Market intelligence will get even smarter by early third quarter when Factiva, a business unit of Dow Jones Reuters Business Interactive LLC, is expected to launch an application dubbed Reputation Management.
In the early adopter phase at the moment, somewhere between post-beta and pre-production, Reputation Management will allow companies to detect and analyze business and social issues that might affect their brand product or service.
The solution combines Factiva’s ability to aggregate large content sets, which number as many as 9,000 horizontal and vertical media outlets in 22 languages, according to Dennis Cahill, associate vice-president of technology and IBM Corp.’s WebFountain technology.
WebFountain provides a single integrated platform of unstructured data, aggregated from various content providers, for text analytics. Once the content is delivered to the unified architecture, called Unstructured Information Architecture, Factiva can plug in numerous text analytics programs of its own and of IBM’s that use natural language understating, statistics and probability, machine learning, pattern recognition, and other artificial intelligence tools.
“If you make unstructured data look enough like structured data, you can apply data mining to it,” said Dan Gruhl, chief architect for WebFountain.
WebFountain technology, a research program which will see its first production version go live with Factiva, uses a process of “disambiguation” to refine searches by looking at the words around the name, said Gruhl.
For example, using disambiguation, a search of Shell Oil Co. would filter out marine sites referencing sea shells, said Gruhl.
Reputation management will both monitor trends from a list of client issues that the client is concerned about and discover issues that the client may not realize are relevant, said Cahill.
The solution discovers the additional trends not listed by the client by finding all mentions of a company, its executives, products and services, and the most frequently occurring words and phrases surrounding those references.
“A lot of early warning signs tend to be ignored,” said Cahill.
For example, Cahill cites the current furor over obesity, where debate escalated from discussion groups, message boards, and blogs all the way to class action law suits similar to the type experienced by the tobacco industry.
“The signs were there with the number of mentions growing over time. It was clear this was coming to a head,” Cahill added.
Despite the recent congressional action to disallow such obesity suits, companies such as McDonalds Corp. have already changed their menu, dropping super sizing for example, in essence to manage their reputation.
Factiva also partners with another market intelligence company, Biz360, which has a slightly different take on how companies can manage the consumer perception of a brand.
Biz360 measures sentiment, both negative and positive points of view, from the media and analysts, using on average 12,000 publications, and turning speech to text to monitor broadcast media in the top 100 markets. Its two solutions are Market360 and Market360 Analyst.
The Analyst solution, introduced four months ago, measures sentiment by monitoring published reports of industry and financial analysts.
Sometimes searching for mentions and extrapolating from there to say whether the article was positive or negative is subjective, according to Nick Patience, senior analyst at The451 research group.
“Biz360 knows that, but they present it from the customer’s point of view,” Patience said.
Biz360 will give a client 30 to 90 articles, asking the client to rate each one as positive, neutral, or negative.
What is considered positive, neutral, or negative varies by company, according to You Mon Tsang, Biz360s chief marketing officer and founder.
“If an article says a company is an aggressive and in-your-face company, Oracle would consider it a positive point of view. Say the same thing about an insurance company and they would see it as negative,” said Tsang.
Today more than ever, a company needs to gather and understand feedback on brand perception from all sources. In the past, perceptions were usually shaped by major publications, said Tsang.
“Today, a random blogger saying a random thing can bring down a company,” Tsang said.
Patience said that there must be a willingness to suspend disbelief in order for a company to put its faith in these types of market intelligence applications, but they do work.
“These guys employ algorithms that use statistical measurements and they have been well-tested in various search engines,” Patience said.
