As steampunk evolved from Victorian-era-inspired sci-fi subculture to mainstream fashion concept -- whimsical mechanical touches making their way into Macy's 2011 Christmas window display and Prada's fall/winter 2012 menswear collection -- IBM tracked the transformation online. Now IBM believes other brands can use the steampunk example to learn how a long-lasting trend develops and how to try to foster one that resonates in the marketplace.

"We noticed a few years back that this subgenre of science fiction called steampunk was appearing in a number of different contexts," said Trevor Davis, consumer products industry expert at IBM. Mr. Davis and his team, based in London, followed the movement of steampunk as it bubbled up into more mainstream consciousness, and in doing so aimed to decipher the differences between a mere fad -- think of what's trending on Twitter or the online memes that come and go on a weekly or daily basis -- and a true trend.

They built a set of models using IBM software based around natural language people typically use when discussing steampunk, such as the Victorian age, airships, steam engines and materials like brass or leather.

Mr. Davis uncovered patterns in conversation on blogs -- yes, there are many dedicated to a variety of steampunk subcategories -- as well as in news sources, on Twitter, Facebook and Pinterest and in other user-generated content. Along the way, his team was proved wrong in some assumptions, such as that celebrities might adopt the steampunk aesthetic and help propel its popularity.

"It seems to be more driven by a few kind of large groups of everyday consumers who are interested in very specific things, like role-playing games or making their own handicrafts," he said. For example, Etsy, the marketplace for craft entrepreneurs, is loaded with steampunk-inspired jewelry, apparel and accessories such as masks and rings.

"In some form or another, we've been doing versions of this kind of analysis probably since 2004 or 2005," said Mr. Davis. "We've been building up libraries of the models" for brands in the food, drink, apparel and perfume categories. "We know there are forms of language that get used time and time again."

IBM is selling the data-analysis concept to clients as "The Birth of a Trend," offering trend detection services on a consulting basis or helping brands use IBM software and hardware to do it themselves. The company's Cognos Consumer Insight application, for instance, could be used by a brand or agency to conduct social media sentiment analysis.

Initially, an IBM analyst dashboard showed how social sentiment was distributed across concepts and "hotwords." The trick, said Mr. Davis, "is a large data set." He said it is also necessary to run models over unstructured social-data sets several times because it can be tainted with misleading information such as incorrect location or gender data, in order to isolate quality data that shows true patterns.

"Patterns must be run over and over and refined, in part to account for the way people naturally use language. "Strunk and White is not on their desktops. .... You have to adjust to the tribal effects," he said, alluding to the grammarian's bible, "The Elements of Style" by William Strunk and E.B. White.

IBM has taken its learning from the steampunk project to assist a brand client in launching a product it hoped would become an instant-classic. "We needed to understand what are the drivers that actually create a classic item in the first place," said Mr. Davis, noting the approach can apply to all sorts of products. To do that, his team built a model for what they thought would be an emerging classic product, then ran the model over different data sets in search of certain patterns. "It's kind of a progression of a conversation," he said.