The Problem
One morning, I opened the Google Form link on my Windows device. There were 1,200 new responses waiting.
I immediately spotted a pattern: the same email addresses appeared multiple times. One buyer had applied 5 times, another 3 times. Checking each one manually would take at least 2 weeks.
Before: The Manual Process
Time required: About 2 weeks
• Scroll through the email column in the Google Sheet
• Use CTRL+F to search for specific domains → count manually
• Compare application dates, products, and requests for each duplicate
• Transfer findings to a separate Excel file
• Final verification round
This approach had blind spots too. For example, when emails differed but company names matched (department-level applications), it was unclear how to categorize them.
After: The Claude Code Bot's Process
Time required: 3 hours (including bot development)
Step 1: Defining the Bot's Role
Input: 1,200 rows of Google Form responses
Filtering criteria:
• Detect duplicates by email address
• Group all entries with the same email
• Sort by application date
• Flag similar company names with notes
Output: Organized Google Sheet + duplicate summary sheet
Step 2: What Claude Code Executed
The bot automatically performed the following tasks.
1. **Grouping by Email Address**
- Record row numbers when duplicates are found
- Log duplicate count in a new column
2. **Analyzing Application Patterns**
- Flag multiple applications on the same date
- Check text similarity in requests (detect copy-paste)
3. **Adding Verification Columns**
- Mark as 「Has Duplicates」 or 「New」
- Assign group IDs (e.g., DUP001, DUP002)
4. **Auto-generating New Sheets**
- 「Duplicate Summary」 sheet: focused view of duplicates
- 「Needs Review」 sheet: different company names but similar email domains
Results
• **Duplicate groups identified**: 187 (about 15.6%)
• **Actual new applications**: 1,013
• **Similar company name cases**: 42 (flagged for review)
What We Did After That
After the bot finished, we only handled these tasks.
1. **Organize duplicate groups with no prior contact records** (about 30 minutes)
2. **Quick review of similar company name cases** (about 45 minutes)
3. **Final buyer list organization and sorting** (about 30 minutes)
Total processing time dropped from 2 weeks to 1 day.
What We Learned
Initially, the bot missed case sensitivity differences (e.g., info@company.com vs Info@Company.com). After adding string normalization to Claude Code, it was perfect.
Patterns that weren't visible when checking emails one by one became crystal clear. Things like sudden spikes in applications at certain times, or common requests from specific regions.
Today's Insight
For repetitive data comparison tasks, bots are far superior. Our job is simply to review the results and make decisions.