"How many users signed up today?"

Simple question. But when you're processing 5 million events per day, the answer isn't simple.

AnalyticsPro solves this: real-time dashboards that turn data chaos into actionable insights.

The Analytics Challenge

Most businesses use Google Analytics. It's free, but it's not built for product analytics:

• 24-48 hour delay in data processing
• Sampled data (not 100% accurate)
• Limited customization (you get what Google gives you)

What We Built Instead

Real-time analytics with custom metrics, no sampling, and sub-second query times.

Core Features

1. Event Tracking

   • Page views, button clicks, form submissions
   • Custom events (purchase_completed, trial_started)
   • User properties (plan, signup date, location)

2. Conversion Funnels

   • Landing page → Signup → Trial → Purchase
   • Drop-off rates at each step
   • Cohort analysis (users who signed up in Jan vs Feb)

3. Revenue Analytics

   • MRR (monthly recurring revenue)
   • Churn rate
   • Customer lifetime value (LTV)

The Technical Architecture

Data Collection

Client SDK (JavaScript/React Native):

analytics.track('button_clicked', {
  button_id: 'signup_cta',
  page: '/pricing',
  timestamp: Date.now()
});

Server SDK (Node.js):

analytics.identify(userId, {
  email: 'user@example.com',
  plan: 'pro',
  mrr: 99
});

Data Pipeline

1. Ingest: Events sent to AWS Kinesis (1M events/sec capacity)
2. Process: Lambda functions validate, enrich, aggregate
3. Store:
   • Hot data: Redis (last 24 hours)
   • Warm data: PostgreSQL (last 90 days)
   • Cold data: S3 (everything else)
4. Query: GraphQL API serves dashboard

Database Schema

Events table (partitioned by date):

CREATE TABLE events (
  id UUID PRIMARY KEY,
  user_id UUID,
  event_name VARCHAR(255),
  properties JSONB,
  timestamp TIMESTAMPTZ,
  INDEX idx_user_time (user_id, timestamp),
  INDEX idx_event_time (event_name, timestamp)
);

Aggregations table (pre-computed metrics):

CREATE TABLE daily_metrics (
  date DATE,
  metric_name VARCHAR(255),
  value NUMERIC,
  PRIMARY KEY (date, metric_name)
);

Performance Optimizations

1. Pre-Aggregation

Don't calculate "signups today" on every dashboard load. Pre-compute it:

// Runs every hour via cron
async function aggregateDailySignups() {
  const count = await db.query(`
    SELECT COUNT(*) FROM events
    WHERE event_name = 'user_signed_up'
    AND timestamp >= CURRENT_DATE
  `);
  await redis.set('signups_today', count);
}

2. Caching Strategy

• Redis for "dashboard opens" (1min TTL)
• CDN for historical reports (24hr TTL)
• Skip cache for custom date ranges

3. Sampling for Exploratory Queries

User exploring data? Sample 10% of events, instant results.
User exporting report? Process 100% of events, queue job.

Real-World Metrics

E-commerce client:

• 2.5M events/day
• 450K monthly active users
• Dashboard loads in 180ms
• Custom funnel queries in 800ms

The Dashboard

Built with React + Recharts:

• Line charts (DAU, MAU trends)
• Funnel visualizations
• Cohort retention heatmaps
• Revenue breakdowns

No-code query builder:

Show me [event: purchase_completed]
Where [user.plan = 'pro']
Grouped by [date]

Cost at Scale

AWS infrastructure for 5M events/day:

• Kinesis: $200/month
• Lambda: $150/month
• RDS (PostgreSQL): $300/month
• Redis: $100/month
• S3 (long-term storage): $50/month

Total: $800/month vs $2,000-5,000/month for Mixpanel/Amplitude

When to Build vs Buy

Buy (Mixpanel, Amplitude) if:

• You need it now
• Team < 10 people
• Budget > $2K/month

Build (AnalyticsPro) if:

• Custom metrics matter
• You want to own the data
• Long-term cost savings (break-even at ~18 months)

Lessons Learned

1. Start simple. Track 5 events, not 50.
2. Cache aggressively. Real-time doesn't mean re-compute every second.
3. Visualizations matter. Good charts > raw numbers.
4. Privacy first. Anonymize IP addresses, respect GDPR.

Need custom analytics for your app?
Explore AnalyticsPro