhaps you have resources, but would prefer to prioritize their activities or put them to work on other tasks.
If those communications are in email, social media or via online forms, this is a natural for text analytics. You can use text analytics to spot high-priority issues or customers on the verge of defecting to a competitor.
Don’t try to be fancy. Subtlety isn’t important. For example, Han-Sheong Lai of Paypal has told a success story about Paypal’s use of his team’s text analytics to reduce churn. He calls it sentiment analysis, but it isn’t like the sentiment analysis you’ll see elsewhere. He doesn’t try to guess the implications of everything said about his brand. He’s looking for customers at risk of closing accounts, so his application searches for statements like, “I’m going to close my account.” He wastes no time on subtlety.
How about you? If you were very effective, say, at finding messages that clearly indicated intention to close an account, could you route those to the right team for remediation and save some accounts? How many do you think you could save? What is each of them worth to your business?
That’s how you start a plan, and that’s how you make a profit in text analytics.
Meta S. Brown is author of Data Mining for Dummies and creator of the Storytelling for Data Analysts and Storytelling for Tech workshops. www.metabrown.com.