Bloomberg News

Recommind Wins Document-Search Deal With SEC in Next Step to IPO

January 30, 2013

Recommind CEO Bob Tennant

Recommind Inc. Chief Executive Officer Bob Tennant poses for a photo at company headquarters in San Francisco on Jan. 24, 2013. Photographer: David Paul Morris/Bloomberg

Recommind Inc. Chief Executive Officer Bob Tennant says his company is taking off, and he’s just clinched the deal to prove it.

His San Francisco-based startup, which provides electronic discovery software to companies and law firms, boosted sales 50 percent last year, surpassing $70 million, while top rival Autonomy Corp. brought about an $8.8 billion writedown at parent Hewlett-Packard Co. (HPQ:US)

Meanwhile, Recommind has just landed one of its most high- profile contracts. The U.S. Securities and Exchange Commission has been testing Recommind’s software since mid-2012, and is now deploying the service to 1,200 employees in the regulator’s enforcement unit, the company said today. Like Recommind’s more than 300 customers, the SEC is using the technology to quickly search for, and identify, relevant documents, reducing the time and costs of investigations.

“A lot of the times, when the SEC misses things, it’s because there’s too much stuff,” Tennant, 43, said in an interview at his office in San Francisco’s financial district. “The traditional ways of dealing with it are no longer workable. They came to that realization.”

Following a year that included contracts with Siemens AG’s U.S. subsidiary and the New York State Bar Association, the SEC deal puts 12-year-old Recommind in position to join the crop of business software companies that have recently sold shares on the public market.

IPO Prospects

Splunk Inc. (SPLK:US), a maker of software that helps businesses analyze data, and cloud-software providers Workday Inc. and ServiceNow Inc. have surged since their initial public offerings last year. Each trade at 18 to 40 times revenue over the past 12 months, compared with a price-to-sales ratio of 2.3 for the Standard & Poor’s 500 Index technology group.

Tennant said he hasn’t ruled out an IPO this year and is also considering a private-financing round.

“We feel pretty comfortable about the valuations we would be able to command, in either the private or public market,” Tennant said.

To help prepare the company for an IPO, Recommind added Oracle Corp.’s former finance chief Jeff Epstein to its board in August. About 200 of Recommind’s 450 employees are in engineering, with the bulk of the company in San Francisco and Bonn, Germany.

The e-discovery market, including software, services and hardware, is expected to grow 35 percent by 2016 to $22.8 billion, from $16.9 billion last year, according to Vivian Tero, an analyst at IDC in Boston. It’s a fragmented market that includes Hewlett-Packard’s Autonomy unit, Symantec Corp. (SYMC:US)’s Clearwell Systems Inc. division, as well as International Business Machines Corp., Oracle and EMC Corp.

Industry Standard

Recommind is seeing less of Autonomy and Clearwell in the market since they were acquired in 2011, Tennant said. Still, the two competitors are backed by giant balance sheets (HPQ:US) and global brands, representing a risk to a startup. For Recommind to continue its growth trajectory, it needs to become an established industry standard. The SEC deal can help Recommind achieve that, according to Tero.

“Any organization that’s constantly being investigated will probably want to use the same type of tools that regulators are using, because it solves a lot of problems,” Tero said.

While Tennant didn’t disclose the size of the contract with the SEC, he said most clients pay in the hundreds of thousands of dollars per year, with the largest deals exceeding $5 million. The SEC is already using the software for more than 100 cases, he said.

Complex Searches

The deployment, part of an agency-wide technology upgrade, is being used to handle the 6 terabytes of data (equivalent to 250,000 boxes of paper) handled each month by the SEC’s enforcement division, according to Adam Storch, the division’s chief operating officer.

“This will change the way enforcement does business,” Storch said. It “makes us more productive and effective, especially in large, complex matters that challenge traditional approaches,” he said.

Started in 2000 by two graduate students in machine learning, Recommind found its niche two years later as a provider of search tools for the legal market. In 2007, the company expanded into e-discovery and investigation, and has bolstered sales between 50 percent and 120 percent each year, Tennant said.

More Efficient

The software is used by large companies facing constant litigation, regulatory agencies and businesses that are in highly regulated industries such as energy, financial services and health care.

For clients like law firm Wilmer Cutler Pickering Hale & Dorr LLP, the key to Recommind’s technology is what’s called predictive coding, a process that goes beyond keyword search, fetching documents and data based on intent and meaning. For example, a search on “credit default swap” would return documents with that exact phrase, as well as related concepts.

Steven Berrent, managing director of Wilmer Cutler’s DiscoverySolutions group, said Recommind’s software can reduce the number of documents reviewed by humans by 80 percent, and save 50 percent on the costs of a case.

“The old-fashioned way is to have a lot of lawyers working in a room and you just go document by document, to the bottom,” Berrent said. “This moves us from purely commoditized document review, which is inefficient, expensive and cumbersome, and enables us to get much more value out of it.”

To contact the reporter on this story: Ari Levy in San Francisco at alevy5@bloomberg.net

To contact the editor responsible for this story: Tom Giles at tgiles5@bloomberg.net


Too Cool for Crisis Management
LIMITED-TIME OFFER SUBSCRIBE NOW

Companies Mentioned

  • HPQ
    (Hewlett-Packard Co)
    • $34.93 USD
    • -0.01
    • -0.03%
  • SPLK
    (Splunk Inc)
    • $63.24 USD
    • 3.41
    • 5.39%
Market data is delayed at least 15 minutes.
 
blog comments powered by Disqus