Published: July 19 2011, 09:09 AM | no comments
by George Watt

Four caveats for those evaluating the costs and benefits of cloud solutions

Recent announcements ranging from mergers and acquisitions, to movement of key personnel, to changes in licensing models have certainly made it anything but a "lazy summer" (or winter, depending upon where you are) in the cloud space.  These changes have sparked even more interesting discussion regarding the costs and benefits of cloud solutions, and of the never-too-far-from-the-forefront debate related to the benefits and advantages of public versus private clouds.

Trap Door

As a service provider, an advisor, and in friendly debates, I have been witness (and participant) to many discussions (debates, slugfests) regarding which type of cloud is less expensive.  While these discussions have been, for the most part, either useful, productive, or simply fun (I really need to get out more often), there are a couple of things about these discussions that are cause for concern, especially in the context of the selection of a business service.  Too often I witness:

1. Loyalty to a specific style of service
2. Tendency for highest order analysis (especially when the debate is heated)

These two can be closely related, and each is potentially as damaging as the other.  Though to avoid falling victim to my second concern I believe further explanation is warranted.

It appears that as cloud services have evolved many people have developed a favorite model and enter these discussions with their decisions locked in.  At times, persuading an individual in this state that a different option might be better, or in some cases even to consider something different, is not possible; or at the very least requires significant effort - at times even when the facts are compelling.  Their position is often the result of a past success, or perceived success (more in a moment), with a previous service. 

I am not suggesting people should ignore past successes, nor am I suggesting they should not begin by looking at something that has worked for them.  I am simply suggesting that we need to keep an open mind, even in those cases.  Again I must admit that this statement might be another example of my firm grasp of things obvious.  Though there is more to consider.

You may have noticed that in the previous paragraph I used the phrase "perceived success".  That belies the second of my concerns, which can itself be a contributor to the first.  In many of the discussions I have witnessed, the full costs (and, in some cases, benefits) of the potential solutions are not considered in the analysis.  The thinking and debate remains at the highest order.  What can complicate this, and make it a more likely occurrence, is that it may often appear that there has been a deeper analysis when that is not the case. 

"Never assume the obvious is true" - William Safire

For example, consider a hypothetical analysis of a SaaS versus on-premise solution.  Debates of that nature often include statements such as: "I only paid $x for this and the internal chargeback would have been $X+ $0.10..."  Though that appears to be logical, complete, and compelling, often it is not because the analysis may have fallen victim to two assumptions:

1. The cost quoted accurately reflects the entire cost of the service
2. The solutions being discussed are exactly the same

These assumptions frequently slink their way into discussions of private versus public cloud solutions.  And either can result in decisions that can be disastrous to a business.

LAN That Time Forgot

Consider the first fallacy and how it might impact your business.  Suppose we are evaluating that hypothetical SaaS service and we learn that it can be delivered for $100/month/user versus $110/month/user for an in-house system.  If we assume the application serves 10,000 users, that may be a fairly compelling case; $100,000 less expensive each month.  Though what if we later learn that the on-premise service quote also included the costs of providing network infrastructure to the users (wiring, switches, power, telecom...). 

I believe that the fact that some on-premise infrastructure may be required, even for a full-featured SaaS service is one of the most frequently unaccounted-for cost elements of cloud service cost analyses.

In addition, what if the on-premise quote also included an allocation for the consumer's client machine (laptop, desktop...) and/or for an ELA for software required by the customer regardless of the solution chosen.  And what if these things accounted for $45 of the internal rate, or even $20?  Is the SaaS service still less expensive?  Finally, even if the on-premise solution is less expensive with all costs considered, does that mean it is the best solution for the business?  And that brings me to the second of the assumptions.

Think Beyond Service Features

I find that people often have these discussions with an implied premise that the solutions being discussed are the same, or similar enough to make discussing that unwarranted.  My concern is not related only to solution features, though they are often (always?) a consideration.  All services of a specific genre (e.g.:  payroll, email, CRM...) are not necessarily created equal.  Though, assume for a moment that in our fictitious evaluation of a SaaS versus on-premise solution they are.  There are many other things to consider such as maintenance, how quickly new features can be made available (agility), service resilience (quality, response, failover...), security, data location, and many others.  In addition, the case can become more compelling (in either direction) when items such as that are considered over a longer period (e.g.:  1 year, 3 years, 5 years, longer...).  And, as Robert Stroud highlighted in his recent post, what about your service desk?  Even an off-premise service can have an impact on an on-premise service desk.

Even Cities Get The Blues

Falling victim to the second assumption is hardly a new phenomenon.  I recall a case many years ago where a municipality had been receiving police services from a regional police force.  I suppose in this case, PaaS was "Policing as a Service"; though the similarities between this and cloud services selection are remarkable (at least in the context of this discussion).  It was time for the contract renewal.  The municipality decided to bring the service back under their own control, and not to renew the contract with the regional force.  (To take their "PaaS" back "on-premise".)  This turned out to be a very bad decision.

To make this decision they had considered the cost (to them) of a municipal officer versus their cost per officer from the regional force.  They had assumed they would get the same service "per officer."  Unfortunately, that was not the case.  The quote from the regional force was inclusive of all aspects of providing the police services.  It included vehicles, IT systems and equipment...  everything.  Not only was the municipality forced to pay for those aspects of the services, they were also faced with the significant start-up effort required in order to deliver those operational requirements, and with the expense that accompanied it.

"If you never change your mind, why have one?" -Edward de Bono

Regardless of our preferences - and it's fine to have preferences, and to use a preferred model ceteris paribus - there is no silver bullet solution.  Not in a public cloud, not in a private cloud, not in any cloud.  There is no substitute for good measurement and analysis, at least commensurate with the level of strategic importance, business value, and cost of the solution under consideration.

I'm sure people get lucky when making decisions of this nature.  I've also seen cases where, for example, a team that was publicly celebrating how inexpensive their solution was learned it could have been provided through a different service delivery model at a fraction of the total cost.  And we need to remember that it's not only about cost.  There may be times when selecting the more expensive service is better for the business, and there are some organizations such as the Cloud Service Measurement Initiative Consortium who are working to provide us with standard ways of enabling us to make useful, consistent, standard evaluations of those.

In summary, I have seen:

• (Sycophantic) loyalty to a specific model
• Analysis at only the highest level
• Incomplete analysis of costs (and/or benefits); and
• Assumption of uniformity of solution completeness

These are some of the most common things I encounter that jeopardize the quality of cloud related business decisions. What are some of the things you encounter that, if left unchecked, could have a negative impact to a business or that, if done, will result in a more positive outcome?

*Image used under Creative Commons License courtesy of tsmall.

This blog is cross-posted at Pragmatic Cloud. Follow @GeorgeDWatt on Twitter.