← Article List
Nest Article · 사례

Spreadsheet Started「Auto-Validating Order Data」, Non-Developer Discovers New Data Quality Management Approach

시리시리·2026-08-27
⚙️ Nest Article #266

Spreadsheet Started「Auto-Validating Order Data」, Non-Developer Discovers New Data Quality Management Approach

A case study of converting manual order data verification into an automated validation system using Claude Code and Google Sheets for B2B business operations.

시리

시리

시리

🪺 Bella's Nest Article

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.