Get Serious About Service Levels


“Are you serious?” It’s a question we often feel like asking clients after they tell us they expect a service level of at least “four nines” from suppliers because that’s what they believe they are achieving. Yet many times, the client doesn’t have the historical performance information necessary to support their belief.

Let’s talk about nines. When someone says that they want “five nines” or “four nines” they typically mean they want availability or up-time of a system or application to be 99.999% or 99.99%, respectively. What does that equate to in downtime? Below is a chart that shows the allowable downtime under different levels of “nines” (assuming no scheduled downtime on a 30 day month):

Availability Level Downtime per Month
99.999% 26 seconds
99.99% 4 minutes, 19 seconds
99.9% 43 minutes, 12 seconds
99% 7 hours, 12 minutes

With certain exceptions (e.g., financial market systems, air traffic control systems, retail systems the day after Thanksgiving, etc.), it’s worth asking how many organizations would be seriously impacted if the unplanned downtime in month exceeded 4 minutes and 19 seconds for their critical systems. Clients all too often start at the four or five nines level and have to be convinced to consider a more rational (affordable) SLA.

There is a direct correlation between the supplier’s price and the SLAs required by the client, especially as it reaches the four and five nines level. Many suppliers will tie all sorts of provisions to providing four or five nine SLAs such that they effectively undermine the value of the SLA. Even so, suppliers will charge a premium for the risk associated with just the expectation that they achieve this level of perfection. Even a small move to 99.98% can help. While it doubles the amount of downtime, it’s still under 9 minutes a month.

Getting serious about SLAs also means getting the internal IT organization to address SLAs like you expect the supplier to. It’s interesting to see how many client organizations looking to outsource a critical IT function do not have their own effective “commercial grade” SLA measures, metrics and reporting. They’re absolutely sure they are delivering at a best-in-class level, but don’t have complete meaningful reporting to demonstrate it. How reliable is their level of confidence? Certainly not enough for the outsource suppliers: “show me the reports”.

Even if an IT function is currently insourced and there are no immediate plans to outsource it, CIOs should insist that a commercial grade SLA measurement and reporting structure is in place. If the function is ever targeted for outsourcing, this information is invaluable for preparing the RFP, obtaining an appropriate, properly priced solution and holding the supplier immediately accountable upon transition of services. If it’s never outsourced, the information will still be highly relevant to the proper management of the function. Peter Drucker said “What get’s measured, gets managed.” If the internal IT organization is not measuring performance consistent with industry best practices, there’s a good chance they’re not managing the function consistent with industry best practices.