Data & Analytics Practical insights
Give your AI analyst a marketing metric dictionary
Define leads, pipeline and revenue before asking an AI analyst questions, with approved calculations, time rules and examples that prevent confusion.
‘How many leads did marketing generate?’ sounds precise until sales and marketing use different definitions of a lead. An AI analyst cannot resolve that disagreement reliably by guessing from database field names. Give it an approved metric reference before allowing natural-language questions to become management reports.
Specify the business object
For each measure, define what is counted, the qualifying event and the exclusions. A form submission, a unique person and a sales-accepted opportunity are different objects. State whether repeat submissions count again, whether employees and test records are excluded, and whether an opportunity can belong to more than one campaign. Include a few approved examples and counterexamples.
Make time and money explicit
Define the relevant timestamp: creation, qualification, close or invoice date. For revenue, state currency, tax treatment, refunds and whether the number describes bookings, recognized revenue or cash received. An AI answer that combines a lead cohort from one month with pipeline created in another may sound sensible while answering no coherent business question.
Connect definitions to approved calculations
Link each term to a maintained query or semantic model, with its owner and effective date. Clear tool descriptions and meaningful outputs help agents use connected systems appropriately. [1] Require reports to identify the metric version, filters, reporting period and data freshness. If a request conflicts with the dictionary, the analyst should ask for a definition rather than silently inventing one.
Test the difficult questions
Try ambiguous requests such as ‘best campaign’ or ‘new customer revenue’ and inspect how the system handles them. Compare its results with approved calculations, including refunds and repeated contacts. Begin with five decision-critical metrics, not an encyclopedic glossary. A small, trusted dictionary creates a practical contract between business owners, analysts and the AI system, and gives future disagreements a specific place to be resolved.
Sources and evidence
Sources checked on 4 October 2026. Proposed workflows and hypothetical examples are editorial analysis.
From insight to practice