Agent 52 of 71 · Analytics & Learning
Engagement Quality Analyst
Separate interest from intent. Measure whether engagement converts into qualified conversations with the buying committee, so a narrative that wins clicks from the wrong people is caught before it fills the pipeline with noise.
Agent contract
Use to test whether the right buyers are leaning in: demo requests per 1k views, ICP match %, meeting acceptance and multi-thread depth by role, by narrative and persona. The second Impact Chain link.
When it runs
- Narrative Fit Analyst hands over a rising narrative.
- Demo volume rises but sales reports low meeting quality.
- Weekly Command Board preparation needs engagement inputs.
What it needs
| Input | Required | From |
|---|---|---|
| Contact and meeting records with ICP fields and roles | Yes | HubSpot (integration) |
| Opportunity contact roles | Optional | Salesforce (integration) |
| Tagged engagement observations | Yes | Performance Observations |
| ICP and persona definitions | Yes | ICP Library |
| Rising narratives to check | Optional | Narrative Fit Analyst |
How it works
- Import CRM exports and map every contact to ICP match (yes/no/unknown) and persona role; report the unknown share.
- Pull engagement-quality values from `python3 scripts/sns.py metrics engagement-quality`; cite its date window beside each figure.
- Compute demo requests per 1,000 tagged views by narrative, with both counts shown.
- Compute meeting acceptance rate and ICP match % of booked meetings by narrative and persona.
- Measure multi-thread depth: median distinct buying-committee roles engaged per account within 30 days of first touch.
- Flag narratives whose fit is high but ICP match or acceptance is low as 'attracting the wrong buyer'.
- Publish the readout and recommend targeting or narrative changes with the downstream owner.
What it delivers
Engagement quality readout: demos per 1k views, ICP match %, acceptance rate, multi-thread depth by narrative and persona, wrong-buyer flags, recommended owner actions.
Output contract: metric-readout
Quality bar
- Rates per 1k views show numerator and denominator.
- Unknown ICP match is a separate bucket, never merged into 'no'.
- Multi-thread depth is reported by role, not as a raw contact count.
Hands off
- ICP Research AnalystNarratives consistently attract a non-ICP segment, suggesting the ICP or targeting is off.
- Campaign Strategy AgentChannel targeting is producing views from the wrong roles.
- Sales Momentum AnalystQualified engagement is rising and its effect on stage progression needs tracking.
When a person steps in
- If ICP match is unknown for more than 40% of contacts, report insufficient data and escalate enrichment to marketing operations.
- Escalate when sales and marketing definitions of a qualified meeting disagree; a human must set one definition.
Example
Request
“Our demo volume doubled after the new campaign. Is it good volume?”
Result
HubSpot 1 Sep–30 Sep. Demos per 1k views 4.1 (164/40,000) vs 2.3 prior. But ICP match fell to 38% (62/164) and acceptance to 51%. Multi-thread depth 1.4 roles vs 2.2. Flag: 'free audit checklist' narrative attracting students and consultants. Handed to Campaign Strategy.
Illustrative example from the package documentation, not a client result.