Anatomy of a startup – PredictableIT Part I

Back in 2004 a friend had told me about moving a small VC firm and one of its portfolio companies to a hosted environment (at a Managed Service Provider, which was another of the portfolio companies).  Not just servers mind you, but rather they had locked down all their PCs (turning them into Thin Clients) and were using Microsoft’s Terminal Services offering to run all client software in the data center.  We quickly decided that it would be interesting to take this model more broadly and formed a company, PredictableIT to do so for Small Businesses.  Initially our thought was that this would be a modest-volume income producing business, not your every day attempt to change the world and create the next Microsoft 🙂  Basically we would take what he had done on a consulting basis and productize it.  We’d self-fund the operation and try a business model that all but eliminated fixed costs.  For example, we’d have no employees and rely on contractors and other third-parties for all services.  We’d use the Managed Services Provider, who would rent us machines, rather than own any hardware (other than some development/test systems).  We’d build a website and some tools to help with marketing, sign-up, and deployment.  Of course just as few battle plans survive the first engagement with the enemy so to do few initial business plans survive their engagement with the real world.  And so this is our story.

By early 2005 we’d started PredictableIT and done all the initial things one needs to do.  We were an LLC, had a bank account, we had made the first capital infusion, and found office space in an Executive Office setup.  More importantly, while during the first couple of months of operation my partner had remained in California he soon moved to Colorado so we could spend 60+ hours a week sharing an office and getting on each other’s nerves.  Just kidding about the “getting on each other’s nerves” part.  We’d worked together before and knew that we could handle so much togetherness.  Most importantly we’d convinced the two companies that were doing this on a consulting basis to transfer their custom solution to us and become our first customers, and so we were up and running pretty much on day one.  We even had arranged it so that they would continue to bear the full cost of the Managed Services Provider’s systems until we started offering the service to other customers.  At that point they would stop paying for the systems and start paying us based on our standard monthly per-user rate, which would greatly reduce their costs.  Not only that, but while we absorbed the overhead of supporting the current systems any custom work required was done on a Time and Materials basis.  We weren’t paying ourselves for this work so effectively the T&M work helped fund the company.  This would turn out to be a mixed blessing as the distraction it caused lead to delays in creating our real service offering.  But for the time being we had a very modest cost structure, with our biggest expense being paying a contractor to build our website.

The offering we were creating seems pretty simple at first glance.  Users would replace their PCs with Thin Clients (or lock down their PCs to operate as thin clients) and use Terminal Services to run client software on our servers.  We would also operate a Hosted Exchange mail system for them.  But as you look at this environment it quickly becomes apparent how complex it is.  For example, Exchange did not include an anti-SPAM system at the time and the third-party software available was both expensive and not geared to multi-tenant environments.  Moreover, even Exchange did not fully support multi-tenant environments at the time.  In fact, it was able to do so primarily via hacks that were extremely fragile.  So we had to put together and manage a SpamAssasin-based anti-SPAM system in front of Exchange.  Initially we manually provisioned Exchange for each new customer, but realized immediately that this was so fragile that we need to automate the process.  The same was true for the directory structure on our Terminal Servers.  Mis-configuration exposed one company’s data to another, and so automation was required to prevent operational error!  Likewise at the time Anti-Spyware was in its infancy and the Anti-Spyware and Anti-Virus software categories were separate.  Our Managed Service Provider included Symantec Anti-Virus software in their offering, but not any Anti-Spyware.  Even though Symantec added Anti-Spyware support in a newer version the Managed Service Provider didn’t upgrade.  Since we were the only customer who really cared about this capability they just prioritized the upgrade very low, so low that it didn’t occur during the life of PredictableIT.  The result was that I had to build (and operate) our own Anti-Spyware solution using a variety of half-solutions.  Why didn’t we just buy an existing Anti-Spyware solution?  Well now there is a business plan issue I’ll get to in Part III.  Not that there were many choices that clearly claimed support for Windows Terminal Services!  Indeed we did a lot of dealing with the fact that much third-party software didn’t work in a Terminal Services environment, something else I’ll cover in a future installment.  And keep in mind that back with Windows Server 2003 Terminal Services was an incomplete solution, with most users opting to license Citrix in order to complete it.  Even simple things like handling remote printers didn’t really work in Terminal Services!  And then there was the Managed Services Provider.  I already mentioned that they didn’t provide Anti-Spyware, but that is just where the weaknesses began.  Their patching policy, for example, was unacceptable for our environment.  They would take weeks after Microsoft released a patch to deploy it to their systems, and then they would apply only the most critical fixes.  Because we had lots of client users across multiple companies using our systems we needed them patched rapidly and completely.  I had to take on the testing and deployment of Microsoft (and other) patches rather than leaving it to the Managed Service Provider.

While the basic idea of the service was quite simple the details of making it work in a way that could easily be resold to many users, and be supportable in that environment, were actually daunting.  We inherited a Phase 0 operating environment out of the consulting project, then spent a great deal of effort creating the Phase 1 operating environment that we could use to take on and support other customers.  In fact, when I think about how I spent my time over the course of the PredictableIT experience I would say that over 50% was dedicated to creating and actually running the operating environment.  My title was President, but I now think I was more VP of IT Operations!  My partner was CEO, yet he too spent a lot of time on the operations side (e.g., he setup and maintained SpamAssasin and a few other things we ran on Linux while I handled the Windows systems).

Thinking back on other time commitments is pretty revealing overall.  Keep in mind that these things are bursty, so there were times when both my partner and I were 90% working on the business plan, or 90% working on testing, or 90% working on some marketing experiments, etc.  But on an average basis I was 50% operations, 10% T&M, 20% Program Manager, 10% Test, 5% Dev, and 5% everything else.  My partner took the brunt of the T&M work and I suspect it turned into 30% of his time.  The rest was probably 10% operations, 30% Program Manager, 10% Test, 20% everything else.    If you see problems with these percentages then so do I.  But that I’ll leave that for the end of the series.

Having real customers and real customer problems tought us a lot in those first few months.  We realized that in order to take on more than a very few additional customers we’d have to automate a lot of things.  And as we thought this through we also realized that once automated we’d be able to take on vast numbers of customers.  And as we plugged this into our business model and realized the potential economies of scale we got stars, or rather  Clouds, in our eyes.  In other words, the work required on our part to handle a few dozen clients was not much different from the work to be able to handle thousands of clients.  So we changed our focus.  We were going to go big.

In Part II I’ll explore the system and service we actually built, and in Part III the business aspects themselves and offer a retrospective on why we shuttered the company.


This entry was posted in Cloud, Computer and Internet and tagged , , , . Bookmark the permalink.

1 Response to Anatomy of a startup – PredictableIT Part I

  1. Pingback: Having a versatile Surface | Hal's (Im)Perfect Vision

Comments are closed.