posthog-reference-architecture（Posthog 参考架构）

PostHog Reference Architecture

Overview

Production-grade architecture for PostHog analytics in a web application. Covers file structure, event taxonomy design, SDK initialization layers, feature flag management, group analytics for B2B, and data pipeline integration.

Prerequisites

• PostHog Cloud or self-hosted instance
• posthog-js and posthog-node SDKs
• Next.js or React application (patterns adapt to other frameworks)

Architecture

┌─────────────────────────────────────────────────────┐
│  Browser (posthog-js)                                │
│  $pageview, $autocapture, custom events, identify   │
│  Feature flag evaluation, session recordings         │
└────────────┬────────────────────────────────────────┘
             │ HTTPS (direct or reverse proxy)
             ▼
┌─────────────────────────────────────────────────────┐
│  PostHog Cloud (us.i.posthog.com)                    │
│  ┌──────────┐  ┌──────────┐  ┌───────────────────┐  │
│  │ Events   │  │ Feature  │  │ Session Replay    │  │
│  │ Pipeline │  │ Flags    │  │ & Recordings      │  │
│  └────┬─────┘  └────┬─────┘  └───────────────────┘  │
│       │              │                                │
│  ┌────┴──────────────┴────────────────────────────┐  │
│  │  Analytics: Trends, Funnels, Retention, Paths  │  │
│  │  HogQL (SQL), Dashboards, Cohorts              │  │
│  └────────────────────────────────────────────────┘  │
│  ┌────────────────────────────────────────────────┐  │
│  │  CDP: Destinations (Webhook, Slack, S3, etc.)  │  │
│  └────────────────────────────────────────────────┘  │
└─────────────────────────────────────────────────────┘
             ▲
             │ posthog-node (server events, local flag eval)
┌────────────┴────────────────────────────────────────┐
│  Backend (API routes, webhooks, crons)               │
│  Server-side capture, group identify, flag eval      │
└─────────────────────────────────────────────────────┘

Instructions

Step 1: Project File Structure

src/
├── analytics/
│   ├── posthog.ts           # Browser SDK init (singleton)
│   ├── posthog-server.ts    # Server SDK init (singleton)
│   ├── events.ts            # Typed event constants
│   ├── flags.ts             # Feature flag key constants
│   └── hooks/
│       ├── useFeatureFlag.ts    # React hook for boolean flags
│       └── useExperiment.ts     # React hook for A/B variants
├── app/
│   ├── providers.tsx        # PostHogProvider wrapper
│   └── layout.tsx           # Root layout with provider
└── lib/
    └── analytics.ts         # High-level tracking functions

Step 2: Event Taxonomy

// analytics/events.ts
// Naming convention: object_action (noun_verb)
export const EVENTS = {
  // User lifecycle (track conversion funnel)
  USER_SIGNED_UP: 'user_signed_up',
  USER_LOGGED_IN: 'user_logged_in',
  USER_ONBOARDING_COMPLETED: 'user_onboarding_completed',
  USER_INVITED_TEAMMATE: 'user_invited_teammate',

  // Core product (track feature adoption)
  FEATURE_USED: 'feature_used',           // with feature_name property
  ITEM_CREATED: 'item_created',
  ITEM_UPDATED: 'item_updated',
  ITEM_DELETED: 'item_deleted',
  SEARCH_PERFORMED: 'search_performed',
  EXPORT_COMPLETED: 'export_completed',

  // Revenue (track MRR and churn)
  SUBSCRIPTION_STARTED: 'subscription_started',
  SUBSCRIPTION_UPGRADED: 'subscription_upgraded',
  SUBSCRIPTION_DOWNGRADED: 'subscription_downgraded',
  SUBSCRIPTION_CANCELED: 'subscription_canceled',
  PAYMENT_COMPLETED: 'payment_completed',

  // Engagement (track stickiness)
  NOTIFICATION_CLICKED: 'notification_clicked',
  FEEDBACK_SUBMITTED: 'feedback_submitted',
} as const;

// Standard property schema
interface BaseProps {
  source?: 'web' | 'mobile' | 'api';
  plan?: 'free' | 'pro' | 'enterprise';
}

// Type-safe capture
type EventMap = {
  [EVENTS.USER_SIGNED_UP]: BaseProps & { method: 'email' | 'google' | 'github' };
  [EVENTS.FEATURE_USED]: BaseProps & { feature_name: string; duration_ms?: number };
  [EVENTS.SUBSCRIPTION_STARTED]: BaseProps & { plan: string; interval: 'monthly' | 'annual'; mrr: number };
};

Step 3: Feature Flag Constants

// analytics/flags.ts
export const FLAGS = {
  // Feature rollouts
  NEW_DASHBOARD: 'new-dashboard-v2',
  AI_SUMMARIZE: 'ai-summarize-beta',
  BULK_EXPORT: 'bulk-export',

  // Experiments
  PRICING_PAGE: 'pricing-page-experiment',
  ONBOARDING_FLOW: 'onboarding-flow-v3',
  CHECKOUT_LAYOUT: 'checkout-layout-test',
} as const;

// Flag → default value mapping (used when flags fail to load)
export const FLAG_DEFAULTS: Record<string, boolean | string> = {
  [FLAGS.NEW_DASHBOARD]: false,
  [FLAGS.AI_SUMMARIZE]: false,
  [FLAGS.PRICING_PAGE]: 'control',
  [FLAGS.ONBOARDING_FLOW]: 'control',
};

Step 4: High-Level Analytics Module

// lib/analytics.ts
import posthog from 'posthog-js';
import { getPostHogServer } from '../analytics/posthog-server';
import { EVENTS } from '../analytics/events';

// Client-side tracking
export function trackFeatureUsed(featureName: string, duration?: number) {
  posthog.capture(EVENTS.FEATURE_USED, {
    feature_name: featureName,
    duration_ms: duration,
    source: 'web',
  });
}

export function trackSignup(method: 'email' | 'google' | 'github') {
  posthog.capture(EVENTS.USER_SIGNED_UP, { method, source: 'web' });
}

export function identifyUser(userId: string, properties: {
  email: string;
  name: string;
  plan: string;
  companyId?: string;
  companyName?: string;
}) {
  posthog.identify(userId, {
    email: properties.email,
    name: properties.name,
    plan: properties.plan,
  });

  if (properties.companyId) {
    posthog.group('company', properties.companyId, {
      name: properties.companyName,
      plan: properties.plan,
    });
  }
}

// Server-side tracking
export function trackServerEvent(
  userId: string,
  event: string,
  properties?: Record<string, any>
) {
  const ph = getPostHogServer();
  ph.capture({
    distinctId: userId,
    event,
    properties: { ...properties, source: 'api' },
  });
}

Step 5: Data Pipeline Integration

// PostHog → External Systems via CDP Destinations
//
// PostHog Cloud Data Pipeline:
// 1. Events captured → PostHog stores in ClickHouse
// 2. CDP Destinations fire webhooks to your endpoints
// 3. HogQL queries available for custom analysis
//
// Common destination patterns:
// - PostHog → Webhook → Your API → CRM sync
// - PostHog → S3 export → Data warehouse
// - PostHog → Slack → Team notifications
// - PostHog → Webhook → Billing system (revenue events)

// Server route to receive PostHog CDP webhooks
export async function handlePostHogWebhook(event: string, payload: any) {
  switch (event) {
    case EVENTS.SUBSCRIPTION_STARTED:
      await syncToStripe(payload);
      break;
    case EVENTS.USER_SIGNED_UP:
      await syncToCRM(payload);
      await notifySlack(payload);
      break;
  }
}

Error Handling

Output

• Organized analytics module with typed events and flags
• Client and server SDK initialization (singleton pattern)
• Event taxonomy following object_action naming convention
• Feature flag constants with safe defaults
• Data pipeline integration via CDP webhooks

Resources

• PostHog Documentation
• PostHog JavaScript Web SDK
• PostHog Node.js SDK
• PostHog CDP Destinations
• PostHog Experiments Best Practices