Skip to main content

Anti-Pattern: Embedded BI (Business Intelligence)

Description​

Embedded BI occurs when complex analytical reporting, aggregation, and historical data processing are implemented directly within the Transactional Monolith's application code, rather than in a dedicated Data Warehouse or BI Tool.

Instead of lightweight queries, the application performs heavy "Number Crunching" (Year-To-Date turnover, retention rates, complex funnels) in C# memory or via heavy SQL execution on the live OLTP database.

Evidence​

  • Module: WA (Work & Travel).
  • Files:
    • WAYTDFeeder.cs (36KB of calculation logic).
    • WADataFeederBiz.
  • Observations:
    • The system calculates "Year To Date" turnover by iterating over raw orders in real-time or near real-time.
    • Logic for "What counts as a sale" is hardcoded in C# if statements inside the loop.

Impact​

  1. Performance Degradation: Running a report locks database tables (Orders, Customers), slowing down the operational usage (e.g., someone trying to book a program).
  2. Inflexibility: Changing a KPI definition (e.g., "Exclude cancelled orders") requires a Code Change, Compilation, and Deployment.
  3. Memory Hogs: Loading thousands of orders into memory to compute a sum is an abuse of the Web Server's RAM.

Remediation​

  1. Extract to ELT/ETL: Move the raw data to a Data Warehouse (or a Read Replica) using a scheduled job.
  2. Use BI Tools: Connect Zoho Analytics (Strategic Choice) to the Data Source.
    • Let the BI engine do the heavy lifting (SUM, GROUP BY).
  3. Materialized Views: If real-time is needed, use SQL Indexed Views or a transparent caching layer, do not recompute in C# on every request.

2026 Audit Findings​

  • Occurrences: 1 Major Component confirmed.
  • Analysis: WAYTDFeeder.cs confirmed to contain heavy turnover calculation logic in C#.
  • Recommendation: Move to Zoho Analytics (Strategic) or Stored Procedure (Tactical).