Just as a sampling, we’ve seen contract boilerplates that prevent measurement by:
- Not having an SLA or metric altogether
- Saying that there’s a certain target for SLA up-time, latency or other performance metric without specifying how it will be measured
- Explicitly stating that the SLA measurement will be left to the supplier’s discretion
- Measurement delayed to when it’s convenient for the supplier. For example, requiring the supplier to receive notification from the client and acknowledge that an outage exists before the clock starts ticking on downtime (in which case, acknowledgment can be timed to coincide with resolution).
- Measuring only sub-components of the supplier’s responsibility with the highest resiliency and leaving the items that actually take effort to maintain unmeasured (e.g., we guarantee that there won’t be a failure at the core power systems, but power distribution units (PDUs) are not covered).