DEV Community

chaanli
chaanli

Posted on

I Lost $15K to Bot Traffic in One Month — Here's What I Built to Stop It

This is a real story. Last year I was running native ad campaigns spending $500/day. My ROAS looked great on paper — until I dug into the data.

The Discovery

I noticed something weird in my analytics:

  • 73% bounce rate on what should be a high-intent landing page
  • Average time on page: 2.3 seconds
  • Conversion rate dropped from 4.2% to 1.1% over 3 weeks

I ran a simple test — added mouse movement tracking to my landing page. The results were shocking: over 40% of my "visitors" had zero mouse movement events.

No mouse movement = no human = bot.

The $15K Math

  • Daily spend: $500
  • Bot traffic: ~40%
  • Wasted: $200/day × 75 days = $15,000

What I Built

I built a three-layer detection system:

Layer 1: IP Intelligence 

def score_ip(ip):
    score = 100
    if is_datacenter(ip): score -= 40
    if is_known_proxy(ip): score -= 30
    if request_frequency(ip) > 10: score -= 20
    return max(0, score)
Enter fullscreen mode Exit fullscreen mode

Layer 2: Browser Fingerprint Validation

Canvas, WebGL, AudioContext fingerprints vs known bot signatures.

Layer 3: Behavioral Analysis

Mouse movements, scroll patterns, click timing — fed into a simple ML model.

Results After Deploying

  • Bot detection rate: 97%
  • ROAS: 1.2x → 3.4x
  • Monthly savings: ~$6,000

I Open-Sourced It

If you spend >$100/day on ads without bot detection, you're burning 20-40% of your budget.

Top comments (0)