Audience Insight Narrator
Outcome
Give content executives a defensible narrative of how a title or franchise is performing by linking engagement, cohort behaviour, campaign spend, and social signal into cited explanations that drive slate and marketing decisions.
| Portfolio attribute | Value |
|---|---|
| Primary persona | Content executive |
| Primary aggregate | AudienceNarrative |
| Cosmos DB containers | 1 |
| Grounding data | engagement, cohorts, campaigns, social signals |
| Development estimate | USD 60-220 monthly dev estimate |
| Runtime options | Python FastAPI and C# ASP.NET Core |
Productivity hypothesis
These measures reflect the workflow, evidence, decision boundary, and industry risk for this use case.
| Impact type | Measure | Conservative internal assumption | Customer measurement |
|---|---|---|---|
| Productivity | Executive analysis cycle time | 8-16% reduction | (baseline median analysis minutes - assisted median analysis minutes) / baseline median analysis minutes * 100 |
| Quality or risk | Evidence discovery time | 10-20% reduction | (baseline median evidence-search minutes - assisted median evidence-search minutes) / baseline median evidence-search minutes * 100 |
Important: Conservative internal pilot assumptions only; not public benchmarks, commitments, or expected customer results. Validate against a comparable baseline or matched control cohort with mandatory human review.
Rationale: Playback, campaign, social, game, and release evidence can be reconciled faster into decision-ready narratives.
Pilot guardrail: Cite executive claims and expose conflicting evidence and attribution limitations.
Customer validation plan
- Baseline the current workflow with representative users and tasks.
- Run the accelerator on the same task mix with human review retained.
- Measure executive analysis cycle time and evidence discovery time with the documented formulas.
- Replace internal assumptions with customer-observed results before a business case.
Why Azure Cosmos DB
This AI solution needs an operational data plane, not only a vector index:
- Hybrid vector and full-text retrieval combines semantic recall with exact identifiers, codes, names, and domain language.
- Hierarchical partition keys and metadata predicates keep tenant and domain boundaries in the normal query path.
- Operational JSON, source chunks, workflow and agent state, feedback, and audit records remain together for low-latency reads and governed writes.
- Change feed reacts to source and operational updates without polling.
- Autoscale throughput, configurable consistency, and multi-region distribution support variable demand and enterprise availability.
Domain workflow

| Stage | Actor | Input → output | Approval |
|---|---|---|---|
| 1. Select the title and reporting period | Content executive | Title or franchise and period of interest → AudienceNarrative shell with decision questions | Policy controlled |
| 2. Assemble engagement and cohort rollups | Metrics assembly service | Pre-aggregated engagement and cohort profiles → Metric pack scoped to the period | Policy controlled |
| 3. Separate marketing effect from organic demand | Attribution agent | Campaign flights and start curves → Attributed versus organic breakdown | Policy controlled |
| 4. Draft the explanatory narrative | Narrative agent | Metric pack, attribution, and social themes → Cited narrative with confidence and caveats | Policy controlled |
| 5. Executive review before slate decisions | Content executive and analytics lead | Draft narrative with cited metrics → Approved narrative informing renewal or marketing decisions | Human required |
Technical architecture

Cosmos DB data model
| Container | Hierarchical partition key | Documents | Search paths |
|---|---|---|---|
audienceNarratives | /tenantId, /titleFranchiseId | AudienceNarrative, EngagementMetricRollup, CohortBehaviourProfile, CampaignPerformanceRecord, SocialSignalDigest | /narrativeVectorEmbedding, /themeVectorEmbedding, /narrativeVector, /themeVector |
Query contract
| Query | Container | Scope | Purpose |
|---|---|---|---|
| Period performance assembly | audienceNarratives | single-partition | Read engagement, cohort, and campaign records for the title and period in one partition query. |
| Drop-off anomaly detection | audienceNarratives | single-partition | Find platforms where completion collapsed against the franchise benchmark. |
| Campaign efficiency ranking | audienceNarratives | single-partition | Rank campaign flights by attributed starts per unit of spend. |
| Sentiment shift watch | audienceNarratives | single-partition | Detect days where social volume spiked with negative sentiment. |
Python and C# services route each document type to its declared container and hierarchical partition key. ETag-based writes protect mutable aggregates, while append-only records and bounded cross-partition queries preserve scale.
Azure deployment

Clone the source repository, authenticate with Azure, and run:
Set-Location accelerators/media-entertainment-gaming/audience-insight-narrator
az login
./Deploy.ps1 -Runtime python -Location eastus2
Change python to csharp to deploy the .NET implementation.
Explore the implementation
- Launch signed-in demo
- Implementation article
- Domain data model
- Workflow and API contract
- Python application
- C# application
- Domain container Bicep
- Platform Bicep
- Cost and configuration
- Source repository