Skip to content
English
Contact Support Customer portal
Contact us
  • There are no suggestions because the search field is empty.

March 2026 Releases

Executive Summary

March's headline delivery is real-time streaming for all AI features. Dashboard Insights, Report Insights, and the Dashboard Builder now stream analysis token-by-token via Server-Sent Events, eliminating timeout issues and showing live tool execution progress as the AI works. AI performance also improved dramatically — data preparation is now 400–500x faster through deterministic trimming, and the dashboard analyzer was refactored to a single-agent architecture for leaner, faster responses. On the user-facing side, a new Global Chart Type Preference lets you set Line or Bar as your default across all dashboards, and property-level PM benchmarking data now renders with its own green color series for clear visual separation. A critical data accuracy fix corrects duplicate OTA property counts.

 

NEW REPORTS & FEATURES

  • New: Global Default Chart Type Preference

    • You can now set a global default chart visualization type — Line, Bar, or system default — in Settings → Preferences. This preference cascades as a fallback across all dashboard charts. Explicit per-report chart type selections always take priority over the global default. The setting takes effect immediately across all open dashboards.

  • New: Real-Time Streaming for Dashboard Insights

    • Dashboard analysis responses now stream in real-time. You'll see markdown content appearing progressively, with live indicators showing which tool the AI is executing (e.g., "Fetching section data…", "Building action items…"). If streaming encounters an error, the system automatically falls back to the standard response. Analysis timeouts that previously occurred on complex dashboards are eliminated.

  • New: Real-Time Streaming for Report Insights

    • Report Insights analysis also streams in real-time with a live activity log showing completed and in-progress tool calls. You'll see reasoning steps, tool execution badges, and incremental content as the analysis progresses.

  • New: Real-Time Streaming for Dashboard Builder

    • The Dashboard Builder chat now streams AI responses in real-time with tool execution visibility. You can see which operations the builder is performing (searching reports, adding items, validating the dashboard) as they happen. A collapsible activity accordion shows the complete tool timeline after each response.

  • New: Template-First Dashboard Builder

    • The AI dashboard builder now checks for matching pre-built templates before manual construction. Five dashboard templates (Company Overview, Enhanced Benchmarking, Inventory Growth & Churn, Market Insights, Marketing Reports) are indexed for instant discovery. You also get smarter report search with category browsing, detailed report lookup, and improved KPI matching that understands common aliases (e.g., "revenue" matches "total_revenue", "unit_revenue", etc.).

  • New: Consolidated Dashboard Builder Preview

    • The builder's preview system now supports previewing dashboards while they're still being built — you no longer have to wait for the build to complete. The preview, save, and navigation flow has been unified into a single component.

  • New: Property-Level PM Data in Benchmarking

    • Benchmarking reports can now display property-level PM data separately from company-level rollups. When property filters are applied, the chart shows individual property data in green alongside the company aggregate, enabling direct property-vs-company comparison.

  • New: PM Unit Grouping Reports Now Show All Totals 

    • Six PM unit grouping reports — Property Location, Marketing Sources Aggregate, Custom, Booking Source, Areas, and Marketing Sources — now consistently display total rows for PM data sources. Three of these previously had totals hidden.

 

Reporting Data Accuracy Fixes

  • Fixed: OTA property counts inflated by duplicate records

    • OTA property breakdown reports (Airbnb, VRBO, Combined OTA) were counting every data row instead of unique properties, causing inflated property counts, bedroom totals, and sleep capacity numbers when duplicate property records existed. Reports now count distinct properties.

    • Why this matters: All OTA info cards, breakdown charts, and supply-over-time visualizations will now show accurate counts. Markets with duplicate property records will see numbers decrease to reflect the true property count.

  • Fixed: Inconsistent legend labels on Market and Hotel Benchmarking reports

    • Data source labels in benchmarking report legends were missing area/market context. Direct data sources now display as "Direct ({area})" and Hotel data sources as "Hotel ({area})", consistent with OTA provider labels.

     

Report & Dashboard Bug Fixes

  • Fixed: Default market changes required re-login to take effect

    • When you changed your default market in Preferences or User settings, the change didn't take effect until you logged out and back in. The system now properly clears all market-related caches when a default market changes, and refreshes the session immediately. This fix applies across all settings paths (Preferences, User management, and team-level changes) and works correctly in multi-instance deployments.

  • Fixed: Additional filters missing in template dashboard previews

    • When previewing template dashboards, reports with additional filter options (like data source selectors) weren't showing the filter configuration. Filter metadata is now embedded directly in the template preview response, eliminating an extra loading step.

  • Fixed: PDF/XLSX exports showed "Market" instead of "Direct" for DMO accounts

    • Scorecard and chart exports for Destination Data (DMO) accounts incorrectly labeled the Direct data source as "Market." The product type is now properly threaded through the export pipeline so labels render correctly.

       

AI & Assistant Features

  • Improved: AI Analysis 400–500x Faster Data Preparation

    • The AI report insights agent's data preparation step was migrated from an LLM-based agent (2–5 second latency) to a deterministic trimmer that runs in under 10 milliseconds. The trimmer handles all report types — charts, scorecards, tables, maps — with type-specific field extraction, stripping redundant raw values while preserving the formatted data the AI needs for analysis.

  • Improved: AI Dashboard Analyzer — Single-Agent Architecture

    • The dashboard analyzer was refactored from a multi-agent orchestration pattern to a single-agent architecture. Analysis happens inline instead of being delegated to a sub-agent, reducing LLM call overhead and enabling deterministic output validation (filler stripping, GUID removal, follow-up normalization) in code rather than via prompt engineering.

  • Improved: Intelligent Query Routing

    • The AI dashboard analyzer now classifies queries into three tiers. Broad overview queries use lightweight reasoning without knowledge base search; analytical queries (comparisons, trends, "why" questions) use deeper reasoning with full knowledge access; and simple factual queries use lightweight reasoning. This reduces cost and response time for routine questions.

  • Improved: Large Leaderboard Optimization

    • Unit leaderboard reports with hundreds of rows are now automatically trimmed to the top 20 and bottom 20 rows before AI analysis, significantly reducing token consumption while preserving the most analytically valuable data points at both extremes.

  • Improved: Dashboard Builder Catalog Validation

    • The AI dashboard builder now validates KPIs and report configurations against your actual report catalog before adding them to a dashboard. Reports are matched to their supported KPIs, date range anchors are checked for compatibility, and invalid configurations are blocked with clear error messages instead of producing broken dashboards.

  • New: Dashboard Builder Session Management

    • You can now delete and rename dashboard builder sessions. Session deletion cleanly removes all associated conversation history and draft artifacts. Session names can be updated from the sessions list.