# Core Metrics Define the two to five numbers you run the business on, so every report and ticket is judged against them, and unlock the Core Metrics Drivers report. **Author:** Cogny Team **Published:** 2026-09-23 **Updated:** 2026-09-24 **Canonical:** https://cogny.com/docs/core-metrics Core metrics are the numbers you actually run the business on. Once defined, every report receives them as metric definitions, the home dashboard shows them, tickets are proposed against them, and the Core Metrics Drivers report becomes available. Without them, the agent optimises whatever the data makes visible, which for many sites is sessions and generic purchase events. ## Where Settings → AI Knowledge → **Metrics**. The page is called Metric Definitions. The toggle **Include metrics in AI reports** is on by default; leave it on. Each metric has: - **Metric name** and whether **higher is better**. - **Source**: BigQuery SQL against your warehouse, or a composed cross-integration query (mcpQL) when the number lives in an integration rather than in BigQuery. - **Definition** in plain language. Write this first; **Discover SQL from Definition** drafts the query from it, and **Run** shows the last 30 days and the last full month so you can sanity-check it. - **Target**: none, a single value, or a range. Use **Run All** after edits and **Save**. ## What to define Two to five metrics. Fewer than two gives the agent nothing to trade off; more than five dilutes attention. Good sets by business type: | Business | Metrics that work | |---|---| | E-commerce | Orders, revenue, blended cost of acquisition, share of orders from organic | | Lead generation | Form submissions, qualified leads (from the CRM), cost per qualified lead | | Subscription or app | Sign-ups, activated accounts, paid conversions, trial-to-paid rate | | Non-profit | Donations by type, monthly donor sign-ups, average gift | | B2B with long cycle | Meetings booked, pipeline created, named-account engagement | A reference non-profit workspace defines three: total purchase events from GA4 (with a definition text explaining which item categories mean which donation type), one-time gifts, and recurring donor sign-ups. The definition text matters as much as the query: it tells the agent what the number means and which segments exist. ## Rules for a good definition - **It must be tracked.** A metric the data does not contain is a tracking ticket, not a metric. Non-e-commerce sites often only have a purchase event configured; for lead generation, define a real lead event first (a form submission, a booked call) and get it firing in GA4. - **Name the event, not the concept.** "Sign-ups" is a concept. "GA4 event `sign_up` where method is not `test`" is a definition the agent can query and defend. - **Prefer BigQuery for anything from GA4.** The BigQuery export has event-level detail that the GA4 API rounds or samples. It needs the GA4 link to have been live for a few days. - **Add a target when you have one.** A cost-per-lead ceiling or a monthly revenue target changes the prioritisation of tickets. - **Wait about 14 days after introducing a new conversion event** before you make it the optimisation target of a paid campaign, so there is history to learn from. ## What core metrics unlock - The **Core Metrics Drivers** report (Executive & Strategy) decomposes each metric into its drivers, finds root causes for changes, and builds an experiment backlog sized to what would move the number. It is only recommended once at least one metric exists with the toggle on, and it sorts to the top of recommendations when it is. - Every other report receives the definitions in a METRIC DEFINITIONS block, so the SEO report talks about organic orders rather than clicks and the paid audit talks about cost per qualified lead rather than cost per click. - The home dashboard shows the current values. ## Keep them honest Re-run the metrics after any tracking change. A reference workspace lost visibility on its most important number for months because a tag was renamed and nobody re-checked the definition. The Data & Tracking Quality report catches most of these, but the person who owns the metric should look once a month. --- In this section: - Set Up Cogny Cloud for Success: https://cogny.com/docs/cloud-setup-guide - Before Your Onboarding: Checklist and Wizard: https://cogny.com/docs/cloud-onboarding-checklist - Connect Your Data Sources: https://cogny.com/docs/connect-your-data-sources - Context and Growth Strategy: https://cogny.com/docs/context-and-growth-strategy - **Core Metrics** (this page): https://cogny.com/docs/core-metrics - Reports and Templates: https://cogny.com/docs/reports-and-templates - Tickets and Approvals: https://cogny.com/docs/tickets-and-approvals - Automations and Content Loops: https://cogny.com/docs/automations-and-content-loops - Coding Agent, GitHub and Cogny Sites: https://cogny.com/docs/coding-agent-and-cogny-sites - Use Cogny from Your Own Agent: https://cogny.com/docs/bring-your-own-agent - Credits and Plans: https://cogny.com/docs/credits-and-plans - Team, Governance and Privacy: https://cogny.com/docs/team-governance-and-privacy - Cogny Cloud FAQ: https://cogny.com/docs/cloud-faq Source page: https://cogny.com/docs/core-metrics All documentation: https://cogny.com/docs Cogny for agents: https://cogny.com/llms.txt