This is a great question as there is no easy answer. It depends on your level of expectation about the AA and it's clearly a value proposition.
There are issues to consider around the methodology (can you iterate it?), the data sourcing, the applications and the data model as well as the financial considerations. The main thing to remember is you will need to augment the solution to make it work AND to fit into the bigger DW strategy. Some of these AAs are no more than packaged PS (professional services) engagements. Most want to ignore the DW strategy, but don't let this happen. How the solution needs augmented depends on its shortcomings in all of these areas. A key concern of ours is the work/time/dependency involved in updating the data model and ETL rules to new vendor releases, especially if the client has done some customization themselves (which most will do).
The packaged approach will be attractive if the perceived delivery of the organization's DW team to deliver is low, when speed to market is paramount, when there are difficult, unconquered source systems including ERP systems, in large shops, when the applications are attractive and when there are no perceived issues with the architecture (see above).
This attractiveness will be negated if the organization is already measurably delivering high value with its DW program, the program has established that it serves a variety of applications well, it is a heterogeneous data source environment (as you mention), in smaller shops where the cost of the package is prohibitive, in non-ERP shops and when there are distinctive ETL, design and application needs.
The packaged approach is valid as a jump-start if its architecture is scalable and the shortcomings are known in advance, when the applications are relevant. We believe the packaged approach is here to stay, but DW is still something you build, not buy.
Come to my Keynote presentation on Packaged BI at the next Data Warehousing Institute World Conference in Orlando: http://www.dw-institute.com/conferences/orlando2001/keynotes.htm
This was first published in October 2001