Poc not customer usage

Active Enterprise PoC 30-Day Usage Skill

See config/methodology.md for the primary metric definition, cohort logic, and display configuration for this deployment. Metabase parameter bindings and warehouse table/column names are configured in config/data_sources.json (override the path via DATA_SOURCES_CONFIG).

Required Environment Variables

• GENERAL_HUBSPOT_APP_TOKEN or HUBSPOT_ACCESS_TOKEN — HubSpot private app token for company owner enrichment (prefers GENERAL_HUBSPOT_APP_TOKEN)
• METABASE_API_KEY — Metabase API key
• POC_BOT_SLACK_TOKEN — Slack bot OAuth token (xoxb-...)
• SLACK_CHANNEL — Target channel (default: #poc-bot)

Optional Environment Variables

• PRIMARY_METRIC_LABEL — display label for the primary metric (default: primary metric)
• PRIMARY_METRIC_EMOJI — Slack emoji for the primary metric (default: :bar_chart:)
• DATA_SOURCES_CONFIG — path to the data sources config file (default: config/data_sources.json inside the skill)
• DB_TEAMS_FACTS_TABLE / DB_SERVICE_AGREEMENTS_TABLE / DB_ENTERPRISE_DEALS_TABLE — override the warehouse table names from config/data_sources.json
• DB_DEALS_TEAM_ID_COLUMN — override the column in the deals table holding the team ID (default comes from config/data_sources.json)
• POC_PIPELINE_LABEL — pipeline_label value identifying PoC deals (default: poc pipeline)
• INTERNAL_TEAM_NAMES — comma-separated internal demo/test team names to exclude (empty by default)
• HUBSPOT_ENGINEER_COUNT_PROPERTIES — comma-separated HubSpot custom company property names used to prefer better-enriched company matches (empty by default)
• INTERNAL_EMAIL_DOMAINS — comma-separated domains whose admin emails are flagged as internal/test (empty by default)
• PERSONAL_EMAIL_DOMAINS — comma-separated personal email domains flagged for review (defaults to common providers)

Workflow

Step 1: Fetch Data

Run:

python3 .agents/skills/poc-not-customer-usage/scripts/fetch_poc_not_customer_usage_data.py

This writes /tmp/poc_not_customer_usage_data.json with one object per company in the active-enterprise PoC cohort.

The cohort logic is:

1. start from active enterprise service agreements (type/status filter values configured in config/data_sources.json)
2. require current_period_end >= current_date()
3. subtract teams with any non-PoC closed-won deal in the deals table (configured in config/data_sources.json; DB_ENTERPRISE_DEALS_TABLE overrides) using the same real-contract filter
4. exclude known internal/demo/test teams by name (via INTERNAL_TEAM_NAMES)
5. enrich the remaining teams with the latest PoC deal when available, but do not exclude teams just because the PoC pipeline record is missing

Each output object includes:

• company_name
• product_team_name
• team_id
• service_agreement_id
• service_agreement_end
• admin_email_domain
• team_size
• active_team_members
• members_using_primary_feature
• deal_name
• deal_id
• poc_pipeline_label
• poc_deal_stage
• poc_seats
• company_email_domain
• owner_name
• hubspot_owner_id
• classification
• review_flags
• date_range
• primary_metric_30d The primary metric is defined in config/methodology.md. The script intentionally does not reconstruct the broader customer-health score; it is specifically a usage view over still-open PoCs on active enterprise service agreements. It queries the configured primary metric tile on the Metabase dashboard (set via METABASE_DASHBOARD_ID, METABASE_PRIMARY_METRIC_DASHCARD_ID, METABASE_PRIMARY_METRIC_CARD_ID) with team_id and date_range parameter mappings only; the parameter targets are read from config/data_sources.json. If that metric query fails, the script exits rather than publishing a misleading zero-usage report.

Step 2: Build Report

Run:

python3 .agents/skills/poc-not-customer-usage/scripts/build_report.py

This reads /tmp/poc_not_customer_usage_data.json and writes /tmp/poc_not_customer_usage_report.json.

The report JSON has this shape:

{
  "text": "Active enterprise PoC 30-day <PRIMARY_METRIC_LABEL> usage summary",
  "main_blocks": [ ...Slack Block Kit JSON... ],
  "thread_replies": [],
  "run_summary": {
    "companies_analyzed": 0,
    "primary_metric_30d": 0,
    "top_company": "",
    "top_company_primary_metric_30d": 0
  }
}

Step 3: Post to Slack

Run:

python3 .agents/skills/poc-not-customer-usage/scripts/post_to_slack.py

This posts one top-level summary message.

Slack output expectations

The top-level summary should stay compact and scannable:

• report header with date
• quick summary of how many companies are in the cohort
• ranked list of all companies by 30-day primary metric in the same post
• each company should use the compact one-line format defined in config/methodology.md

Keep all sections readable in Slack. Use n/a for missing fields rather than leaving blanks.