BigDataNYC debate: Are Hadoop upstarts the future or road kill?

BigDataNYC Panel

Capital Markets Day panelBig Data analytics is going to change business as we know it, but there’s no guarantee that the companies that dominate the market today will be around to take a victory lap.

That was a consensus of a Wikibon Capital Markets panel at the BigDataNYC 2014 conference last October, in which a tech entrepreneur, a veteran Wall Street research analyst and a Big Data evangelist from Cloudera, Inc. sharply differed on the outlook for the hyper caffeinated Hadoop market.

Although the panel took place more than five months ago, the discussion seems even more relevant in light of recent developments in the Hadoop market, including the initial public offering of Hortonworks, Inc. and the formation of the Open Data Platform consortium as a direct challenge to Hadoop kingpin Cloudera, Inc.

One thing all panelists could agree upon is that Big Data is much more than a cheaper way to dig through information. It’s a game-changer that’s on par with the enterprise resource planning (ERP) revolution of the 1990s.

Transformative technology


Former Cowen analyst Peter Goldmacher drew that analogy when he cited a study his company had conducted across ERP from two decades ago. “ERP gave these customers the ability to automate and scale, and it let them get much bigger,” He said. “It created an entirely new kind of company where R&D spend at as a percentage of revenue declined by a third, revenue was up 8X and profits were up by 5X.”

Amy O’Connor, a Big Data Evangelist at Cloudera, agreed that Hadoop transforms the way a company thinks about data and, by association, the way it operates. “You can store a lot more data,” she said. “You can act on the lot more data than you could in the past, and you can build eye-popping scale on commodity servers without having to buy all the equipment up front.”

Equally important is that consolidating data breaks down silos and enables organizations to share more fluidly. “That leads to true innovation because you can get at the data and processes a lot more quickly,” she said.

The promise of this kind of scalability is that even companies that count their customers in the millions can think about addressing them individually. “If Google can personalize search for two billion people, the least we can do is personalize banking for 100 million,” said Abhishek Mehta, founder and CEO of Tresata, Inc. and a former managing director at Bank of America.

ROI stumbling block


If the future is so bright, then why are initial results from the field so modest? A Wikibon research report published shortly before the panel took place showed that most enterprises had achieved a return of just fifty-five cents on the dollar on their Hadoop projects. While users were still bullish about the potential of their Big Data investments, the research also showed that nearly half said they had realized only partial value from their deployments.

Be patient, panelists agreed. Disruptive change doesn’t happen overnight, and it’s natural for users to concentrate early deployments on cost-saving applications. Cloudera’s O’Connor noted that the Financial Industry Regulatory Authority (FINRA) “will save $20 million this year just by moving their data to a Hadoop environment. That’s more money for them to invest in other things.”

Mehta agreed. “Even a 1% operational improvement can be a very big number,” in some businesses, he said. “We see very few customers talking return on investment. We see many customers talking about reduction in investment.”

Looking at Hadoop as just a way to save on storage is missing the bigger picture, said Wikibon analyst Jeff Kelly. “You would miss a huge part of the value proposition.”

Mehta agreed. Calling Big Data “the fundamental building block of the next Industrial Revolution,” he said successful companies in the future will be those that recognize unique value in the data they collect and use it to transform industries rather than just achieve incremental efficiencies.

He gave the example of one banking customer that has a 50% market share in the payments business. This gives it unique insight into the world economy. “When you have visibility like that, I would buy the stock,” he said.  “We are rebuilding the IT stack,” around data. “There will be multiple $100 billion companies built from this.” Google, Facebook and Amazon are three early examples.

But is there a business?


Where panelists differed was on the business models that are evolving amid an investment frenzy around Hadoop and the big data ecosystem, and which companies will ultimately dominate the landscape.

Goldmacher played the provocateur, asserting that sky-high valuations are forcing some companies to grow at unsustainable rates and sacrifice profits for revenues. “ does $5 billion in sales, doesn’t make money and has a $35 billion market cap. Nobody gives a s*** about making money anymore,” he said. “They just want to get big.”

The estimated $500 million in sales that Hadoop-related companies realized last year “probably came as a result of a billion dollars in spending,” Goldmacher added. “If I’m selling dimes for a nickel, I’m going to expect to have a pretty good business.”

Not surprisingly, Cloudera’s O’Connor differed. She pointed out that Cloudera used some of the capital from Intel Corp.’s massive $900 million investment to buy security startup Gazzang, Inc., plugging a gap in the company’s product line and opening up opportunities to sell into the financial services and health care markets. Intel also abandoned its own Hadoop effort when it invested in Cloudera, reducing confusion in the market and anointing Cloudera as the clear market leader.

But will the new batch of Hadoop upstarts ultimately be the big winners? Panelists sharply disagreed on the outlook for the Oracles, Microsofts and EMCs of the world, which are often said to be vulnerable to Hadoop incursion, but which have massive installed bases.

Shadow of giants


Goldmacher was again the fire-starter. Even the biggest Hadoop players are tiny in comparison to the giants of the IT industry, he said. “Cloudera says they’re doing better than Hortonworks because they have 100 field sales reps and Hortonworks has 75, but Cisco has 20,000 and Oracle has 35,000,” he noted. If the Hadoop market doesn’t focus on profitable business models, a crash is inevitable and the installed leaders will enjoy the spoils. “If these startups are spending all their money cramming stuff into the market, they are accelerating the inevitable trough of disillusionment, and then their assets will be available for a lot less than when Intel thought Cloudera was worth $4 billion,” he warned.

Panelists agreed that emerging Hadoop players need to pull together for the common good. “There is no partnering in the Hadoop ecosystem. The small players get no support from the big players,” Mehta said.

Customers also perceive too much complexity because of fragmentation that requires them to negotiate with a dozen different vendors to integrate Hadoop with their legacy systems, Goldmacher noted. “When someone has the whole fabric, then things go mainstream,” he said. The question is whether the fabric will come together organically and or be woven by a dominant company like Oracle.

Goldmacher also asserted that the vast majority of industrial companies don’t care about leading-edge technologies like Hadoop, but O’Connor said her experience is quite different.

“I spend a lot of time in the middle of the country with companies that make the food that we eat, the equipment that moves oil and gas and the companies that provide health care,” she said, and these companies are rethinking their businesses.

“Retail companies are looking at unified customer experience. Healthcare companies are thinking of new ways to connect doctors and patients and pharmaceutical companies are thinking of new ways to manage trials,” she said. “The highest value comes when people think about data in terms of how it can transform our lives.”

Watch the full panel discussion (begins at 46:30, duration 50:26)