Why PostHog pricing is a usage question, not a plan question
PostHog's model charges by usage within each product it offers - product analytics events, session replay minutes, feature-flag requests, surveys, and the rest - rather than the seat-based licensing most legacy analytics tools use. That means two companies of the same size can land on very different bills depending on how much of the product they turn on and how they instrument it.
We are not going to restate PostHog's current free-tier allowances or per-unit rates here. Vendor pricing pages change on their own schedule, and a number frozen into a guide like this goes stale quietly. Check PostHog's own pricing page for the current figures, and treat anything below as the shape of the decision, not a quote.
The real cost drivers, in order of how often we see them matter
| Driver | Why it moves the bill | What we do about it |
|---|---|---|
| Event volume | The core meter for product analytics; autocapture-everything setups generate far more events than a designed taxonomy | Scope autocapture vs. explicit events before instrumentation ships |
| Session replay minutes | Recording every session at full fidelity is the fastest way to inflate a replay bill | Sample replay by cohort or funnel step instead of recording everything |
| Feature-flag requests | A flag evaluated on every page load across high-traffic surfaces adds up fast | Audit flag evaluation points and consolidate redundant checks |
| Data retention & export | Warehouse export and long retention windows are a separate line from day-to-day usage | Scope retention against actual analysis needs, not a default |
Self-hosted as a pricing lever, not just a deployment choice
Self-hosting removes PostHog's usage-based billing in exchange for infrastructure and operational cost - compute, storage, upgrades, and whoever is on call when it breaks. Whether that trade is worth it depends on volume and on whether the team already runs comparable infrastructure. Our self-hosted guide walks through how we scope that comparison honestly, including the engineering-time side most vendor comparisons leave out.
Where an implementation review pays for itself
The single highest-leverage move on PostHog spend is usually architectural, not operational: deciding upfront what should be autocaptured, what needs an explicit event, and what does not need to be tracked at all. That decision is cheap to make before instrumentation ships and expensive to unwind after a year of noisy events.