One of the dangers of a lawyer blogging about technology issues is that you risk being told that your views represent a warped legal view of reality. So it's with trepidation that I stray into the vexed world of service level agreements (SLAs), where some might say lawyers don't belong.

In most ICT agreements, the SLA may be one of the most technically complex schedules. It's usually drafted exclusively by people with a deep understanding of the services and the manner of delivery. For that reason, it's often the most impenetrable.

And I'd argue that each of those features is wrong. To be workable, a good SLA has to be relevant, achievable and capable of relatively mechanistic application. And it needs to be understood by all stakeholders, not just the experts who drafted it.

The aim of an SLA should be to objectively define how well the service provider needs to perform and the immediate financial consequences of failure. It's easy to see the process of setting service levels as a scientific exercise: merely recording what should exist. But good service levels require art, not just science.

An SLA works at many different levels depending on the type of services, how integral they are to the business and what other functions depend on the services being measured. SLAs exist to reflect the user experience, act as a benchmark to measure service improvements, and operate as a fair way to tweak the price quickly to reflect service not delivered. The closer a service level scheme is to achieving those goals, the better it will be. I don't regard "compensating for loss" as a key goal of an SLA because that's not a achievable objective of an SLA.

As lawyers, one of the most basic question we usually get asked is whether service credits should be an exclusive remedy. I don't think that's even a valid question: I don't think of service levels or service credits as a remedy at all. Strong service levels or high service credits will not fix a poorly thought-out deal - so don't expect them to do so.

So what makes a good service level? It should be:

  • Capable of objective and reliable measurement.
  • Important to the customer's business objectives.
  • Prioritised for key business events or touch points to the outside world.
  • Realistic.
  • Part of the end-to-end service.
  • Specific – not a "goal" or a "target".

Prioritisation is particularly important. The SLA should focus on the aspects of service performance that are most important to the business. Customers ought to be realistic in setting service levels, and understand that the SLA is directly linked to supplier's cost base. The post-Lehman recession led to much better understanding of the cost consequence of SLAs. Customer "wish-lists" of redundant hot-stand-by systems or 24x7 availability cost money and may not be necessary. Try to negotiate good value from SLAs.

My main advice to clients is that SLAs aren't intended to cover all aspects of the services, and should not measure all things that can be measured. Not everything that's worth tracking should be linked to a financial penalty. Instead, focus on the most important parts of the services, and those parts capable of objective measurement. The commonest SLA mistake that I see is for SLAs to be too big and too ambitious.