How HP is learning to predict customer behavior by merging social media and company data

A pilot project completed by HP Labs and HP Global Customer Intelligence shows that data from social media can be merged with company data to predict customer behavior with up to 90% accuracy.

The pilot demonstrates that collaboration between marketing and IT departments is critical to unlocking the business value of social media. The system used can correlate social media conversations about specific product features to actual customer transactions in real-time.

HP is expanding the scope of the project internally as the underlying technology is being piloted by HP customers in the media and entertainment industries.

A key marketing opportunity for the digital age

The owner of a local restaurant can easily read a few dozen reviews of her business on Yelp, understand them, and act on them.

For larger, global companies – that are the subject of tens of thousands of comments every day – a human-only approach doesn’t work. The alternative is to put software analytics in place. But without merging that analysis with other business data, insights from social media remain disconnected from the rest of the enterprise, limiting the opportunity to act.

Project Fusion: Results

HP called its pilot “Project Fusion” because it combines two different kinds of data:

1. unstructured data (Amazon.com reviews, customer surveys, customer support logs, and other natural-language text)

2. structured data (customer support tickets, sales transactions, customer demographics)

Using HP Labs’ new text analytics technology, the team first converted the unstructured data into a structured format. Then, human analysts could use standard data-mining and statistical tools to analyze the two data sets together. In one case, social signals predicted support tickets with 90% accuracy. In another case, different social signals were highly correlated to sales.

“The results are significant enough to act on,” explains Prasanna Dhore, VP Customer Intelligence, HP. “Applying this methodology can improve both the customer experience and the bottom line.” For example, if HP can anticipate product issues, it can deploy specialized support staff before customers call for help, he said. The signals could also be used to fine-tune the company’s marketing spend while a campaign is still running.

How HP is learning to predict customer behavior by merging social media and company data

View more presentations from Hewlett-Packard

How HP’s system works

Like many companies, HP has already wrung most of the useful information out of its existing customer database. To tap into the richness of social media, HP had to invent new technology.

The text analytics HP Labs created are part of a broader “Live Customer Intelligence” project that includes data visualization, parallel processing, and other key components.

“Prasanna needed the ability to drill down inside a customer review to identify specific product attributes and assign sentiment to them,” says Meichun Hsu, Director of Intelligent Information Management at HP Labs. “That’s what our technology does. Because it can identify sentiment more granularly than other solutions, we could convert text into structured numerical data.”

The software is also designed to work in real-time and works with database technology like Vertica’s, which HP acquired in January 2011. (In fact, customers of Vertica include social media sites like Twitter and Groupon).

Bringing the solution to HP customers

“We see an enormous opportunity for this in every industry,” says Jeff Edlund, CTO, Communications and Media Solutions, HP. “There are particularly interesting applications for media and telecommunications companies, which already have detailed customer databases. The more customer data you have to begin with, the more insight social media signals can add.”

As Dhore notes, though, the reality is that bringing this to life is 5% vision, 95% execution. “Transformative projects like this, which cross organizational boundaries, require intense focus and cooperation,” he notes.

The challenge is not lost on Edlund, but it’s one he feels equipped to meet. “One reason we’re so excited is because HP has all the capabilities customers need to integrate social media with enterprise IT.”

Complementing the capabilities created in the lab, HP further expanded its portfolio of unstructured data analytics by acquiring Autonomy earlier this month. Autonomy is a leader in enterprise information management and meaning-based computing.

“The exploding growth of unstructured and structured data and unlocking its value is the single largest opportunity for consumers, businesses and governments,” said Meg Whitman, HP CEO, in a recent statement about Autonomy. “We are committed to helping our customers solve their toughest IT challenges.”