SELF-SERVICE ANALYTICS PILOT WORKSHEET
Reviewed October 11, 2026. Blank editorial aid; not an AYD import or measured product report.
Use anonymized request identifiers. Keep private data in approved systems.

SCOPE (agree before collecting outcomes)
Team / owner:
Observation dates / comparison dates:
In-scope question families / eligibility rule:
Out-of-scope categories and escalation owners:
Approved reporting surface / grain / metric-definition version:
Access validation evidence / processing approval:
Definition of correctly resolved without analyst repair:
Acceptance thresholds / stop conditions:

ONE RECORD PER REQUEST (retain failures)
Request identifier / question family:
In scope? If not, reason and escalation route:
Source snapshot / refresh time / metric version:
Reference answer / reviewer:
Initial response reviewed correct? Evidence reference:
Resolved without analyst repair / after repair / unresolved:
Clarification required? Who supplied it?
Analyst repair minutes / setup effort recorded separately:
Elapsed time to reviewed answer (or unresolved):
Access or freshness issue:
Decision supported / next action:

AGGREGATE
All in-scope requests N:
Correctly resolved without analyst repair A:
Correctly resolved after analyst repair B:
Unresolved C:
Check A + B + C = N; assign each request exactly one final outcome.
Independent resolution = A / N; eventual resolution = (A + B) / N.
If N = 0, both rates are undefined. Do not report 0% or remove failures.
Out-of-scope count and reasons (separate):
Total analyst repair time / setup and maintenance time:
Time-to-reviewed-answer comparison for comparable tasks:
Repeat usage / access issues / freshness failures:
Decision: expand / repair definitions / retain reporting / stop:
Owner and next review date:

ILLUSTRATIVE ARITHMETIC ONLY
N=12, A=8, B=2, C=2: independent 66.7%, eventual 83.3% (one decimal).
8/10 excludes unresolved requests and is not the independent-resolution rate.
No customer outcome or AI benchmark is represented.