In a recent survey by AMR Research it was found that “Fifty three percent of companies had active SOA projects by the end of 2007.” The companies that adopted SOA spent an average of $1.4 million to implement SOA on software and services in 2007. These findings were confirmed in a survey prepared by BEA Systems. The BEA survey reported that half of all enterprises with revenue exceeding $1 billion have shelled out over a million on their SOA efforts and expect to spend even more over the next 12 months.

Financial Services organizations were least likely to be adopt SOA but those that did spent significantly more than their peers in other industries.

So does that mean that most organization will face such a large expense to implement SOA? What is the real cost for a SOA implementation?

In a traditional software estimate one would consider: Software size, Complexity, Technology, and Constraints. These individual factors are counted, weighted, and combined to identify an estimate of total effort. The total hours are then multiplied by cost factors for a total cost estimate.

In the SOA world the underlying fundamentals are much the same. By summing four essential estimation components an estimated cost can be derived. The four components are: Data Complexity, Service Complexity, Process Complexity, and Enabling Technology. (your terms, or a citable source here?)

As with more traditional estimation methods, size matters. In a traditional estimation approach one would identify the size of the problem by estimating lines of code, function points, or by some other method. In the SOA community one is concerned with the number of data elements, the complexity of the data storage for those elements, and the cost to understand and refine each element.

David S. Linthicum of the SOAInstitute.org provides a simple formula for the calculation.

Cost of Data Complexity = (((Number of Data Elements) x Complexity of the Data Storage Technology) x Labor Units))

• The "Number of Data Elements" is the number of semantics you're tracking in your domain, new or derived.

• Express the "Complexity of the Data Storage Technology" as a decimal between 0 and 1. (For instance, Relational is a .3, Object-Oriented is a .6, and ISAM is a .8.)

• "Labor Unit" is the amount of money it takes to understand and refine one data element. Dave said this could equal $100, for example.

Example: ((2,000 data elements) X .6 complexity) X $150) equals $180,000 for the total Data Complexity Costs.

Now repeat the same formula for the Service Complexity, Process Complexity, and Enabling Technology. The final total should be within 10 – 20% of the actual costs. However, consideration should always be made for changing requirements, scope creep, changes in technology – and the myriad of additional real life factors that have become lessons learned in traditional software development projects.

The factors that go into a SOA estimate are similar to a traditional estimate in many ways. As was demonstrated above, consideration must be made for size, complexity, staffing, and many other parameters. Products such as SEER for Software™ by Galorath can be used to help identify and quantify the underlying components that make up a SOA estimate. Within SEER for Software the estimate can be built, evaluated, assessed for risk, and delivered as part of a repeatable estimation process.