Multitenant, yes. Hosted, not necessarily
Noted Silicon Valley blogger and SAP emissary Jeff Nolan didn't enjoy sharing a panel about Web 2.0 in the enterprise with Salesforce.com's Adam Gross last week:
"All throughout the discussion he kept dinging me as a representative of SAP and insisting that nobody should install software. I find this position a little mystifying ... Here's the real problem that I see SFdC running into with their obsession with multi-tenant hosted software, it's really their religion and just like you can’t decide it's convenient to switch religions when the opportunity avows itself, they will not be able to offer anything other than what they are currently doing when customers ask them to."
Reading through the comments on Jeff's post,To center the debate on hosted versus installed is to miss the point quite a few other attendees were fed up with the way the panel descended into a SaaS vs software confrontation. I guess they were eager to hear more about the social software, data sharing and mashup aspects of Web 2.0, rather than listen to Salesforce.com claim it as an endorsement of their own SaaS model.
Even worse, this is the wrong argument, anyway. Or perhaps the right argument, but framed in the wrong terms. This later comment that Jeff added to the post highlights what's really at issue:
"Jeez, I missed a huge opportunity last night to remind Adam that iTunes is installed software when he said that buying software should be as easy as buying a song on iTunes."
Well if it comes to that, Salesforce.com is installed software too. Lots of it. Marc Benioff was quite open about it when I interviewed him in January: "We're not talking about small amounts of code ... We're talking about seven years of code." The difference is that customers don't have to install it. It's physically installed in sfdc's own data center.
So to center the debate on hosted versus installed is to miss the point. What matters is not where it's installed, it's how it's architected. That's why I distrust SaaS as a term, because it gives the wrong impression. It's not just applying a different delivery mechanism to the same old software.
Once you've got the right architecture, it doesn't really matter where it's installed, although there are issues around economies of scale, expertise and intellectual property that might persuade a vendor like salesforce.com that there would be more to lose than gain from offering an on-premises option.
The distinction that matters is multitenant. Because that's where all the other benefits flow from: the platform independence, the requirement to configure for each customer using metadata rather than custom coding, the focus on scalability and performance irrespective of the size of customer, the enforced abandonment of per-server licensing and maintenance as a business model.
It's not a matter of religion, it's competitive and commercial necessity. If you don't do these things, you won't survive as a purveyor of business automation. I think in its heart of heart, SAP realizes this, which is why it's migrating its customers to a multitenant business services architecture based on NetWeaver. The question mark hanging over SAP is, will it get to a truly multitenant architecture fast enough? Followed by the subsidiary question, how many of its customers by then will still choose to host their own software? And the final question, can its business model survive those customer decisions?
PS: If you're coming to SaaScon in San Francisco on September 25/26, I hope you'll join me on the second afternoon, when I'll be discussing this and other matters with Jeff Nolan along with Oracle's Rob Reid at the closing panel of SaaScon (disclosure: I'm on the advisory board of SaaScon but have no commercial interest in the conference. The organizers are contributing some of my travel and accommodation expenses).