I like some of what Attensity does, though I am not a raving fan, yet.
The Palo Alto company provides a suite of applications that use natural language processing and other technologies to derive meaning from unstructured data. Good for them.
The natural language processing idea is something that I have been watching for almost ten years. In the early part of the last decade, it seems like a lot of smart people from the Stanford Linguistics Department (of some other department) went commercial with their research and began giving application developers a way to learn what their customers were telling them without having to actually apply a human ear or eye to the situation.
Today, Attensity sent me an email saying they’d analyzed 15,000 tweets about the new Nexus phone from Google and showed me the results in charts and graphs. This was a significant and powerful demonstration of sentiment analysis. The brief report showed what people were tweeting that they liked and disliked about the new phone.
Now, sentiment analysis is a very useful tool and I am a big advocate of anything that gives a vendor the ability to listen to customers. I am an even bigger advocate of vendors that actually listen to customers and base their product and messaging decisions at least partly on what customers tell them. I call this, undramatically, using social technology as a stethoscope rather than as a megaphone, which is too often the first thing that people try to do with social tech.
Every technology, every idea has its limits though and the Attensity example shows some limitations of the research method or assumptions used in sentiment analysis though not the product itself. As I was going through the report the data on why people liked the phone and why they didn’t seemed very credible. But when I got to a section on purchase intent, school was out, as they say. There was no data label on the pie chart but it looks like about eighty percent of the 15,000 tweets gave some indication that their authors intend to buy the phone.
This is where you begin to see the difference between sentiment analysis and real research — science based research that uses controls on its variables. While I am sure the folks at Google would be delighted to sell phones to eighty percent of 15,000 people (or was it tweets? How many people do these tweets represent?) or about 12,000 phones, might want to do a deeper dive to determine true demand.
I wouldn’t build 12,000 phones based on this analysis. I’d like to know many other things first such as the ability of these people to buy — are they in contracts with other providers now? And, of course, how many people do these tweets represent? Sentiment analysis is a good start but I want to know more.
So what does this mean?
Well, we have new and very powerful technology that can give us quick answers. But now we need some discipline and methods to ensure that what we get is useful information. We’ve invested a lot of time and treasure over the last few years in the customer experience and much of that was money and effort sell spent. Now as we try to innovate our way out of the recession it might be good to turn our attention to doing the rigorous work that ensures we don’t over invest in sentiment when what we need is analysis. Like any tool, this Attensity product can deliver powerful results if we use it right. I hope we do.