What changes the metricIndependent model example
One metric. Two answers.
Five on-time responses can mean 71.4% or 62.5%. The difference is how you count the missing timestamp.
The same five on-time responses.
Eight fictional records. Results are rounded to one decimal place.
Known outcomes only
71.4%5 / 7 · unknown excludedUnknown counted as a miss
62.5%5 / 8 · all eight enquiriesThe same five on-time responses. A different number of included records.
- CHECK-014hOn time
- CHECK-0210hOn time
- CHECK-0322hOn time
- CHECK-0431hLate
- CHECK-056hOn time
- CHECK-0646hLate
- CHECK-078hOn time
- CHECK-08?Missing
Should a record with no timestamp count?
Of eight invented enquiries, five were answered within 24 hours and two later. The last has no response time. The responses stay the same; the calculation rule changes.
What if you find the missing timestamp?
Enter a hypothetical response time for CHECK-08. The limit stays at 24 hours; the original example above stays unchanged.
What changes the metricFind out why the timestamp is missing.
Exclude the unknown and the result is 5/7, or 71.4%. Count unknown as a miss and it becomes 5/8, or 62.5%. That does not establish a late response. First find out why the timestamp is missing.
Missing record
Was the response late, unrecorded or handled through another channel?
Measurement boundary
What starts and stops the clock? Do non-working hours count?
Check
Can you reproduce the result from source records and explain the exceptions?
Origin of this sample
A rework of the original standalone Evidence Note. All eight fictional records and both calculations are retained; no client data was used.
Limits of the conclusion
The model shows no client performance or measured improvement. The service owner must agree the definition.