01 June 2009

BI Software as a Service

Of late, there is more and more talk about BI running in SaaS mode, which is Software as a Service. Let's look at the pro's and cons of this approach and where it makes sense to pursue this way of thinking.

How often do you worry about the water supply infrastructure in your home? What if the main water supply to your home is disrupted or needs an upgrade? No problem, you call the utility company and they fix it up for you with no cost and minimal bother to you. This is delivering water as a service whereby you are charged for the amount of water you use (and perhaps a small fixed equipment service charge too, if your utility provider is 100% transparent in their billing). Software as a Service (hereafter called SaaS) is very similar to this approach.

SaaS means that as an end user or customer, you only pay for what you use without having to worry about the infrastructure of how it gets to you.
This is all well and fine when it comes to using your household utilities, but running BI as a service has a few more considerations.

Firstly, there is data acquisition, as in getting the data from its source. Every organization I have dealt with in the past considers their data to be the most important thing in the world. Where the organization has already used the benefits of a hosted system or a hosted environment, the idea of having the information that drives BI going off to some cloud storage / computing device somewhere in the never-neverland of the WWW has some chance, though slim. The environments I have worked in have to comply to information privacy laws and cannot afford for any hosting vendor to say 'oops, we leaked the information, so sorry', or 'due to no fault of ours, some hackers got to the data and now it is plastered all over the internet, so sorry!'. The reality we deal with is that large organizations could be completely folded or shut down for breach of compliance laws. "There are various flavours of hosting vendors" some will say, however I will be the most surprised person to ever see one of our large financial institution customers ever agree to a situation where the data being extracted will be allowed to go to some off-site, not-controlled-by-them environment. Am I missing something in this picture? If so, please comment and let me know.


What we have offered that is closer to a service, is the ability to rent the licenses, so that the organization can add or remove licenses as they see fit. This way, the client organization will host the BI implementation at their premises on their machines, making it secure and ensuring that it conforms to all standard operating environment guidelines, and they bump the licensing requirements up or down based on their periodic need. Besides reducing the customer's risk to near-zero, this approach also helps our customers in that the software is always an expense, not an asset, so therefore does not land up in the balance sheet and does not need to be accounted for in terms of capital asset depreciation, which is a nebulous concept at best. It helps us in that most departmental managers have the authorization to approve the expense as an operational cost and that way no board-level decisions are required for approval, usually. The rental fee we charge also includes full support and all upgrades to future versions / releases (provided the rental fee is paid up to date).

Another trend that seems to be gaining some traction is the idea of a BI appliance. This is not new anymore as there are many organizations that use DW appliances, whether these are GreenPlum, Dataupia, Netezza (ok, not limited to DW here), the new (2009) DW applicance offered by Sybase to host Sybase IQ or the Oracle-Sun various appliances. This often solves the problem of large-organization beuracracy (where a $4,000 machine can land up costing $24,000 by the time all the prescribed packages and modules have been applied and tested to death, whether the server needs it or not) and is often based on the business wanting a plug-and-play device that involves minimal IT intervention. If this is sold to the organization as a service whereby they can rent (as a periodic expense) to the business the space and power that is needed, we're talking about a form of SaaS.

What has been your experience of BI implemented in a Saas fashion?

Comments welcome.