Twill Brain

KPI Sync

2 min read

KPI Sync

Date: 2025-05-13 · Duration: 90.08000183105469 min Organizer: Nader Attendees: Ahmad, Mike, Nader

Summary

  • Filter Schema & Column Mapping: Front end uses ‘filter schema’ in widget JSON for filters. Column mappings managed per KPI/widget, separate JSON for clarity. Martin updating schema for team review. Decision made to maintain separate mappings for flexibility.
  • Widget Layout Flexibility: Minimal widget layout defined (rows/columns); optional UI layout allows front end to control rendering. Discussion on layout logic placement; standardization of widget sizes (small, medium, large, tall) prioritized for consistency.
  • KPI Granularity & User Interaction: Granularity variations (daily, weekly, monthly) supported; quarterly complexity deferred. Front end enables user-driven frequency changes. Prioritized KPI view adjustments for enhanced interactivity.
  • LLM Context & Memory: Plan for context JSON generation per KPI for LLM; guidelines on screenshots vs. JSON for insights. Long-term goal: persistent business context for proactive insights, nightly updates planned. Future alert system to trigger based on single-value KPIs.
  • Implementation Phases: Focus on SQL-based KPIs and LLM context; future phases include alerting and anomaly detection. Documentation of architecture and tasks for iterative development to be organized in analytics repo.
  • KPI Comparison Features: Exploring month-on-month/year-over-year comparisons in KPIs. Back end must support additional data needs for front end; feature definitions for comparison tools require refinement.
  • Action Items: Martin to push updates for

Action Items

  • Martin Elias Costa Push the branch with changes related to column mapping and filter schema updates for widgets, including frequency support if possible (20:20) Add a framework/example for generating a context JSON for each KPI, including all relevant metrics and findings, to be used for LLM context ingestion (01:25:45)

Nader Atrchin Review and revise the architecture and action plan for analytics, incorporating feedback from this meeting and adding alerting/notification system considerations (01:29:19)

Ahmadreza Abdoli Share print statements or examples of additional metrics calculated in Python implementations for KPIs, to help inform the context JSON structure (01:29:06)

Graph View