Analyst research 'distorted' against open source

JBoss World: A top HP executive has said that open source software is being under-represented in market share statistics because analyst firms are using outmoded measurement techniques
Written by Ingrid Marson, Contributor on

HP is hoping to persuade analyst firms to change the way that they measure the software market to take into account the open source business model.

Martin Fink, the vice-president of HP's Linux division, said in a keynote speech at the JBoss World conference in Atlanta on Tuesday that analyst firms are currently under-representing the impact of open source, as their statistics rely on how much companies spend on licence fees. This discriminates against the open source business model where companies make money from support and services, rather than licences.

"Because the business model is changing with open source, they [analyst firms] have to change how they measure market share," said Fink.

Large analysts firms such as Gartner and IDC claim that the open source application server JBoss has a market share of around 2 percent, but the actual figure is probably much higher, according to Fink.

"This figure [of 2 percent] is really distorted," said Fink. "We are challenging the industry analyst system to change. Hopefully industry analysts such as IDC and Gartner will do this."

He pointed out that research by analyst firm BZ Research found that JBoss' market share was significantly higher. BZ Research interviewed over 750 development managers in November 2004 and found that 34.8 percent of respondents were using JBoss, while 33.9 percent, 28.7 percent and 22 percent were using IBM WebSphere, BEA WebLogic and Oracle Application Server respectively.

HP, which is partnered with IBM WebSphere, BEA WebLogic and JBoss, said it has seen a noticeable increase in requests for the open source application server. "We are seeing an increase in demand, almost an avalanche around JBoss," said Fink.

Fink later told ZDNet UK that until recently licence fees were a good way of measuring the market share of products as it took into account the number and size of deployments. He said there is no easy way at present of measuring market share that takes into account commercial deployments of both open source and proprietary vendors, but he is working with analyst firms to see how this can be resolved.

"Right now I've identified the problem, but we have to work on the solution," said Fink. "Counting downloads is OK, but it's not a measure that I could use to do business planning."

Editorial standards