A comment recently by Vinnie that Oracle is ‘going retro’ with its plethora of ‘on-premise’ apps, as opposed to SaaS (Software as a Service) apps, makes me wonder about the on-premise vs. Software as a Service (SaaS) debate as applied to enterprise software.
I can see why an app like FreeAgentCentral can be effective as a service, but how could that be applied for big apps that need significant customisation? In the space that my company operates – complex manufacturing and project ERP – many installations have unique customisations and need guarantee of high availability, neither of which I can see SaaS offering. But there is a bunch of commentators - Phil Wainwright, Nick Carr et al – who seem to think that SaaS is the way to go. One analyst group is also putting SaaS as the next wave of adoption for enterprise software.
In the on-premise vs. SaaS argument, Dennis made the point that this is ‘not an either or world’. Add to that coverage by Thomas of Hasso Plattner’s comments on the Lego analogy . Although this refers to the argument by the Software As Services (SOA) proponents that ”the ‘enterprise’ is a series of interconnected businesses that can be snapped together on demand”, it seems to me to be something that the SaaS followers also use. As a key founder of SAP, Plattner knows how enterprises work and has said that Lego is “not the model for corporate or enterprise software. Lego bricks are not the model for architectural models. No architect in the world uses Lego for models, and they are a few magnitudes simpler than enterprise software.”
SaaS apps for business tend to be focusing on specific requirements within certain functions. That is creating the old problem of ‘silos of information’ within those companies that have adopted SaaS. One reason why companies in the 1990s adopted ERP was to get away from those silos to a system that promised an integrated view of the organisation.
Since Gartner’s prediction in 2001/2 of the rise of ERP 2 (ERP between enterprises – ERP 1 was within enterprises), which still hasn’t happened, I am sceptical of analysts like the one on the Sandhills blog who forecasts mainstream adoption of SaaS in the next few years.
The key point to my mind is the magnitude of complexity of enterprise systems. So what do you do to handle complexity, you abstract it to smaller blocks or layers that can be more easily managed. What then would be the layers for an enterprise SaaS? Could they be technical (as in database, application, rules, presentation) or functional (payments, receipts, order management, inventory control, etc)?
When it comes to customisation, how would a vendor provide a highly configurable, multi-tenanted system where an implementer could also provide add-on applications for unique requirements? There are probably some very bright people who know this answer, but I haven’t heard from them.
I am not discounting that SaaS will have an impact on us enterprise software geeks. The guys at Microsoft are getting into SaaS for enterprise apps , and Thomas regularly comments about SAP’s excursions into SaaS. But I doubt that, outside the US and a few other places, we will see companies changing their on-premise apps for SaaS ones in the next few years. New enterprise systems require not only extensive technical work but also extensive change management, and organisations have been through a lot of that with traditional ERP in the recent times to want to go through it again soon.