99.95% Available, Seven Hours Down
Northwind Retail’s IT infrastructure team reports 99.95% server availability for the order-entry service and celebrates it monthly. Yet the order-entry service was unusable for seven hours last month. The outage was traced to a third-party payment gateway that IT does not monitor; the service desk logged 312 tickets but recorded only 40 incidents because agents closed duplicates and treated “gateway slow” as a known error. The CIO now asks for a “single dashboard” of IT performance and wants to use ticket volume as the main productivity KPI for the service desk.
1. Assess whether the 99.95% availability figure is a valid indicator of service performance. What is it actually measuring?
Read the worked answer
The 99.95% figure measures server availability inside IT’s monitoring boundary. It is not a measure of the order-entry service because:
- The gateway is outside the monitored estate even though it is inside the user’s value stream.
- Component availability is not service usability. Servers can be up while customers cannot order.
- Demand was hidden. The service desk recorded 40 incidents from 312 contacts by closing duplicates and labelling the issue a known error. That removed user pain from the reported picture.
The metric answers “were our servers up?” not “could customers order?” It is a component/output indicator, useful as a diagnostic, but invalid as the headline measure of service performance.
2. Design a small set of performance indicators for the order-entry service that would give the CIO a balanced view. Identify at least one outcome, one leading, and one experience indicator, and explain how they relate.
Read the worked answer
A balanced set could include:
- Outcome: order-entry service success rate - the percentage of attempted orders completed without error, measured end-to-end including the payment gateway. Alternatively, business-hours service availability as experienced by users.
- Leading: percentage of critical dependencies, including the third-party gateway, covered by active synthetic monitoring and alerting. Another leading indicator is the number of problem records past their target date.
- Experience: user-reported ease of ordering or satisfaction with the service when it is needed.
- Operational guardrail: incident and contact volume by service, used as a demand and health signal, not as a productivity target.
The leading indicator predicts whether the outcome is likely to hold; the outcome shows whether value is being delivered; the experience measure catches what telemetry misses. No single number should be used as the sole target.
3. What risks arise from using ticket volume as the service desk’s main productivity KPI? How would you mitigate them?
Read the worked answer
Ticket volume as a productivity KPI invites several behaviours:
- Splitting one issue into several tickets to inflate counts.
- Discouraging self-service and automation, because both reduce ticket volume.
- Closing tickets prematurely to keep the queue moving.
- Pushing work to other channels so it is not counted.
Mitigations include pairing volume with quality measures such as reopen rate, first-contact resolution, and backlog age; measuring demand rather than individual productivity; rewarding outcomes such as reduced repeat contacts; and auditing ticket classification. The CIO should see volume as a demand signal, not as proof of value.