SEO Testing Tool v1.3.0

Did your SEO change actually work — or was it just random fluctuation?

seo-testing-tool connects to Google Search Console, compares traffic before and after your change, and tells you whether the difference is statistically significant. Local SQLite database, zero configuration.

Vision

Reliably answer the question: "Did this SEO change likely improve or hurt performance on Google?"

The tool does not promise certainty, but supports informed decisions through rigorous statistical analysis.

Installation

npm install -g seo-testing-tool

Requires Node.js >= 18. On some platforms better-sqlite3 may require build tools (python, make).

The SQLite database is automatically created at ~/.seo-tool/data.db on first run. No configuration needed.

Getting Started

# 1. Configure Google OAuth2 credentials (interactive wizard)
seo-tool setup

# 2. Connect your Google Search Console account
seo-tool login

# 3. Create a new SEO test (interactive prompts)
seo-tool add

# 4. Fetch data from GSC and analyze
seo-tool run

Quick Start (Demo)

Want to try the tool without configuring Google Search Console?

seo-tool demo

Generates two experiments with 70 days of simulated metrics (realistic statistical noise via Box-Muller):

• Positive Experiment — +46% click increase in the post period (p-value < 0.01)
• Neutral Experiment — no significant difference between pre and post

Then explore the results:

seo-tool list
seo-tool status <ID>
seo-tool export <ID>

The ID is shown at the end of seo-tool demo. You can use just the first 8 characters.

CLI Commands

| Command | Description | | ---------------------- | -------------------------------------------------------------------------------- | | seo-tool setup | Interactive wizard to configure Google OAuth2 credentials | | seo-tool demo | Generate demo data (2 experiments, 70 simulated days — no Google account needed) | | seo-tool login | Connect Google account (OAuth2 for Search Console) | | seo-tool add | Create a new SEO test (interactive prompts) | | seo-tool list | Show table of all tests | | seo-tool status <id> | Test detail with statistical analysis and ASCII chart | | seo-tool run | Sync all active tests (fetch GSC + analyze) | | seo-tool export <id> | Export metrics to Excel or CSV | | seo-tool delete <id> | Delete a test (with interactive confirmation) |

All commands that accept <id> support partial IDs (e.g. seo-tool status abcd).

Examples

# Configure credentials
seo-tool setup

# Connect Google Search Console
seo-tool login

# Create a new test
seo-tool add

# List all tests
seo-tool list

# View detail with ASCII chart
seo-tool status abcd1234

# Sync data from GSC
seo-tool run

# Export to Excel
seo-tool export abcd1234 --format xlsx

# Export to CSV
seo-tool export abcd1234 --format csv

# Export (interactive prompt to choose format)
seo-tool export abcd1234

# Delete a test (asks for confirmation)
seo-tool delete abcd1234

Development (from source)

git clone https://github.com/svilupp0/SEO-Testing-Tool.git
cd SEO-Testing-Tool
npm install
cp .env.example .env

Environment Variables (.env)

GOOGLE_CLIENT_ID=xxx.apps.googleusercontent.com
GOOGLE_CLIENT_SECRET=xxx
GOOGLE_REDIRECT_URI=http://localhost:3000/auth/callback

# Optional: default ~/.seo-tool/data.db
# DATABASE_URL=file:/custom/path/data.db

# Optional: webhook for notifications (Slack, Discord, n8n, Zapier, etc.)
# NOTIFICATION_WEBHOOK_URL=https://hooks.slack.com/services/xxx/yyy/zzz

Development Commands

# Run commands without build
npx tsx src/cli.ts <command>

# Run tests
npm test

# Tests with UI
npm run test:ui

# Tests with coverage
npm run test:coverage

# Manual database setup (normally auto-migrates)
npm run db:push

# Verify DB connection
npm run smoke-test

# Lint
npm run lint

# Format
npm run format

Deploy on Railway (optional)

The project includes configuration for Railway with a nightly cron job and embedded SQLite database.

Railway Setup

1. Create a new project on Railway
2. Connect the Git repository
3. Railway will automatically detect railway.json and Dockerfile
4. Configure environment variables in the Railway panel (Google credentials, etc.)

Cron Job

The cron job (railway.json) runs npm run cli:run every night at 03:00 UTC:

• Fetches new data from Google Search Console
• Runs statistical analysis (Welch's t-test)
• Sends notifications if results are statistically significant

Manual Docker Build

npm run build
docker build -t seo-testing-tool .
docker run --env-file .env seo-testing-tool

Architecture

src/
  cli.ts                          CLI entry point (commander)
  cli/commands.ts                 Commands: setup, demo, login, add, list, status, run, export, delete
  cli/formatters.ts               Colors, tables, ASCII charts
  demo.ts                         Demo data generator
  index.ts                        Cron job entry point
  stats/
    StatisticalEngine.ts          Welch's t-test, outlier detection
    TDistribution.ts              t-distribution (p-value calculation)
  config/
    AnalysisConfig.ts             Configurable thresholds
    env.ts                        OAuth2 config
  database/
    db.ts                         Drizzle + SQLite connection
    schema.ts                     Database schema (Drizzle ORM)
    DatabaseService.ts            CRUD, transactions, aggregates, pagination
    TimeSeriesService.ts          Gap detection, recovery
  orchestrator/
    SEOExperimentOrchestrator.ts  GSC -> Analysis -> Results
  services/
    ExportService.ts              Excel/CSV export
drizzle/                          SQL migrations and metadata
railway.json                      Railway deploy config (cron 03:00 UTC)
Dockerfile                        Node.js 18+ container build

Minimum Traffic Requirements

The Welch's t-test needs enough data to detect real changes. As a guideline:

• ~50 click/day minimum (averaged over the monitoring period)
• 7+ days of data in both the "before" and "after" periods
• Sites with fewer than ~1,500 clicks/month are unlikely to produce significant results

The tool will warn you if data is insufficient, but won't prevent you from running the test.

Declared Limits (by design)

• Does not work on small sites (insufficient traffic for statistical significance)
• Not realtime (requires time to collect data)
• Does not guarantee results (provides probabilities, not certainties)

These limits increase trust in the tool.

Documentation

• PROGRESS.md — Implementation progress
• blueprint.md — Full technical specification

License

MIT

Version: 1.3.0 Last Updated: 2026-06-22