Sign in →
Intelligence1 min read

Intelligence — Overview

See what is happening with usage, revenue, and unit economics. Dashboard, Analytics, COGS & Margins, Reports, Usage Intelligence, Ingestion, Data Exports — one section, seven lenses.

Updated 2026-06-30Suggest edits
Docs Intelligence Overview

Intelligence is the read side of the platform. Catalog, Pricing Studio, and the gateways feed it — and these seven pages are how you make sense of what came back. Executive snapshot, per-customer drill-down, gross-margin auditing, scheduled CFO reports, anomaly detection on usage, the ingestion pipeline that fuels it all, and the export path for everything else. Use the right lens for the question, not all of them at once.

What Intelligence covers#

Three buckets: operator visibility (Dashboard, Analytics, Usage Intelligence), finance visibility (COGS & Margins, Reports), and data plumbing (Ingestion, Data Exports). Each page is built around a specific question; the section index below maps page to question so you can jump straight to the right one.

Pages in this section#

Dashboard
Daily executive snapshot. Revenue hero, MRR bridge, attention strip, top customers, activity feed.
Open
Analytics
Per-customer + per-product drill-down. MRR & Cohorts, Customer Health Signals, Usage Anomalies, Pricing Intel.
Open
Reports
Scheduled exports, CFO-ready summaries, recurring CSV / PDF reports by email.
Open
Usage Intelligence
Anomaly Detection, Trend Analysis, Cohort Comparison — usage-side equivalents of Customer Health.
Open
Ingestion
The pipeline metering hits before it shows up anywhere else. Recent events, billable-unit health, quality, performance.
Open
Data Exports
On-demand + scheduled exports to S3, GCS, Azure Blob, Snowflake, BigQuery, Redshift, Databricks. Parquet / CSV / JSONL.
Open

Pick by question#

Almost every visit to this section is driven by one specific question. Find yours below and skip to the right page:

QuestionPageSurface
Are we growing or shrinking this month?DashboardRevenue hero + MRR bridge
Which customer drives the most MRR?AnalyticsMRR & Cohorts → top customers
Is any customer about to churn?AnalyticsCustomer Health Signals — weekly drop tile
Are we losing money on a customer?COGS & MarginsBy-customer tab + margin pill
Is one provider eating our margin?COGS & MarginsBy-provider donut + cost composition
Is there a usage anomaly worth investigating?Usage IntelligenceAnomaly Detection tab
Why is the dashboard zero on metric X?IngestionBillable Units tab — NO_DATA card
I need the raw rows in our warehouse.Data ExportsNew export → S3 / Snowflake / BigQuery
Email finance the same report every month.ReportsScheduled reports → CSV / PDF

Real-time vs. batch#

Some surfaces refresh on every page load; others are populated by overnight jobs. Knowing which is which saves a lot of "the number is wrong" tickets.

SurfaceRefresh cadenceBacking store
Dashboard → Live Revenue tiles60s polling, current-month bucketusage_events (PostgreSQL)
Subscription drawer → Current Spend60s polling, projection onlyusage_events (PostgreSQL)
Event Log (Operations)live SSE streamplatform_events (ClickHouse)
Analytics → Customer Health Signalsdaily detector, 7-day rolling windowweekly_drop_log (PostgreSQL)
COGS & Margins (most tiles)nightly revenue backfillcogs_events with backfilled revenue_usd
MRR & Cohortscomputed live from invoicesinvoices (PostgreSQL)
Usage Intelligence anomalieshourly scanusage_events aggregates
INFO
Live tiles label their math as a projection and footnote when any active subscription uses tier or allowance pricing (those are upper-bound estimates until the period closes). Invoiced numbers come from the bill-run pipeline and may differ.

Storage tiers under the hood#

Most reads go to PostgreSQL. High-cardinality time-series (event log, platform metrics, MCP tool invocations) tier into ClickHouse when CLICKHOUSE_ENABLED=true on analytics-service. The query router prefers ClickHouse and fails open to PostgreSQL — nothing ever shows "service unavailable" because the analytics tier is degraded; it just falls back.

Where the data physically lives

  • PostgreSQL — invoices, subscriptions, customers, products, billable units, COGS events, the source of truth for everything financial.
  • ClickHouse — platform_events, uptime_checks, health_metrics, usage_records, mcp_tool_invocations, storefront_conversion, api_endpoint_metrics. High-volume, append-only, batched in.
  • Redis — short-TTL caches (current-spend, public-status-page, live-revenue buckets). Cache misses fall through to the primary store.

Intelligence answers "what happened." If you're trying to change what happens, you want a sibling section:

  • Cost going up? Read COGS & Margins, then tighten policy in .
  • Customer churning? Analytics → Customer Health, then act in Storefront → Customers + Trials & Conversion.
  • Ingestion is broken? Operations → Event Log + Metering Health diagnose the failure mode.