Costly, Painful and Worth It

In fact, big companies often bring in not only third-party consultantsbut the EAI vendors themselves. For General Motors' SeeBeyondimplementation, for example, the company called inPricewaterhouseCooper's and Cap Gemini Ernst & Young, as well aspersonnel from the software vendor, for deep technical expertise.Since GM is heavily outsourced, Musser says the company will continueto depend on outside help to maintain and update its links because,like all other software, EAI does require maintenance.

TransUnion's Lombardo expects to keep the software in order with oneor two full-time employees. "There can be decreased cost based onpersonnel, but more commonly those costs are displaced rather thandecreased," says Gold-Bernstein. In other words, the folks who werehand-coding interfaces frequently get reassigned to maintain the EAIlinks. Same head count, different tasks.

Many companies like TransUnion are banking on connections to otherbusinesses to make EAI worthwhile. While the potential forintercompany payoff may be great, it takes extra efforts to make thoseconnections work. As Benner points out, data definitions usually varyfrom company to company or even within a single company. Drawing on hisdays at Fluor for an example, Benner says, "I may sell a pipe that'scertified for $10 and a noncertified version for $7. They havedifferent part numbers, but under the covers it's the same pipe. Thepart-number issue can drive people nuts. That's the bigger problemthan being technically able to pass data from one system to theother."

If a company has trouble defining a "part" or "customer" (and almostall do), imagine the disparity across company borders. EAI tools havethe capacity to address those problems through their ability totransform data as it moves from point to point. However, the legworkof describing the differences and training the EAI tool to map datacorrectly from one system to another is something that won't beautomated.

TransUnion is fortunate in this regard, so far. Lombardo didn't facedata mismatch problems to that extent in his first Vitria projectbecause the data from Axciom is simply packaged with TransUnion data.The two data sets are different, so there are no overlap or mismatchissues. However, data still requires extensive testing to ensure itsadherence to TransUnion's standards. Other companies probably won't beso fortunate and will have to carefully examine all data definitionsinvolved in EAI integration projects.