Outsourcing - made-to-measure SLAs
Service levels represent the performance required for supplier service delivery against which the performance of those external service providers can be measured. If they are not set out at the right level, this can result in additional cost and compromises over quality. The better the preliminary work, the smoother the partnership with the provider.
Outsourcing has firmly established itself as a core strategy to procure IT services, and that trend is sure to continue in the next few years. The benefits are well known: the supposedly greater process efficiency and economies of scale achievable by the relevant suppliers can create a situation from which both sides benefit – at least in theory. This can, however, quickly turn sour if service providers fail to meet the expectations placed on them, especially when they have set those measures for themselves. For the partnership to be successful, intensive work must be carried out in advance.
In an ideal world, IT service providers will tailor their services proactively to the client’s requirements, and will always do so flexibly, to a high standard and at a continually falling price. In this ideal world, however, clients also know what services they want, at what volume and at what level of quality, so that there is never any friction between customer and contractor, either during or after the end of the contract term. In all other more realistic worlds, it has become the norm to group service levels together in a Service Level Agreement (SLA) in order to determine how a service or a process is to be delivered.
Success from the SLA
As the basic building block of a service agreement, service level agreements should benefit both parties as they provide a foundation for entitlements and a disciplinary framework for their dealings with one another. While the SLA protects the provider against excessive claims by the client, on the other hand the client can monitor and control the service provider’s performance of the service. How the SLAs are formulated in advance of an agreement is thus a major criterion for success, as there will be constant scope for interpretation and hence for an escalation of disputes during the term of any agreement.
As each client will have different requirements, there can really be no such thing as an off-the-shelf service level. Service levels need to fit perfectly, and hence be tailored to the particular enterprise and its needs with the aim of giving the service provider sufficient latitude to do its work effectively. These basic requirements suggest that some partnerships are put under excessive strain by this balancing act.
SLA – OLA – UC
According to ITIL v3, a Service Level Agreement (SLA) is a contract (interface) between a provider and its client on the quality of the services provided. The service provider in turn agrees services provided by its internal departments by means of an Operational Level Agreement (OLA). If the service provider also commissions an external service provider within the SLA, it covers these requirements in an Underpinning Contract (UC).
In consulting practice it is common to find cases in which the client – perhaps under pressure of time – has adopted the service levels of the provider, ascribed the wrong priorities, or set the qualitative and quantitative hurdles for the service provider so low that they can easily be cleared. Service levels that deliver a wealth of key ratios over the years have often been defined, but the reports are not controlled because far too much information is provided. Furthermore there is frequently a lack of precision in the description of the responsibilities of the client and of the provider.
The key question for the client, though, is simple: What exactly do I want to achieve with this particular service level? Service levels that do not have a real benefit for the business side merely puff up administrative expense. This makes it absolutely essential for the IT and procurement team to sit down at a table with the business and determine which elements of the service are business-critical for the user and what the concrete effect might be of any failure. Few services and applications have to be delivered to the exact second; sometimes they are not even worth the administrative expense of a service level. If a client puts all its trust in the provider and the provider’s SLAs, and does not have a firm grip on the demand side of IT as a result, it will not be able to deliver a good job.
A frequently-used example of the content of service levels is availability, for instance of servers. What is important is the period (month, year) in which the measurement is carried out, whether it is done during operating hours, and whether fixed maintenance intervals need to be excluded from the calculations. These definitions within the SLAs are extremely important, yet often they are forgotten or not formulated precisely enough. Even small numbers can have big consequences, such as the maximum downtime of a particular process. An availability of 99.9 per cent works out at more than eight hours over the year – and it would be a disaster for a retailer, for example, if those eight hours fell on the Saturday before Christmas!
One client – a retailer – had an outsourcing contract with a supplier who was able to hide business-critical performance failure behind multiple operational service levels. The client did not care whether the problem was with hardware, software or communications – all it cared about was that its shops were unable to do business on a Saturday afternoon. After intervention from the Maturity team, the client and supplier agreed far more business-relevant service levels such as “retail hours lost at peak business periods as a result of IT failure”.
The needs of the business
When assessing the provider’s performance, however, it is not enough to concentrate solely on the technical side of the services. Clients should clarify in advance, for instance, what exactly “availability” means in an agreement. If just one of 16 processors in the server is working, the computer and the applications on it are still notionally available. But the user, who needs significantly longer to enter his data onto the system, is not interested in whether the problems lie in the server, the network or the application. In this case, we would recommend mapping the transaction time for the user of the service, rather than just the status of the hardware, in a comprehensive – or end-to-end – service level. This would make the service meet the needs of the business user much more closely.
It follows that specific definitions of the terms used, such as how the achieved availability is calculated or how the response time is explained, must likewise be recorded in each SLA, as there is no generally valid definition of these terms. Nevertheless, potential clients should strive for moderation: SLAs extending to more than 100 pages that cover all aspects of the service provided become unmanageable and, what is more, are not sufficiently flexible to allow adjustment to changes. In an ideal world, the service level always strikes a balance in this regard between detail, readability and flexibility.
Less can be more
The more service levels and key ratios an outsourcing agreement contains, the more expensive it will be for the client to control. Some years ago Forrester Research quantified the expense for the retained organisation at three to seven per cent of the contract volume. This may be perfectly true for smaller deals, when the retained organisation accesses traditional functions of the company (such as the legal department), where the costs arise elsewhere. In the case of large and fully equipped control groups (a sort of company within a company), all costs are visible because they impact directly on the retained organisation and hence the overall consumer. In this case, it is more realistic to put the expense for the retained organisation at up to ten per cent of the outsourcing volume.
If the service provider does, in fact, trip over one or more hurdles in the SLA, penalty clauses come into play. Contract penalties in the event that the promised performance is not delivered may seem attractive, but they are not necessarily recommended. Where, for the client, is the actual attraction of penalties? They are, in fact, a good way of exerting pressure in order to “punish” the service provider relatively easily for mistakes, and of putting the particular service manager in an unpleasant situation with his superior. But how does this really help the client?
One thing, though, must be clear: the service provider factors these potential penalties and the risk into its prices. If the service provider then delivers its work according to contract, it will charge a hidden “margin” for anticipated penalties. For every service level, the client must therefore ask whether triggering a penalty payment will cover its potential loss. This can only be assumed in the rarest of cases. Triggering a penalty is nowhere near as good as getting it right first time; and exercising a clause may not only fail to be cost-effective, but may also leave to a significant administrative burden to manage.
Special service levels for the end of the contract term protect the client if it does not wish to renew the contract; get-out clauses offer a special right of cancellation. The latter are used relatively often. Service levels by which the transition to another provider is to be safeguarded, however, are still the exception. It is useful to have a clause agreed at the start of the contract that imposes a duty of collaboration on the service provider in the event that the contract is cancelled. This clause must state beyond doubt whether the service provider is to be actively involved in the transition and who must meet any resulting costs.
To conclude, the driving forces behind service levels are not just control and penalty, but expense and benefit at the individual service level. It is counterproductive to receive 500 unimportant “key” ratios from the service provider every month when it is clear that these cannot be appropriately controlled and discussed with the provider. Two to three real headline key ratios which the service and the relationship manager can use as a basis for their interaction are normally sufficient for each service level. Ideally, the client will, after the contract has been running for a certain length of time, be able to identify which service level can change. For example, money might be saved by the business – with its full agreement and involvement – by declassifying a service level from silver to bronze. This only works where the client manages the provider through a few key targets. It is therefore in the client’s own interest to think long and hard in advance about how the flood of figures that the provider can generate at the press of a button can be channelled through meaningful service levels and KPIs that effectively support the business side of the organisation.
It is always helpful if, whether or not this is part of an outsourcing project, usable SLAs/OLAs are agreed between internal IT and its clients. Unfortunately, this does not always happen. Having professional agreements on service levels with colleagues is recommended as a protective measure: written agreements allow IT to demonstrate its quality and deliver counter-arguments to excessive demands of clients, so that the constant pressure for self-justification falls. In return, the understanding of in-house performance capabilities and cost structures rises if internal SLAs/OLAs have to be defined. This preliminary work will pay off in every outsourcing deal.