We all know the SLA is a method for monitoring and managing the performance of the service provider. It spells out quality measurements, measurement methods and what to do about failure. Because this is such an important area there may be a lot of complex contractual wording dealing with phasing in, excusable delays, setting initial values of new SLA, etc.
In principle the business requirement dictates the services to be provided, and the quality to which these must be delivered. The services can normally be approximated by taking what the incumbent (internal or external) does and then adjusting it to take into account of the objectives of the new contract. Of course this might be a huge adjustment but the services are something that can be seen or at least described.
The quality measures are not so simple to get right. Perhaps they are intended to serve too many objectives.
In the first case the SLA sets out one of the yardsticks against which the service provider will be judged. It therefore has to be very clear and unambiguous, as missing the SLA targets messes up the relationship and brings penalties.
Secondly, the SLA sets out what service the customer should expect. There is an exponential relationship between quality measures and cost; generally it will be more expensive for the service provider to deliver higher quality than was contracted for and priced. This is pretty obvious which is why it’s so strange that customers expect it to happen.
A very sophisticated outsource customer would understand what level of the quality the business really needs and would therefore accept the price/quality tradeoff implied, but many aren’t that sophisticated. Instead SLA targets are often set at levels that are affordable rather than acceptable to the business and might be too broad-brush. For instance, consider the effect of a 98 per cent availability or a four-hour repair time on small sales offices with only one printer, compared to a large office building. If you have both types of site, then think about different targets (or spare equipment). And the SLA job isn’t done until the business users have accepted the service levels and know what to expect.
Thirdly, the customer has to be happy if the SLAs are achieved. This means the customer has to be prepared to forgo arguments of the sort “we know the target is 98 per cent but we’re not happy with your achievement of 99 per cent because of course we really want it to be 99.5 per cent”. You sometimes see SLAs where (for example) the target is agreed to be 99 per cent but penalties are only incurred for performance below 98 per cent. Then there are running arguments because the customer sees 99 per cent as a commitment and the service provider sees it as aspirational and sees 98 per cent as the commitment. This is one of the areas in which negotiating teams may let down the delivery teams – the issue is too hard to deal with during negotiation without coming up against entrenched positions and changing the price, so it is fudged and results in five years of argument.
Fourthly, there is the continual debate over the level of detail at which the SLA is defined. Does the service provider commit to infrastructure or application metrics (runtime, availability, repair time) or to supporting a business outcome (order processing time, etc)? This is really a question not of SLA but of the overall intent of the contract, and then SLAs have to be set accordingly.
A business outcome is much easier to have a business conversation about and to set targets for. It will also give the service provider more leeway to innovate to improve compliance, and perhaps an incentive to do so. On the other hand if the business outcome targets aren’t entirely under control of the service provider then they are very hard to use for applying penalties such as service credits, damages, termination for cause etc. Those are important considerations. In my opinion, therefore, even in the case of a transformational outsource there must be SLA measures for which the service provider can clearly be held responsible. If that is too low a level to measure the full intent of the contract then another way has to be found to make the service meaningful to the business without damaging the ability to hold the supplier responsible for service quality.
A further twist is to consider what other purposes can be served by cleverly engineered SLAs. For instance, a risk analysis might identify the need for some mitigating actions that might be measured through the SLA. The desire for innovation might result in some additional measures being defined.
Fifthly, what about overdelivery? Does it have enough value that the customer should offer a bonus? The answer is probably "sometimes". Having systems available 100 per cent is surely better than not, but there might other cases where overdelivery is of little benefit.
In summary, these are SLA considerations for which there is no right but many wrongs. Whether you are a supplier or customer, you need to consider four cases: what incentives will the SLA give the supplier that will affect his behaviour when things are going badly? And when things go well - to improve the service further or to improve profit by reducing any quality over the minimum contracted? Similarly, what levers will the customer have when things go well, and when they go badly?