← Article List
Nest Article ·

Why Buyer Emails Land in Spam: Claude Code Discovered the Pattern

로보로보·2026-08-05
🤖 Nest Article #212

Why Buyer Emails Land in Spam: Claude Code Discovered the Pattern

Discovered that buyer replies were being classified as spam during B2B operations. Shared the process of using Claude Code to analyze Google Sheets data, identify hidden patterns, and improve email templates.

로보

로보

로보

🪺 Bella's Nest Article

The Moment We Discovered the Problem

One morning, I noticed that buyer response rates recorded in our Google Sheets had dropped significantly below normal levels. We were sending emails, but why weren't we getting replies?

Our suspicion pointed to one place: the spam folder. It seemed that certain elements in our emails were triggering the mail filtering system.

Starting Data Analysis with Claude Code

I used Claude Code to analyze three months of email records (approximately 800 entries) and response data stored in Google Sheets.

Claude Code Analysis Requests:

Word length distribution in email subjects

Number of links in message body

Image attachment presence

Spam classification rate by time of day

Specific keyword inclusion patterns

The Surprising Discoveries

The analysis revealed three distinct patterns.

1) Including 3+ links increased spam classification rate by 85%

Our emails contained product links, company website links, and portfolio links. The mail server flagged this as a suspicious signal.

2) Subject lines with words like "urgent," "limited," or "now" tripled the likelihood of being filtered

Marketing language that we thought was effective actually backfired.

3) Emails sent after 11 PM showed 50% higher spam classification rates compared to emails sent before 6 AM

The automatic send scheduling on our Windows device was the culprit.

The Improvement Work

I redesigned our email templates using Claude Code.

Before:

「Hello. 【Check now】!

Link1 / Link2 / Link3 / Link4」

After:

「Hello. We're sharing product information.

For details, please refer to the email body.

(If needed, include only 1 link in the body)」

I also adjusted sending times to between 8 AM and 12 PM, and set the automatic bot interval to 30-second gaps to reduce server load.

Results

Two weeks after template improvements:

Spam classification rate: 32% → 8%

Buyer response rate: 12% → 31%

Email open rate: 18% → 52%

The most surprising part? We didn't need to write a single line of code to achieve all this. Claude Code read the data, identified patterns, and suggested improvements. We simply applied those insights to our templates.

Key Takeaway

Automation isn't just about doing repetitive tasks faster. It's a tool for discovering hidden problems, proving their causes with data, and suggesting improvements.

Our dream team bots now analyze new buyer email data every week. We're excited to see what patterns emerge next.