AI Intelligence
AI growth analysis and Aria chat in Behavior Intelligence: Groq jobs, structured recommendations, run history, and analytics-grounded assistant conversations.
AI features overview
AI Intelligence bundles automated growth analysis jobs and the Aria assistant chat. Analysis jobs consume aggregated analytics, funnel summaries, SEO signals, and anomaly context to produce structured recommendations you can push to Growth Plan. Aria answers ad hoc questions in natural language with read-only grounding in your data.
AI does not replace dashboards—it accelerates interpretation. Always verify suggestions against raw reports before major business decisions.
Enabling AI
Requirements:
- Toggle
ai_analysis_enabledin Settings → Advanced. - Configure a Groq API key in Imaginit Core AI settings.
- Ensure sufficient historical data (typically 14+ days) for meaningful analysis.
Optional: set growth analysis schedule for automatic weekly runs.
Latest growth analysis run
Start a job manually or wait for schedule. Progress UI shows pipeline stages: context build, site behavior synthesis, opportunity ranking, validation. Output log displays structured sections—executive summary, top opportunities, funnel insights, SEO notes—with links to supporting pages.
Failed validation retries with stricter schema checks; persistent failures log errors without corrupting prior runs.
History and comparison
History lists past runs with timestamps and status. Select two runs to compare side-by-side—useful after implementing recommendations to see whether AI narrative shifted. Runs store hashed inputs metadata for reproducibility without storing full PII.
Aria assistant
Aria embeds in the app shell and Overview. Ask questions like "Why did organic drop last week?" or "Which pages leak mobile users?" Aria pulls AnalysisContextBuilder snapshots—metrics, anomalies, opportunities—not arbitrary SQL. Conversation history persists per user via ConversationStore.
Responsible use guidelines
- Do not paste customer PII into chat.
- Treat API costs as operational—schedule jobs off-peak if needed.
- Review AI-generated Growth Plan items before auto-assigning engineering work.
- Disable AI in regulated environments if policy requires on-prem-only analytics.
Prompting Aria effectively
Ask specific, data-bound questions: "Summarize organic mobile change last 14 days" beats "How is SEO?" Reference page URLs or funnel names Aria can ground. Follow up asking which dashboard page to open for verification.
Cost and scheduling
Scheduled growth jobs consume API tokens. Start monthly, increase frequency only if teams act on outputs. Disable AI on staging clones accidentally pointing at production keys.
Additional best practices
Behavior Intelligence works best when tracking coverage, funnel definitions, and team review rituals stay aligned. Revisit this guide after major site changes—new theme, consent plugin, or checkout provider—and confirm Realtime, Data Quality, and Tracking Plan still agree.
Share direct help links from the in-app help panel so teammates read the same canonical definitions. Press Ctrl+K in the help center to jump between related guides quickly.
- Align date ranges before comparing metrics across pages.
- Document segment names when sharing screenshots.
- Verify fixes in Realtime before closing Growth Plan items.
- Keep UTM conventions consistent for Marketing and Traffic.
- Schedule AI analysis after enough post-change data accumulates.
Working with the global toolbar
Every major Behavior Intelligence report page shares a global toolbar controlling date range, comparison mode, saved views, and active segments. Understanding this toolbar prevents the most common reporting mistake: discussing metrics that look contradictory because two pages use different ranges or filters.
Date ranges support presets (last 7, 28, 90 days) and custom intervals. Comparison mirrors the prior period of equal length. When comparison is off, KPI deltas hide—do not assume zero change. Saved views capture all toolbar state; teach teammates to check the view name in the header before screenshotting for leadership.
Connecting insights to action
Analytics value appears when insights change behavior. Behavior Intelligence routes findings through Growth Plan, Anomalies, and AI Intelligence into prioritized work. A metric without an owner becomes wallpaper. Assign DRI (directly responsible individual) per Growth Plan card and set verification dates.
Cross-link investigations: Traffic anomaly → Sessions sample → Journeys path → Tracking Plan coverage → Funnel step fix. This sequence resolves most conversion mysteries without external tools.
Data freshness expectations
Raw events ingest in near real time on Realtime. Daily aggregates power most dashboard KPIs and typically finalize after processor cron completes overnight in site timezone. Intraday numbers on Traffic are directional, not audit-final. Export via Report Builder after period close for finance-adjacent reporting.
Security and access control
Report data stays inside WordPress authenticated admin. REST endpoints require configured capabilities. Do not expose API keys or disable capability checks for convenience. View-only roles should not access Settings, Tracking Plan writes, or AI job triggers unless policy explicitly allows.
Module updates and compatibility
After Imaginit Core or Behavior Intelligence updates, revisit Getting Started verification steps: Realtime test, Data Quality health, one saved report run. Release notes may add event types or metrics—refresh Report Builder definitions if preview validation fails. Theme changes always trigger Tracking Plan re-crawl before trusting click metrics.
Keyboard shortcuts and help search
From any Behavior Intelligence admin page, open in-app help links to jump to the matching guide on the public help center. Within the help site, press Ctrl+K or Cmd+K to open Spotlight search across all articles, headings, and keywords. Search for metric names, page titles, or tasks like "export CSV" to find the right workflow without browsing the sidebar.
Bookmark frequently used guides for your role: content editors often need Content Performance and Organic Search; engineers need Tracking Plan and Data Quality; leadership prefers Overview and Reports email summaries.