Easy ways to get the answers you need.
Or call us at:
Insights from unstructured data: An interview with SPSS CEO Deepak Advani
Delaney Turner 270002T14M email@example.com | | Tags:  ibmsoftware spss iod_emea
0 Comments | 2,817 Visits |
Deepak Advani is the CEO of SPSS, an IBM company and IBM Vice President of Predictive Analytics. Back at Information on Demand EMEA, I had the chance to talk to him about the business potential of unstructured data and his vision for how predictive analytics can improve business outcomes. The transcript has come back and below, I've presented it for your persual.
Also, I'd be remiss if I didn't mention that the Business Analytics Forum @ Information On Demand 2010 is a great opportunity to discover what predictive analytics (not to mention the rest of our offerings) can do you for your organization. Finally, I'd be doubly remiss if I didn't mention that the early bird deadline is August 31. There are some significant savings for IBM SPSS and IBM Cognos customers, so be sure to mention them in your business case to go.
One thing I've been really struck by is the emphasis on unstructured data. How long has IBM been working on analyzing things like blogs and tweets?
We've been saying that about 80 percent of your data is unstructured, but the context was call center notes, open-ended survey responses, video and audio, and e-mail. Now, social media usage is increasing at incredible rates.
I'll give you an example: in 2005 I helped do the sale of IBM's PC business to Lenovo, and part of the agreement was for me to become their Chief Marketing Officer. One of the key things we focused on was brand transition from IBM ThinkPad to Lenovo ThinkPad. We didn't talk that much about it publicly, but one of the biggest decision-making criteria was what was being said online in social media.
We analyzed the sentiment from the blogs and looked at how many people were writing about IBM ThinkPad, compared to the number of people using ThinkPad or Lenovo ThinkPad. Over time, we saw the mentions of IBM ThinkPad go down and ThinkPad or ThinkPad with Lenovo go up. At the same time, we were watching the sentiment around ThinkPad itself. We knew it would be disaster if the sentiment toward ThinkPad itself started to weaken.
We watched that every day. And what was great was that even as the association with IBM was weakening, the sentiment around Thinkpad remained strong, which allowed us to be more aggressive in moving off the IBM logo.
Companies now want to bring that same type of sentiment analysis into their decision-making - to compare their brands against others and look at the trends. But the really exciting part is taking those insights and feeding them into a predictive model that actually lets you anticipate online buzz and take early action. Not a lot of companies are doing that yet, but that's where things are headed.
A quick follow-up on that: In one customer session the speaker talked about analyzing social media as part of their risk management processes. Is that something new to you? Is that where predictive analytics could go?
Yes, we've always said that the more data you capture, the better your view of your customers. So even if you're working in Finance you need attitudinal information. You need behavioral information. You need to understand the interactions people have with you.
A lot of companies now offer the ability to analyze social media content. What makes the IBM offering stand out?
There's probably more than a dozen companies that do client sentiment analysis in social media, and they've been doing it for at least five years. One way we differentiate ourselves is by having higher-quality algorithms. The text analytics capabilities in our new product, IBM SPSS Modeler Premium, can read emoticons. We can also now understand slang terminology and detect sarcasm in what people are saying.
We're working with clients to incorporate these insights into predictive models. For example: Say you want to launch a new product. Even if you're already monitoring the buzz within the community, with our solutions you can predict which date or tactic will give you a higher lift in positive sentiment. Then you analyze these results and feed them back into your model so the next time you launch a new product the results are even better.
Think about it this way: what if a CMO could go to the Board and say “We have four campaign choices. We've run each through a predictive model and we believe option three is going to give us the best lift and have the most positive impact on revenue, so that's how we're going to proceed.” I think that changes the game. And we're the only ones really talking about it.
You'd be surprised how many C-level people now recognize the significance and the impact of social media on their business. We've actually struck a partnership with Yale Business School, and we're going to be doing a series of seminars with CMOs to get them a little bit deeper into analytics and show them how social media can help with what you do every day.
"Predictive" is a really popular term right now. Can you define it in the IBM context?
Sure. A lot of people think when we say “predictive,” we're talking about predicting the future - that in two months there will be an earthquake. But we're really talking about statistics and probabilities. It's much more about the likelihood of a customer leaving your organization because they're showing signs and symptoms of somebody who would leave. We're not predicting one big future event, like a fortune teller looking at a glass globe. It's creating statistical probabilities based on correlations and patterns.
How would you describe their value to someone who's new to the field? How are they applied in business?
Predictive analytics help people become more proactive in their decision-making. A lot of these decisions are around customers: Who are my more profitable? Which are most likely to leave? Are they the ones I want to keep? What can I do now to make sure they don't leave?
You can also expand into fraud prevention, because there are patterns that tell you where and when it's going to happen. Risk management is another perfect example because all these things are somewhat predictable. HR is another example. Which of your top employees are likely to leave? You can detect and analyze the patterns to take pre-emptive action. Really, predictive analytics can be applied to almost any industry and business problem.