Product Analytics: PostHog vs Mixpanel vs Amplitude, Funnel Analysis, and A/B Testing

Product analytics answers the questions that database queries can't easily answer: Why are users dropping off at step 3 of onboarding? Which features are correlated with long-term retention? Does version B of the checkout flow convert better than version A?

Choosing the right tool and setting up a clean event schema are the two decisions that determine whether your analytics investment pays off.

---

Tool Comparison (2026)

For most SaaS startups: PostHog (self-hosted or cloud). It covers analytics, session recording, feature flags, and A/B testing in one platform — and the free tier handles most early-stage needs.

---

Event Schema Design

The most important analytics decision you'll make. A good schema is easy to query; a bad one creates a mountain of data that can't answer your questions.

Naming conventions:

Object-Action format: noun_verb (past tense)
  ✅ project_created
  ✅ member_invited  
  ✅ subscription_upgraded
  ✅ payment_failed
  
  ❌ create_project (imperative)
  ❌ ProjectCreated (PascalCase)
  ❌ project-created (hyphens — hard to query)

Properties standard:

// Every event should include these standard properties
interface BaseEventProperties {
  // User context
  userId: string;
  anonymousId?: string;  // Pre-login
  tenantId?: string;     // For B2B
  userPlan: 'free' | 'starter' | 'pro' | 'enterprise';

  // Session context
  sessionId: string;
  pageUrl: string;
  referrer?: string;

  // App context
  appVersion: string;
  platform: 'web' | 'ios' | 'android' | 'api';
}

Domain event schema:

// Define all events with their specific properties
const events = {
  project_created: {
    projectId: string;
    projectName: string;
    templateUsed?: string;
    memberCount: number;
    source: 'onboarding' | 'dashboard' | 'api';
  },

  member_invited: {
    projectId: string;
    inviteeEmail: string;  // Hashed before sending to analytics
    inviteeRole: 'admin' | 'editor' | 'viewer';
    inviteMethod: 'email' | 'link' | 'bulk_import';
  },

  feature_used: {
    featureName: string;
    featureCategory: string;
    // Avoid one "feature_used" event per feature —
    // creates hundreds of similar events.
    // Use this only for genuinely cross-cutting features.
  },

  subscription_upgraded: {
    previousPlan: string;
    newPlan: string;
    billingCycle: 'monthly' | 'annual';
    upgradeSource: 'paywall' | 'settings' | 'admin' | 'api';
    mrr_change_cents: number;
  },
};

---

PostHog Implementation

// lib/analytics.ts
import posthog from 'posthog-js';

// Initialize (client-side)
posthog.init(process.env.NEXT_PUBLIC_POSTHOG_KEY!, {
  api_host: process.env.NEXT_PUBLIC_POSTHOG_HOST ?? 'https://app.posthog.com',
  loaded: (ph) => {
    if (process.env.NODE_ENV === 'development') ph.opt_out_capturing();
  },
  capture_pageview: false,  // Handle manually for SPA
  persistence: 'localStorage',
  autocapture: false,  // Be intentional about what you track
});

// Wrapper with TypeScript types
type EventName = keyof typeof events;

export function trackEvent<T extends EventName>(
  name: T,
  properties: typeof events[T] & Partial<BaseEventProperties>
): void {
  posthog.capture(name, {
    ...properties,
    appVersion: process.env.NEXT_PUBLIC_APP_VERSION,
  });
}

// Identify user (call on login and on page load if authenticated)
export function identifyUser(user: {
  id: string;
  email: string;
  name: string;
  plan: string;
  tenantId: string;
  createdAt: string;
}): void {
  posthog.identify(user.id, {
    email: user.email,
    name: user.name,
    plan: user.plan,
    tenantId: user.tenantId,
    createdAt: user.createdAt,
  });
}

// Server-side tracking (for backend events)
import { PostHog } from 'posthog-node';

const serverAnalytics = new PostHog(process.env.POSTHOG_KEY!, {
  host: process.env.POSTHOG_HOST ?? 'https://app.posthog.com',
  flushAt: 20,
  flushInterval: 10000,
});

export async function trackServerEvent(
  userId: string,
  name: string,
  properties: Record<string, unknown>
): Promise<void> {
  serverAnalytics.capture({
    distinctId: userId,
    event: name,
    properties: {
      ...properties,
      $lib: 'posthog-node',
      appVersion: process.env.APP_VERSION,
    },
  });
}

---

Funnel Analysis

A funnel tracks how many users progress through a sequence of steps:

Signup funnel:
  signup_started     → 10,000 users
  email_verified     → 7,200 (72%)
  profile_completed  → 5,100 (51%)
  first_project_created → 2,800 (28%)  ← big drop-off here
  member_invited     → 1,400 (14%)
  activated          → 1,200 (12%)

Build funnels in PostHog:

1. Create Funnel insight
2. Add steps in order: signup_started → email_verified → project_created → member_invited
3. Set conversion window: 14 days (trial period)
4. Break down by property: userPlan, platform, country

The insight: A 72% email verification rate with only 28% project creation means the drop-off is between verification and first use — not at the top of the funnel. Focus onboarding improvements there.

---

Cohort Retention Analysis

Retention cohorts show what % of users who signed up in a given week are still active N weeks later:

-- Build retention cohort in PostgreSQL for custom analysis
WITH cohorts AS (
    SELECT
        DATE_TRUNC('week', MIN(created_at)) AS cohort_week,
        user_id
    FROM events
    WHERE event_type = 'user_signed_up'
    GROUP BY user_id
),
activity AS (
    SELECT DISTINCT
        c.user_id,
        c.cohort_week,
        DATE_TRUNC('week', e.created_at) AS activity_week,
        EXTRACT(EPOCH FROM (
            DATE_TRUNC('week', e.created_at) - c.cohort_week
        )) / 604800 AS weeks_since_signup  -- 604800 seconds in a week
    FROM cohorts c
    JOIN events e ON e.user_id = c.user_id
    WHERE e.event_type NOT IN ('page_viewed')  -- Exclude passive events
)
SELECT
    cohort_week,
    COUNT(DISTINCT user_id) AS cohort_size,
    COUNT(DISTINCT CASE WHEN weeks_since_signup = 1 THEN user_id END) AS week_1,
    COUNT(DISTINCT CASE WHEN weeks_since_signup = 2 THEN user_id END) AS week_2,
    COUNT(DISTINCT CASE WHEN weeks_since_signup = 4 THEN user_id END) AS week_4,
    COUNT(DISTINCT CASE WHEN weeks_since_signup = 8 THEN user_id END) AS week_8,
    COUNT(DISTINCT CASE WHEN weeks_since_signup = 12 THEN user_id END) AS week_12
FROM activity
WHERE cohort_week >= NOW() - INTERVAL '90 days'
GROUP BY cohort_week
ORDER BY cohort_week;

Industry benchmarks (B2B SaaS):

• Week 1 retention: 70–85% (good)
• Week 4 retention: 40–60% (good)
• Week 12 retention: 25–45% (good)
• D30 < 20%: churn problem — investigate

---

A/B Testing with Feature Flags

// PostHog feature flags for A/B testing
import { useFeatureFlagVariantKey } from 'posthog-js/react';

function PricingPage() {
  // Assigns user to variant — consistent across sessions
  const variant = useFeatureFlagVariantKey('pricing-page-test');

  return (
    <div>
      {variant === 'annual-default' ? (
        <PricingToggle defaultBilling="annual" />
      ) : (
        <PricingToggle defaultBilling="monthly" />
      )}
    </div>
  );
}

// Track the outcome event
function handleUpgrade(plan: string, billing: string) {
  trackEvent('subscription_upgraded', {
    newPlan: plan,
    billingCycle: billing as 'monthly' | 'annual',
    upgradeSource: 'pricing_page',
    mrr_change_cents: calculateMRR(plan, billing),
    // PostHog automatically associates this with the feature flag variant
  });
}

A/B test decision criteria:

• Minimum sample: 1,000+ users per variant before declaring results
• Statistical significance: p < 0.05 (PostHog calculates this)
• Practical significance: effect size matters (1% conversion lift may not be worth shipping complexity)
• Run for at least 2 weeks to capture weekly variation

---

The Viprasol Method

We set up product analytics infrastructure for SaaS products — event schema design, PostHog or Mixpanel implementation, funnel and retention dashboards, A/B testing frameworks, and data pipelines to your warehouse. Analytics infrastructure built correctly pays dividends in every future product decision.

→ Talk to our team about product analytics implementation.

---

You Might Also Like

• SaaS Metrics and KPIs — the metrics that connect analytics to business outcomes
• B2B SaaS Onboarding — using analytics to improve activation
• Feature Flags — A/B testing infrastructure
• SaaS Churn Reduction — using retention analysis to reduce churn
• Web Development Services — SaaS product development