Sramana Mitra: The problem you describe is clear and easy to understand: unstructured data that comes through open-ended comment boxes in market research forms. You help process those comments and categorize them.
Tell us more about what other trends are driving behavior change inside the corporations that are prompting you to move more in that direction.
Rick Kieser: This is coming to social media, big data, etc. There are lots of different ways. About a decade ago, nobody could process this data very effectively. It was all generated through survey comments or solicited feedback, either through telephone, paper, or online interviews. The rise of unsolicited feedback coming from the Internet, social media, or call centers in the last decade has posed a challenge for a number of technology firms to enable digital researchers to become more efficient in managing unsolicited feedback.
This has been a huge trend over the last decade. I have observed it because I led pioneering investments and text analytics firms back in 1999. It is amazing to see how this industry has evolved over the past 15 years and how automated technologies are enabling even higher coding and classification in terms of responses per hour that we can do today.
Having said that, within our platform we use a hybrid approach. We don’t choose just one technology. We use a base of semi-automated technologies that represents a core that humans read and process. But then we overlay two other automated technologies in our platform, which nobody else has done. One of them is a machine based learning approach. On top of that there is also a natural language processing technology. All three of these approaches – NLP, machine learning, and semi-automated coding – have different characteristics, depending on the channels of feedback, whether it is social media, survey responses, call center–transcribed feedback, responses from marketing research online communities, or responses as simple as electronic documents or emails. Each of those channels has different characteristics to which you need to apply different technologies to be able to optimize your insights.
SM: What is the workflow of all this? How are you going to market with it?
RK: We have a services arm, in which we code and classify open-ended social media comments for corporations. We use our own platform for that. But more important, at the core of our business there is SaaS. We license our product and embed it into client’s natural workflows.
If you think about it from a market research perspective, companies interface with the customer, design a survey, and then go to the field and implement it. Then [the data] comes into a survey-gathering platform, which we don’t do, and the open-ended comments are spit out of that and injected into our platform, in which they are coded either manually or using automated technologies. Once those are completed and classified on a code book, they go back into the survey platform for the market research front. We are very much embedded in the natural workflow because this is quite complicated to do on a scaled level.
SM: What percentage of your business is products, and what percentage is services?
RK: In terms of revenue it is 60/40. But from a transactional perspective, our economic model is based on how many comments are classified in the Ascribe platform. In this perspective it is 90/10. Ninety percent of our volume comes from our SaaS offering.