Top Ten Service Level Agreement Considerations

Here are our top ten considerations for authors and managers of service level agreements (SLAs):

1. MECE – SLAs should be Mutually Exclusive and Completely Exhaustive (MECE). By this, no two SLAs should measure the same thing, and there should be a SLA for every important aspect of the program. Too often we see SLAs that overlap, creating double jeopardy situations and misleading positive or negative performance reports. In addition, we often see important areas that are left unmeasured due to oversight.
2. Defined – Service levels must be thoroughly defined. For example, abandonment rate within a call center ought to be defined so that parties know whether blocked calls (trunk blockage), calls that terminate in the IVR, calls that terminate in the first 5 seconds (we don’t agree with this, but someone people do), or calls that answered by agents but are the “wrong number” are considered abandoned calls. Check this Q&A section out if you need some idea of how varied definitions are. If you don’t define the service level, you don’t know what you’re measuring.
3. Calculated – Even though you spend time defining service levels, you need to define any calculations. For example, if the service level is abandonment rate, the calculation should defined as “the number of Abandoned Calls divided by the total number of Offered Calls”. Where we use capitalized terms, these are predefined earlier in the SLA section. Vendors and vendor managers should never, ever be surprised by calculations. Give examples in your written SLAs.
4. Measurable – Don’t waste anyone’s time with SLAs that can’t be measured. For example, if you’re measuring customer satisfaction in a call center environment (typically done via a 3rd party after the primary call is concluded), but you don’t have the means to measure customer satisfaction (e.g., your call center doesn’t have the ability to use 3rd parties or your agents aren’t trained to collect satisfaction data), its a waste of time. Typically, angry customers will whip out unmeasurable SLAs and argue that vendors failed to achieve them, which is a huge waste of effort.
5. Easily Measurable – It’s one thing to measure something, its another thing to spend oodles of dollars to measure the same thing. If the cost of measurement doesn’t warrant the benefits of the SLA, don’t use it. The best example of this was a measure where customers call to complain about a bad transaction they’ve recently received in the mail. The vendor didn’t manage the call center, just the backoffice transactions. So, the call center needed to take notes on bad transactions and track them in a manner that allowed auditors to identify if the vendor was the cause of the bad transaction or not. Since the vendor only handled one of six steps in the backoffice process and the mainframe systems didn’t track transaction history, it was impossible to determine who caused the error without significant system modifications.
6. Time Frames – SLAs should cover a specific time period. Daily, weekly, monthly, quarterly, annually, etc. They should also only be assessed once. The example provided in #4 is a bad example of this, since customers could call to complain months after the transaction was completed, making it difficult to understand when to assess a month’s quality number. Essentially, the vendor would be in jeopardy forever, since a customer could complain at any time about a month – and every complaint would only lower the quality score, until the vendor had to pay penalties.
7. Singled Barrels – A SLA should contain only one measure, not two, three, or even four measures. If you’re SLA is “99% of transactions must meet quality standards and achieve customer satisfaction requirements” you need to track both conditions, which is nightmarishly difficult. In questionaire terminology, these are called double-barreled situations, and typically provide misleading or inaccurate pictures of operational performance.
8. Serve a Purpose – In some contracts, a minimum number of SLAs are required (to reduce the vendor’s risk, of course). That’s great when you need 3 or 4 SLAs, but what if you only need two SLAs and are therefore required to make-up another one or two to meet the contractual guidelines? These typically become “gimmes” and are a waste of time. Every SLA should serve a purpose.
9. Actionable – Every SLA should be capable of being influenced through actions of the vendor or the company. If the SLA can’t be influenced, don’t bother. A bad example may be measures of employee satisfaction with compensation in a HR outsourcing relationship, where the vendor has no control over compensation or employee /supervisor communication/training program. Since all employees will naturally dislike their compensation to some degree, the vendor has very little ability to create positive results.
10. Realistic – Look, we all want to be perfect, but those who belong to the cult of zero defects don’t understand contracting and real-life BPO and ITO. Achieving 100% of anything is simply unrealistic in most situations. Your goals can be aggressive (or evenly progressively more aggressive over time), but they should achievable.

We want to hear your comments! Let us know what you’re thinking by commenting below.

Related posts:

1. Another Tale from “When You Don’t Have Vendor Management Governance”