How to Build an Anomaly Detection System for Business Metrics

This anomaly detection system for business metrics flags unusual patterns before they become problems. I use it to catch revenue drops, traffic spikes, and support surges in real time.

What You Need Before Starting

• Python 3.8+
• Data source API access
• pandas and matplotlib installed
• SMTP for email delivery

Step 1: Connect Data Sources

Set up API connections for anomaly detection.

import requests
from datetime import datetime

def fetch_data(api_config):
    results = {}
    for source in api_config:
        response = requests.get(source["url"], headers=source.get("headers", {}))
        if response.status_code == 200:
            results[source["name"]] = response.json()
    results["fetched_at"] = datetime.now().isoformat()
    return results

Step 2: Process and Calculate

Transform raw data into the metrics you need.

import pandas as pd

def calculate_metrics(raw_data):
    df = pd.DataFrame(raw_data)
    metrics = {
        "total": df["value"].sum(),
        "average": df["value"].mean(),
        "trend": df["value"].pct_change().tail(7).mean(),
        "period": datetime.now().strftime("%Y-%m-%d"),
    }
    return metrics

Step 3: Generate the Report

Build the report using your template.

from jinja2 import Template

REPORT = Template("""
<h2>{{ title }} - {{ date }}</h2>
<table>
{% for metric, value in metrics.items() %}
<tr><td>{{ metric }}</td><td>{{ value }}</td></tr>
{% endfor %}
</table>
""")

def build_report(metrics, title):
    return REPORT.render(title=title, date=datetime.now().strftime("%Y-%m-%d"), metrics=metrics)

Step 4: Add AI Commentary

Use Claude to explain what the numbers mean.

import anthropic

def add_narrative(metrics):
    client = anthropic.Anthropic()
    message = client.messages.create(
        model="claude-sonnet-4-20250514", max_tokens=500,
        messages=[{"role": "user",
            "content": f"Write a 3-sentence analysis of these metrics. Be specific.\n{json.dumps(metrics)}"}])
    return message.content[0].text

Step 5: Schedule Delivery

Automate report generation and distribution.

import smtplib
from email.mime.text import MIMEText

def send_report(html_content, recipients, subject):
    msg = MIMEText(html_content, "html")
    msg["Subject"] = subject
    for recipient in recipients:
        msg["To"] = recipient
        with smtplib.SMTP("smtp.gmail.com", 587) as server:
            server.starttls()
            server.login("[email protected]", "app-password")
            server.send_message(msg)

What to Build Next

Feed anomaly alerts into Slack so your team sees them immediately.

Related Reading

• How to Use AI for Anomaly Detection - ai anomaly detection business
• How to Build Automated Alerts That Actually Help - automated alerts that help
• Setting Up Automated Inventory Alerts - automated inventory alerts setup