The ROI Of SOA: Get It Right!

In which this reporter is taken to task for exaggerating (negatively) the claims in a research report about the return on investment for service-oriented architecture. Thirty-seven percent is not nothing!Last week I wrote a blog about the ROI of SOA -- or lack thereof. SOA is one of those BIG I.T. INITIATIVES, the ones that involve the entire enterprise, represent massive expenditures of time, effort, and money, and are generally hard to measure in terms of, well, return on investment.

The blog was based on three things: a presentation at the InformationWeek 500 Conference two weeks ago by the No. 1 company on the InformationWeek 500 list of most innovative users of IT, Con-way; an interview with the CIO of Coty, the New York marketer of fragrances and cosmetics; and the results of a research report by Nucleus Research.

To sum up: Con-way is using SOA technology from Tibco to expand its logistics business; Coty used SOA technology from iWay Software to integrate disparate systems inherited in an acquisition of Unilever; and Nucleus Research published a report that said the return on investment from SOA is practically nonexistent.

Well, not quite. "I think you might have overstated the negativity," says David O'Connell, the author of the Nucleus Research report. O'Connell says his report, which was based on interviews with 106 organizations, found that 37% of the companies interviewed actually are realizing ROI from their investments in SOA technology and programming.

Full disclosure: I didn't read the report. ("It's only available for a fee," says O'Connell. "You gotta pay for it.") I read the press release Nucleus Research put out about the report (which can be found here). The headline of the press release is this: "SOA's ROI in the Enterprise is MIA, Nucleus Research Survey Finds."

"There is some ROI, but most companies aren't getting it," O'Connell says. That's because, even for those companies getting benefits from exploiting a service-oriented architecture, the benefits are limited to pockets within an organization, project areas or business units, and rarely extend out to the entire enterprise, which is where the real ROI would be realized, says O'Connell.

The limiters to realizing ROI from SOA, according to O'Connell:

1) SOA gets stranded on a project basis, so the ROI gets left on the table.

2) The reason SOA gets stranded is that nobody has the mandate to spend the money and time on the infrastructure that helps broaden the adoption of SOA -- infrastructure such as registries and repositories.

3) There are cultural issues, such as overcoming developer resistance -- most developers like to use their own code, not reuse someone else's code.

My colleague Charlie Babcock posted a blog about BT and its massive SOA conversion, based on Charlie's interview with George Glass, BT's chief architect. It's a great read about how BT is using SOA to create customer-centric systems and deconstruct product-centric ones.

It sounds like Glass got buy-in from the top to enforce the SOA initiative enterprise-wide and therefore avoid the "stranded on a project basis" limiter O'Connell was talking about. No specific ROI figures, but there is this: "By designing and building general-purpose services that could be invoked in the sale of different products, BT started to cut down on its software overhead and phase out older systems."

Sounds like ROI to me. I guess BT is one of the 37%.