I recently came across this interesting article on the poeticcode blog. In short, it's author claims that BI as service (SAAS-BI) does not make sense because of the difficulties with incremental data loading and higher network bandwidth requirements.
I do not want to dispute here the difficulty of the incremental data warehouse loading. It is difficult in certain situations. It is not that difficult in others where the nature of data help solving it in elegant ways. Moreover the incremental loads are necessary for many BI applications no matter if these are deployed in house or as a hosted service. This integration "baseline" (from the labor and cost standpoint) is the same for both in-house and SAAS. SAAS then removes the labor and costs associated with the data warehouse and analytical stack implementation/maintenance. The multi-tenancy leads to improved HW utilization. Only these two factors lead to huge savings and far better return of investment.
The Internet network connection bandwidth argument is ridiculous: "Not to mention, all the extra network bandwidth needed to encrypt and transfer the data from your data center to the SAAS-BI data center. That also means factoring your internet pipes for much more peak-bandwidth else your potential customers visiting your corporate website might have network problems and worse, you might lose sales and loyal customers."
This argument does not even apply to my naively and quickly implemented FAN (Family Area Network). :-)
Summary: It is certainly good to know about the data integration issues associated with the (SAAS) BI. You should not overlook them. However, I do not believe that these issues are the key decision points like cost of ownership, risk management, return on investment, implementation time etc.