Mission: Find Zava's Hidden Product-Quality Crisis

Mission/open goal Description

I selected the mission: "Find Zava's Hidden Product-Quality Crisis."

The goal was to investigate Zava's SQL Server 2025 PromptathonDb database and identify a product or category with meaningful sales impact but hidden customer dissatisfaction.

I analyzed sales, support, customer conversations, and document data to find a risk cluster where revenue, support burden, satisfaction scores, recurring complaints, and similar negative reviews converged.

The business question was:
Which Zava product should leadership prioritize for immediate quality investigation, and what evidence supports that decision?

The investigation used:

• SalesOrders and SalesOrderLines for sales impact.
• Products for product and category mapping.
• SupportTickets and SupportChats for customer pain signals.
• Docs for review/support evidence.
• Vector similarity search to confirm complaint clustering.

Harness and model

I used GitHub Copilot Chat in Agent mode inside GitHub Codespaces. The agent used SQL MCP tools to investigate the SQL Server database and analyze the PromptathonDb data. Model used: GitHub Copilot powered by GPT-5.5.

Turn-by-turn journey

1. Prompt:
   Explore the PromptathonDb database structure. List tables, row counts, primary keys, important columns, and relationships.

Agent response/action:
The agent inspected the SQL environment and identified the core entities:

• Products
• Customers
• Employees
• SalesOrders
• SalesOrderLines
• SupportTickets
• SupportChats
• Docs

Result:
I understood the database schema and identified the relationships needed for the investigation.

2. Prompt:
   Analyze sales performance using SalesOrders, SalesOrderLines, and Products. Find high-impact products by revenue, quantity sold, and order count.

Agent response/action:
The agent aggregated sales data and identified revenue and volume leaders.

Result:
The analysis showed that B2B products had high revenue, while Premium and Elite categories had stronger transaction volume.

3. Prompt:
   Analyze support impact by comparing products against SupportTickets and SupportChats.

Agent response/action:
The agent mapped support activity to products and calculated ticket counts, satisfaction scores, and priority distribution.

Result:
Premium Short Sleeve Men's Top (SKU: ZCPTM-SS-M-BW) was identified as the strongest risk candidate with:

• 9 support tickets
• 1.67 average satisfaction score
• Multiple high and critical priority cases

4. Prompt:
   Analyze complaint themes from SupportChats.MessagesJson and Docs for the identified SKU.

Agent response/action:
The agent extracted customer language and identified recurring themes.

Result:
The main issue was smart-fabric functionality failure:

• Smart features stopped working after washing.
• App pairing failed.
• Sensor/connectivity issues caused dissatisfaction.

5. Prompt:
   Select a negative document and run vector similarity search using find_similar_docs_by_doc_id.

Agent response/action:
The agent selected DocId 39:
"Smart fabric stopped connecting after one wash"

Result:
Vector search returned multiple highly similar reviews for the same SKU, confirming a repeated complaint cluster rather than isolated incidents.

Final decision:
Recommend engineering and quality investigation of Premium Short Sleeve Men's Top smart-fabric components.

Completion

• Yes, the agent completed the mission or goal.
• No, the agent did not complete the mission or goal.

Bonus work

Beyond the core mission, I documented the full investigation workflow and created an evidence-based executive brief.

Additional work completed:

• Compared sales impact against customer dissatisfaction instead of relying on a single metric.
• Used multiple data sources to validate the finding.
• Used vector similarity search to confirm complaint clustering.
• Identified limitations and suggested additional data that would improve root-cause analysis.

The issue documents a completed investigation of PromptathonDb using SalesOrders, SalesOrderLines, Products, SupportTickets, SupportChats, and Docs. Start by reviewing the reported SQL MCP workflow and DocId 39 similarity search; the documented result is an evidence-backed recommendation to investigate the Premium Short Sleeve Men's Top smart-fabric components.