Assess performance

Canonical email assessment is input-only. Runtime retrieves aggregate provider metrics through Zeocore/ZEOconnect, Newsroom persists normalized observations and retrieval receipts, and Creator receives those artifacts in creator.assess_email_program@4.0.0.

Supply observations

Use AssessEmailProgramRequest with the exact EmailCampaignPlan and finalized EmailCampaignRelease, the expected normalized operation receipts, a tuple of EmailMetricObservation, the observation window, expected metric names and an explicit timestamp. The expected operation set and released messages define the coverage denominator; missing observations must remain partial. No connector service is required or consulted. The installed reference workflow demonstrates the complete acquisition-boundary simulation and capability invocation.

Each observation carries organization/publication, campaign revision, operation binding, retrieval-receipt reference and digest, provider definition/version, aggregate segment, window, completeness, coverage and attribution assumptions. It contains no subscriber identities or provider response body.

Supported names cover attempted, accepted, delivered, deferred, bounced, complained, unsubscribed, opened, unique link clicks, CTA conversions, sequence enrolments, step completion/exit, campaign conversion and attributed revenue. Open observations require reliability notes. Revenue requires an attribution policy. Rates require a nonzero denominator and an exact numerator/value relation.

Interpret conservatively

Partial observations remain partial. Missing expected metrics become data gaps. An empty input produces an unknown, limited-confidence assessment. Provider acceptance is not delivery, an open is not reliable evidence of attention, and correlation does not establish campaign lift.

Creator records hypotheses for future review. It does not select a winner, alter a live campaign, enlarge an audience, schedule work or activate a sequence. Assess each publication and campaign revision separately.

Legacy v1 compatibility

creator.assess_performance@1.0.0, MetricsQuery, MetricObservation and PerformanceAssessment retain their original schemas and behavior. That legacy capability uses creator.metrics_source; existing applications may supply an in-memory view of previously collected observations. It is not the canonical email-marketing path. Do not pass a live connector or ZEOconnect proxy into the new email capabilities.

The old credential-free example remains available as python -m zeo_creator.examples.assess_performance. New email integrations should run python -m zeo_creator.examples.email_marketing and use the input-only request above.