The is a very good synopsis of the issues related to the SaaS model by Don Fornes titled “The Software as a Service Dilemma”. Don’s comments are related to the risks to the large incumbents such as SAP, Microsoft, and Oracle in transitioning to a SaaS model, largely drawing on the work of Clayton Christensen in The Innovator’s Dilemma. Don summarizes the challenge for the large incumbents as follows:
I have been in the enterprise software market, principally with supply chain management vendors, for 20 years so have lived through the rapid adoption of this technology. There is no doubt in my mind that the biggest disruption caused by SaaS is transitioning to a subscription model. This is so tied up with cash flow and maintenance, not to mention Wall Street expectations and methods of evaluating performance, that the transition to subscription pricing has to be handled extremely carefully.
We have only to witness the backlash from SAP and Oracle customers on their recent attempt to increase maintenance fees to over 25%, not to mention the consequences to the SAP management board. These companies are hooked on maintenance fees, not only the initial license fees.
According to their 10-K, for the 3 months ending Feb 28, 2010 Oracle reported $1,718M in “New software licenses” and $3,297M in “Software license updates and product support”. In other words, Oracle’s maintenance fees were nearly twice as much as their revenue from new licenses. How do you transition from that ratio? In their annual report for 2009, SAP reported €2,607M in “Software revenue” and €5,285M in “Support revenue”, which is very close to twice as much as their new license revenue.
The technology issues are related to cost and therefore more in the control of the software vendor. One can still deliver on-premise solutions, but use subscription pricing. Without a doubt, over the long term a transition must be made to web technologies so that the solution can be hosted.
While client/server technology facilitates hosting to some extent, as Ron points out there is a lot of time and cost associated with installing a client on everyone desk who needs to use the application. For solutions that are computationally heavy, I am not yet convinced that virtualization is not a suitable substitute for multi-tenancy.
The operating costs will be higher because of the need to deploy software updates on each of the instances, but this can be automated. Multi-tenancy has gained a strong foothold in computationally light applications such as CRM. I am not sure what would happen if an MRP run was thrown at a multi-tenant engine. Can the clock cycles be managed as effectively in a multitenant model as they are in a virtualization model?
We are seeing quite a bit of interest in SaaS from large multi-billion dollar companies. Some legacy customers have moved from on-premise to on-demand, and there is increasing interest in on-demand from prospects. As your graph shows, large enterprises are slower to adopt SaaS.