The Problem
Every morning when orders arrived, the same issues kept repeating.
Contact fields left blank, quantity numbers suspicious, shipping addresses incomplete, duplicate orders mixed in.
Team members had to manually verify each order one by one, categorize problematic ones, then send confirmation emails to buyers. Roughly 3 hours per week were spent just on this verification task.
Before: Manual Verification Workflow
Order data received
↓
(Employee manual check)
↓
Note problem items
↓
Draft buyer confirmation email
↓
Wait for resolution
In this process, errors were often missed, and repeating the same verification work was inefficient.
After: Claude Code Auto-Validation System
When new orders are added to Google Sheets, Claude Code automatically performs the following in the background.
Step 1: Data Field Validation
Check required fields (contact, quantity, shipping address)
Validate number format (is quantity between 0-999999)
Validate email format (is it a valid email structure)
Validate date format (is shipping date valid)
Step 2: Logic Validation
Did same buyer place duplicate order within 1 hour
Is quantity 10x higher than typical order volume (typo suspected)
Does shipping address match buyer's previous transaction country
Step 3: Auto-Classification and Alerts
Valid orders → Check "approved" column
Warning level → Record reason in "needs review" column
Error level → Document error details in "rejected" column
Real Results
Time Saved:
• Manual verification: 3 hours per week
• Automated verification: instant (2 minutes in background)
• Freed up time: team now focuses on genuine relationship building
Error Reduction:
• Before: 3-5 problematic orders per month causing delays
• After: 0-1 missed errors per month (auto-filter catches almost everything)
Buyer Satisfaction:
• Confirmation emails arrive faster, shortening sales cycles
• System-generated precise error messages make corrections easier for buyers
Lessons Learned
1. Perfect automation doesn't exist
Initially we hoped automation would catch every error, but in reality it handles about 90%, with the remaining 10% requiring human intuition. For example, that gut feeling of "technically valid but something seems off" can't be automated.
2. Validation rules need constant adjustment
Initial rules aren't permanent. Every time a new buyer type appears or seasonal order patterns shift, rules need tweaking. It's worth setting aside monthly rule review time.
3. Windows device and Mac mini cooperation
When developing automation scripts using Claude Code on the Windows device and running actual automation on Mac mini, the two platforms collaborate beautifully through Google Sheets, each leveraging its strengths.
Next Steps
We're now planning to extend this validation system further, allowing valid orders to automatically proceed to invoice creation. Since data validation is now reliable, we're confident about taking the next automation step.