[Originally published in the Clarabridge Bridgepoints newsletter, Q1 2008.]
Hosted/on-demand text analytics is the preferred service-delivery model for a rapidly growing segment of text-analytics users. Let’s call this approach, a variant on the increasingly popular Software as a Service (SaaS) model, Text Analytics as a Service (TAaaS).
SaaS comes in a number of flavors. With hosting, you contract another organization to install and maintain software for your use. With on-demand, you pay for only the services you use. Both styles of SaaS may be delivered via Web services interfaces or via more-traditional data-integration approaches, that is, where the provider in effect hosts selected business processes.
TAaaS adoption, like other moves to SaaS models, is a response to efficiency, cost, and rapid-delivery business drivers. Users don’t have to install and manage software, allowing a focus on business goals rather than on IT. They can get started quickly, relying on the TAaaS provider’s expertise, without incurring full software-licensing and training costs. Uptake therefore not only represents a change in how software is used and paid for; it is a key enabler that is accelerating the adoption of text analytics by organizations that would otherwise not have the resources to license, design, and manage in-house solutions.
Fern Halper, an analyst with Hurwitz & Associates, said that a text-analytics survey her firm conducted last year asked how prospective users view SaaS models. “We asked companies that were planning to implement solutions in the next year whether they would consider utilizing text analytics in a SaaS mode. A little more than half said yes. 20% don’t understand enough about the SaaS model and 12% replied ‘don’t know.’ Only about 16% said no. So there are definite possibilities for success here.”
That much of the text used in certain business domains is on-line and distributed rather than “behind the firewall,” where installed software typically resides, has only accentuated the trend toward TAaaS. Even though “most business don’t yet like on-demand applications reaching back inside their firewalls for documents” according to SaaS guru David Linthicum, only 8% of Hurwitz respondents would consider TAaaS “only if the documents were external to my company,” with 43% saying they would consider TAaaS regardless where documents are held.
Customer Experience Management (CEM) applications used by consumer-facing businesses are among those that rely heavily on text-rich on-line sources – blogs, forums, reviews, and news and social media – as well as on previously isolated, “behind the firewall” sources such as contact centers that generate large volumes of text. Not surprisingly, analyst Fern Halper reported that more than 70% of respondents to Hurwitz’s 2007 survey expressed interest in text analytics for customer-care type applications. CEM is clearly promising territory for both text analytics and application of SaaS models.
Hosted Text Analytics for Customer Experience Management
Customer Experience Management seeks to capture the voice of the customer from on-line media such as blogs and forum postings; from e-mail, on-line chat, and contact-center dialogues; and from surveys and other mechanisms for collecting customer feedback. CEM incorporates but goes beyond traditional transaction-based Customer Relationship Management (CRM) to create a fuller picture of customer and market attitudes about an organization’s products and services. According to Clarabridge CEO Sid Banerjee, industries such as travel and hospitality and retail “live and die on customer experience.”
Gaylord Hotel’s text-analytics implementation illustrates the possibilities afforded by a hosted text-analytics/CEP program. Business at the company’s resorts and convention centers is 80% meeting driven according to Tony Bodoh, manager of operations analysis, and the company sees definite competitive advantage in being able to analyze the full spectrum of information generated by customer interactions.
Gaylord’s work in text analytics has initially involved customer-satisfaction programs where the company has used third-party vendors in addition to internal resources. They needed an approach they could test for a time with minimal impact on IT operations. And they needed text technologies that would integrate well with their existing analytics infrastructure, not only with BI solutions from their existing vendors but also with industry- and application-specific approaches. They found that most vendors either didn’t work in hospitality or were not focused on customer-satisfaction applications. A particular drawback was limited categorization models, that is, the ability to classify customer-satisfaction feedback related to the hospitality industry.
An initiative to bring contact-center information into the analyses is on the horizon for Gaylord; Bodoh sees hosted text-analytics as best assuring his company’s ability to keep up with a combination of rapid business expansion, broadened analytical scope, and evolving technology while still allowing the possibility for the company to bring analytical operations in house at a later date.
Philip Russom, senior manager at TDWI research, sees demand for hosted text analytics fueled in part by the head-start provided by vendors’ domain-adapted analytical models. According to Russom, “Hosted text analytics would make sense, if it helps user organizations minimize the development commitment involved in creating and maintaining a vocabulary of entities, facts, and relation types. Since users of text analytics seem to congregate in specific industries – namely, federal government, insurance, automotive, and healthcare – hosted text analytics should provide lexicons and relationship models for these, plus for cross-industry applications like call centers and e-commerce.” Yet for many organizations the focus is on information-centric services and on outsourced business processes rather than on the installed vs. hosted question.
Some on-demand text-analytics users aren’t even aware that they’re using text analytics, for instance,
- users of content-search and analysis services such as Factiva and Silobreaker and clients of information-services providers including Thomson and LexisNexis,
- users of a spectrum of leading publishers and media outlets, and
- users of a spate of new entrants in the media-monitoring and reputation-management spaces.
These are essentially research services supporting all manner of information consumers, typically using custom-developed text-analytics and information-retrieval solutions. We see another promising direction for Text Analytics as a Service: Emerging TAaaS products will make it easier for solution providers to offer analytical capabilities, packaged with domain-related information and workflow, to customers whose primary concern is business needs rather than technology.
In sum, Software as a Service provides a compelling text-analytics solutions model. Whether implemented via a hosted platform or working behind-the-scenes to add analytics to an information service or line-of-business solution, the SaaS model can deliver high-impact analytics with reduced IT overhead. Users in a variety of business domains have responded to the possibilities of Text Analytics as a Service, paving the way for wide adoption in the years to come.