Friday, June 20, 2008

SaaS - Multitenancy Debates

There has been a really interesting discussion started by Phil Wainwright on Why Multitenancy Matters for SaaS companies...the latest post with a response from the Intacct CTO, Aaron Harris, is very enlightening. Harris points out that many of these decisions on infrastructure are made both on the technical merits but also with a strong emphasis on the impact to the economics of the SaaS model. It is worth a read.

Phil in his earlier post walks through the various back-end definitions of what a SaaS multitenant solution could look like. If you look at the 15 SaaS portfolio companies of Hummer Winblad you can see that we have a pretty strong religion on the approach that is best...this includes Intacct, Omniture and Aria Systems. The approach outlined by Harris - few instances of the back end and every customer on the same instance of the code - is the one where we see the best patterns for success.

There was also a post by the constant instigator, our own Dave Rosenberg, CEO of Mulesource (one of our portfolio companies and a regular ZDNet blogger). He argues that customers dont care. At the risk of setting off a fiery debate with Dave I will respectfully add some thoughts...that customers care because of the derivative effects of good backend choices.

Why customers DO care...

Market Leadership

Market leadership in a category can be a derivative of the efficiency of the SaaS providers organization. If you read the response provided by Harris on their multitenancy choices it boils down to economics. With a lean and mean back end infrastructure the SaaS provider has less administrators, less mixed hardware, less outages, etc and an overall lower cost to provide the service - do the customers care?

I would argue they do - most companies want to partner with the best companies in their category. The derivative reasons are a safer vendor choice, less risk of having to change later, often better pricing and usually a stronger product development capability.


There is often a temptation for SaaS companies to bend to large customers who want separate notification and processes for upgrading code. If you play out this scenario for the SaaS provider you end up in a situation quickly where a few customers don't want the upgrade, or delay it by six months, or advocate for a different feature mix...and the small decision to allow different upgrade paths blossoms into a large subset of code versions across customers a few years out. I believe the saying is that if you let the nose of the camel in the tent and it wont be long before he is sitting beside you.

If you play out the scenario, every support call starts with the old software model question "what version are you running" and "was it Bob who knew how to deal with this"? This slows down support, diverts development resources to support and slows down the innovation at the company. Not good.

I don't expect all customers will start asking all these questions about how a SaaS company manages their back end infrastructure during the sales process...but I do expect them to make choices based on market leadership, support time, pricing, product development, etc.

Friday, June 13, 2008

SaaS Platform Wars

Chris Keene (CEO of our portfolio company WaveMaker) pointed me to a recent McKinsey report on the SaaS Platform Wars.

The reference, and link to the McKinsey report are on Chris' blog entry called SaaS Platforms For ISVs - Who Wins? from May 21st.

It was very thought provoking and gave our firm a basis for examining our work in SaaS. We have been very active SaaS investors with about 15 companies including Omniture, Employease and Aria Systems. The latter is a great example of the common services layer being delivered as a SaaS infrastructure layer as described as one of the core areas of the stack.

I think the internationalization piece related to SaaS is another core asset that platforms can provide worth noting. For instance Aria provides other SaaS companies with the ability to do business in many countries and currencies with one integration - to date they have done a billion transactions in over 150 countries. Without good SaaS platform components each vendor would have this complicated mess to untangle for each area that they want to do business - and it slows down the adoption.

The battlefront that I see as the one developing the most quickly in SaaS platforms is the one around cloud computing. We are seeing different approaches and different levels of the stack exposed by all the vendors…Amazon is a utility model (but proprietary queuing and other bits), Google forces Python and the persistent storage, Opsource wants to build lock-in through application peering, etc. We still don’t know where Microsoft, IBM and others will enter…it will be a dynamic market for a bit yet. We are investors in Elastra that provides a third party layer for working with cloud computing environments.