Why Data Context Matters for AI

Table of Contents

• Overview
• What we mean by “data context”
• How context helps the Data Intelligence Agent
• Concrete examples
• Checklist: What to add first
• Best practices

Overview

AI assistants are only as useful as the information they can reason about. For analytics and implementation guidance this means the agent needs structured, accurate context about what you track and how you name things. Adding Attributes, Events, and Views in Anamap gives the Data Intelligence Agent the vocabulary and structure it needs to make precise recommendations, generate queries, and spot gaps in your implementation.

What we mean by “data context”

Data context is the structured metadata about your tracking implementation:

• Attributes: named fields (e.g., user_tier, product_category, price) and their types
• Events: actions or occurrences you record (e.g., purchase, sign_up, view_item)
• Views: logical groupings of pages or screens where events occur

Together these describe what data exists, where it comes from, and how it should be interpreted.

How context helps the Data Intelligence Agent

1. Better understanding of domain: When the assistant knows that user_tier is a string with values free|pro|enterprise, it can answer "How many pro users converted last month?" without guessing column names.
2. Fewer ambiguous recommendations: Without context the agent may suggest broad, incorrect fixes. With context it can recommend exact attribute names, show example filter logic, or highlight missing events.
3. More targeted queries: The assistant can generate accurate analytics queries (GA4, Amplitude, SQL) using real attribute and event names, saving time and reducing iteration.
4. Improved troubleshooting: When a tracking gap exists, the agent can point to the exact view or event where instrumentation is missing.

Concrete examples

• Example 1 - Attribute clarity
  • Without context: "Filter by plan" is ambiguous.
  • With context: knowing the Attribute user_tier exists with values free, pro, enterprise allows the agent to build exact filters and suggest cohort definitions.
• Example 2 - Missing event detection
  • With Views and Events registered, the agent can detect that checkout_started exists on Cart View but checkout_completed has no matching event and recommend adding it.
• Example 3 - Better mapping suggestions
  • If you have a product_sku attribute and view_item event defined, the assistant can recommend attaching price and category to Views where product analytics are needed.

Checklist: What to add first

1. Add Attributes for your core user identifiers (e.g., customer_id, user_email).
2. Add Events for conversion actions (e.g., purchase, signup, add_to_cart).
3. Add Views for major page types (e.g., product page, cart page, checkout page).
4. Document common attribute values (enums) for categorical fields.

Best practices

• Use clear, consistent naming conventions (snake_case or camelCase). Keep names short and descriptive.
• Prefer explicit types (Text, Integer, Decimal, Boolean) and add validation where possible.
• Group related Attributes in Views so the agent can reason about where events happen.
• Keep your attribute descriptions concise. Explain the meaning and typical values.

If you’d like, use this page as a checklist when you open the "Improve your data context" menu. The agent will immediately benefit from any Attributes, Events, or Views you add.