· analytics / posthog / amplitude

PostHog vs Amplitude: Product Analytics Guide 2026

PostHog is the default for solo teams: free 1M events/month, transparent pricing. Amplitude's real cost averages $64K/year. Here's the full breakdown.

By

2,301 words · 12 min read

For most solo founders and small dev teams: start with PostHog. The free tier covers 1 million events per month, 5,000 session recordings, and unlimited team members — all resetting monthly, no credit card required. The pay-as-you-go pricing is transparent enough to predict your bill before signing up. Amplitude’s published $49/month plan sounds competitive until you look at what teams actually pay once they’re in contract.

Switch to Amplitude if you specifically need its cohort analysis depth or enterprise ML features, and only after you’ve confirmed your monthly tracked user count is stable and predictable. If you’re not yet sure what an MTU is, you’re not ready for Amplitude’s pricing model.

Who this is for

Solo founders and small dev teams with a $0–$500/month analytics budget who’ve outgrown free GA4 and are choosing their first real product analytics stack. If you’re at Series B with a sales-negotiated Amplitude contract, the pricing analysis here doesn’t apply to your situation.

What we compared

PostHog Cloud is continuously deployed with no versioned release cadence — see posthog.com/changelog for recent updates. Amplitude is SaaS with no versioned release cadence. Pricing data pulled directly from posthog.com/pricing and amplitude.com/pricing on May 26, 2026. Real-world Amplitude contract data from Vendr’s deal database (372 deals).

Feature matrix

FeaturePostHogAmplitude
Free events/month1,000,000~2,000,000 (10K MTU cap)
Free session replay5,000 web recordingsIncluded; quota undisclosed
Free feature flags1,000,000 requestsUnlimited
Free seatsUnlimitedUnlimited
Free data retention1 yearNot published
Session replay AIPostHog AI (2K credits/mo)Built-in AI summaries
LLM observabilityYes (100K events free/mo)No
Feature flags standaloneYesYes
Advanced cohort analysisBasicPlus and above
Predictive analytics / MLNoGrowth / Enterprise only
Group Analytics (B2B)YesEnterprise only
Self-hostingLimited (Docker Compose only)No
Open sourceYesNo
Pricing modelEvent-based, pay-as-you-goMTU-based
Transparent pricingYesHeadline only — see below

Pricing: what you’ll actually pay

PostHog

PostHog has two modes: free (forever, no card, 1-project limit) and pay-as-you-go (same monthly free allowances, metered billing above them). Both reset monthly.

After free allowances:

ProductPrice
Product analytics$0.00005/event
Session replay (web)$0.0050/recording (5K–15K), declining to $0.0015 at 500K+
Feature flags$0.0001/request
Surveys$0.10/response

At 5 million events/month — a medium SaaS with roughly 2,000–3,000 daily active users — that’s around $200/month. PostHog’s pricing calculator gives you an exact number before you enter a payment method. No sales call required for any tier.

Amplitude

Amplitude’s $49/month Plus plan is the number that shows up in every comparison article. Here’s what that number actually covers.

Plus caps you at 300,000 monthly tracked users (MTUs). Viral traffic spikes — which you’re supposed to be measuring — get billed at 20–50% above your contracted per-MTU rate. Standard contracts include an 8% automatic annual price uplift (negotiable — multi-year terms often reduce it to 3–5% or remove it entirely). And the features most teams actually want gate above Plus: behavioral cohorts require Plus, Feature Experimentation requires Growth, Group Analytics for B2B requires Enterprise.

Real-world data from 372 Amplitude contracts tracked by Vendr:

MTU bandTypical annual cost
50K–150K MTUs$30,000–$70,000/year
250K–750K MTUs$70,000–$150,000/year
1M+ MTUs$150,000–$250,000+/year

Median actual contract across all 372 deals: $64,724/year.

This isn’t Amplitude hiding anything — the Plus plan at $49/month is technically available. But the structural traps (MTU overage billing, annual uplift, add-on gating) mean the headline price has essentially no predictive value for your actual bill once you’re past early traction. Name this explicitly in any internal evaluation.

Data model

This matters more than any feature table will tell you.

PostHog uses a flat event model. Call capture(), set whatever properties you want, query it later. No required taxonomy. Fast to get running for teams still iterating on their event schema.

Amplitude uses a structured taxonomy: events, user properties (persisted across sessions until changed), event properties (per-event only, not persisted), and group properties (account-level attributes for B2B analysis). The structure is a feature — it’s what makes Amplitude’s cohort and funnel analysis deeper than PostHog’s.

Two Amplitude constraints to know before committing:

  1. Group property changes apply only to new incoming data. They don’t retroactively update historical records. Restructure your B2B account taxonomy after a year of collection and historical queries won’t reflect it.
  2. Max unique group types per project: 5. Max group values per event: 10.

For early-stage teams still iterating on instrumentation, PostHog’s flat model is faster to get running. For B2B SaaS that needs account-level analysis from day one, Amplitude’s group structure is worth the setup overhead — but Group Analytics is Enterprise-only, so budget accordingly.

Session replay

Amplitude has closed the gap with PostHog here more than most comparisons acknowledge.

PostHog gives you 5,000 web recordings/month free (explicit cap, resets monthly). Paid recordings start at $0.0050/recording, declining to $0.0015 at 500K+/month. Retention on the paid Cloud plan is 3 months; free Cloud retention is not clearly documented. Platform coverage is wide: 20+ SDKs including iOS, Android, Flutter, React Native, Webflow, and Framer. Privacy defaults mask input fields; password inputs are always masked.

Amplitude includes session replay in all tiers, including the free Starter plan. The free monthly recording quota is not published — documentation references 2.5M sessions in a hypothetical but doesn’t state an actual free limit. If you need a hard number for budget planning, you won’t find one in Amplitude’s docs.

Amplitude’s differentiator: AI summaries are built into the replay viewer and surface friction points and user sentiment per session without you having to watch recordings manually. PostHog offers AI analysis via PostHog AI (2,000 free credits/month), but it’s invoked separately rather than appearing alongside the recording.

One Amplitude gotcha: the default session replay sampling rate is 0%. You need to explicitly configure a rate to capture any sessions — the recommended starting value is 1%. Deploy Amplitude without setting this and you’ll collect zero recordings with no error or warning.

Verdict: PostHog for teams that need an explicit free-tier recording budget. Amplitude for teams who want AI-assisted replay analysis with minimal configuration. Amplitude’s opacity on free quota is a minor annoyance at small scale; at enterprise scale, not knowing your limit is a real problem.

Feature flags

PostHog’s feature flags are the cleaner standalone product.

They work with no analytics enabled — separate SDK, separate free quota (1M requests/month), SDKs for 17+ languages. Supported flag types: boolean, multivariate, percentage rollouts, user/group targeting, remote config with JSON payloads, scheduled changes, flag dependencies, and early access management.

Server-side local evaluation (Node, Go, Python, Ruby, C#/.NET, PHP, Java, Rust) polls flag definitions every 30 seconds. Each definition-fetch counts as 10 flag requests but covers unlimited flag evaluations until the next poll. Cold start window: up to 30 seconds before definitions load. Edge and serverless environments (Lambda, Cloudflare Workers, PHP) need an external cache layer for local evaluation — without one, every invocation hits the remote API.

Amplitude includes unlimited feature flags in all tiers, tightly integrated with its experimentation and analytics layer. If you’re already deep in Amplitude, that integration is a genuine advantage — flag targeting against behavioral cohorts from the same interface. If you’re not on Amplitude, the integration has no value.

For dev teams who want feature flags without analytics lock-in, PostHog wins. The explicit free quota, 17-language SDK coverage, and zero dependency on analytics make it the right choice for teams shipping flags to production alongside any analytics stack.

Self-hosting PostHog

Don’t.

PostHog deprecated Kubernetes support and no longer offers it for new deployments. Docker Compose is suitable for lower event volumes — minimum specs are 4 vCPU, 16GB RAM, 30GB storage.

PostHog’s own documentation is clear: “We don’t offer any sort of guarantees around it working in certain ways on your infrastructure, and you assume all responsibility and risk.”

All paid features — extended retention, more projects, email support — are Cloud-only. Self-hosted gets you the free-tier feature set plus the full ops burden.

Community data documents the failure modes clearly. One developer published 6 undocumented failure modes discovered after 12 hours of debugging a fresh self-hosted install: a hidden CDP_API_URL environment variable defaulting to a non-existent Kubernetes URL, a Node.js crash loop from plugin server image mismatch, an encryption key double-encoding bug, static IP assignment causing PostgreSQL binding failures, Celery Beat scheduler lock conflicts in Redis, and Nginx DNS caching causing 502s after container recreation.

Cost math at 80M events/month: self-hosting runs ~$450/month in infrastructure; PostHog Cloud runs ~$2,183/month (posthog.com/pricing calculator, 2026-05-26). The ~$1,733/month gap disappears at 9–12 hours/month of DevOps time at standard engineering billing rates. Self-hosting makes sense only above 10M events/month with dedicated DevOps capacity, strong Kubernetes expertise, and a genuine data-residency requirement.

Amplitude doesn’t offer self-hosting. The decision doesn’t apply there.

LLM observability

This is the one area where PostHog has no competition in this comparison.

PostHog LLM Analytics tracks AI conversations, token counts, model costs, latency, and traces. It works with OpenAI, Anthropic, and other providers. The free tier covers 100,000 LLM events/month. Lovable uses it in production to debug LLM traces — see their customer story.

Amplitude has no equivalent. You can approximate token tracking with custom events, but you’re hand-rolling a schema that PostHog ships out of the box — including trace visualization, model cost roll-up, and AI evaluation tooling.

If you’re building AI agents, LLM-powered features, or anything that involves model calls you want to measure and optimize: PostHog is the only analytics tool in this comparison that handles it natively.

Teams shipping AI products typically also need error tracking and APM alongside analytics — see Sentry vs Datadog for the observability stack that pairs with PostHog in production.

PostHog Cloud EU

EU data residency is available, configured via serverZone: "EU" in the SDK. Feature parity between the EU and US regions is not explicitly confirmed in PostHog’s documentation — treat it as unverified. If feature parity is a hard compliance requirement, verify directly with PostHog before committing.

Competitive context

Mixpanel is the “neither fits” option. Cohort analysis is stronger than PostHog’s without Amplitude’s contract structure. Historically cheaper than Amplitude at equivalent MTU volume. Worth evaluating if you need behavioral analytics depth but aren’t ready to absorb Amplitude’s real-world pricing.

June.so is worth a footnote for teams under roughly 1,000 active users — lightweight, free at that scale, minimal instrumentation. Not a replacement for either tool at meaningful product volume.

For teams who also need lightweight, privacy-first web analytics — especially for GDPR-sensitive EU audiences — Plausible vs Google Analytics covers the natural pairing when PostHog handles your product analytics layer.

Verdict

Start with PostHog Cloud. The free tier is large enough that most early-stage teams don’t pay anything for the first 12 months. Pay-as-you-go pricing means you scale without a sales call, and the pricing calculator lets you model costs against your event volume before you commit to anything.

Move to Amplitude if you eventually need its cohort analysis depth — and only after you’ve modeled your MTU count against the Vendr ranges above. Budget a minimum of $30K–$70K/year once you’re past Starter in any meaningful usage pattern. The $49/month headline is not the price you’ll pay at growth-stage scale.

Pick PostHog if…Pick Amplitude if…
You’re pre-revenue or early-stageYour MTU count is stable and predictable
You need feature flags without analytics lock-inYou need enterprise cohort / ML features
You’re building AI products (LLM observability)You’re B2B SaaS and need Group Analytics
You need EU data residencyYou can budget $30K+/year
Budget is effectively zeroYou have a dedicated analytics team

Self-hosting PostHog: don’t. It stopped being a cost-saving strategy when Kubernetes support was sunsetted.

Caveats

PostHog Cloud EU feature parity with the US region is unverified — the region exists, but explicit parity confirmation wasn’t found in documentation as of May 2026. Verify directly before making EU residency a deciding factor.

Amplitude’s free session replay quota is not published. Budget modeling on free-tier replay volume is not possible without contacting Amplitude.

This article contains an affiliate link to Amplitude. The /go/amplitude link earns toolchew a commission on Starter/Startup signups only — commission does not apply to Growth or Enterprise deals. Affiliate status doesn’t change the verdict: PostHog is the right default for the target reader here; Amplitude earns a recommendation only in specific scenarios where it’s genuinely the better tool.

References