← Blog

Statsig pricing: what the free tier doesn't cover

Statsig pricing runs on three tiers: a free Developer plan with 2 million metered events a month, a Pro plan at $150/mo for 5 million events, and a custom-quoted Enterprise plan. Feature flag and gate checks are free on every tier, by design: what you pay for is logged events, not flags. The free tier is genuinely usable, but it skips entire categories of capability, not just a lower cap on the same features. This post breaks down what each tier actually includes, what counts as a billable event, and what Developer leaves out.

Key Takeaways

  • Statsig’s free Developer plan costs $0/mo: 2 million metered events, 50,000 session replays, 1-year analytics retention, and unlimited seats. Flag and gate checks are free on every tier.
  • Pro costs $150/mo for 5 million events, then $0.05 per 1,000 over that. It’s the first tier with unlimited analytics retention, CUPED and other advanced statistics, and change-review workflows.
  • Developer skips SSO, RBAC, warehouse-native deployment, automated rollouts, and stale-gate management entirely. None of those show up at a lower cap; they’re missing.
  • A billable event is an exposure (checking a gate or experiment with Metric Lifts enabled, or a partial rollout) or a custom event logged through the SDK. A flag held flat at 0% or 100% generates no metered event.
  • A growth team running five concurrent experiments against 300,000 monthly users can clear both the 2 million free cap and the 5 million Pro allowance in the same month. The math is below.

Statsig pricing in 2026, in plain numbers

Developer is free, no credit card required: 2 million metered events a month, unlimited flag and config checks, 50,000 session replays a month, core experimentation (A/B and multivariate tests), dynamic configs, 1-year analytics retention, and unlimited seats.

Pro costs $150/mo, billed monthly in advance, with 5 million events included and $0.05 per 1,000 events beyond that. It adds advanced experimentation (custom queries, holdouts, multi-armed bandits), advanced statistics (CUPED, Bonferroni correction, switchback and non-inferiority tests), automated rollouts, change-review workflows, and unlimited analytics retention.

Enterprise is a custom quote, on an event-based or experiment-based contract with volume discounts. It adds warehouse-native deployment, outgoing data integrations, SSO, RBAC, SCIM provisioning, priority support, and HIPAA eligibility (a signed BAA).

The full rate card is at Statsig’s pricing page.

What counts as a billable event

Statsig meters two kinds of events, and both are scoped to your production environment only, per Statsig’s billing FAQ. An exposure event fires when the SDK checks a user against a gate, experiment, or dynamic config that has Metric Lifts turned on, or that’s running a partial rollout. A custom event is anything you log through the SDK, a data-warehouse import, or a data collector, billed once no matter how many experiments read it.

The part that surprises teams: a flag held flat at 0% or 100% with Metric Lifts off never generates a metered event. That’s the mechanism behind “unlimited flag and config checks” on every tier. The moment you turn on Metric Lifts or split traffic to measure a change, that same check starts counting against your cap.

What the free Developer tier doesn’t cover

Developer isn’t a smaller version of Pro. Several categories are absent outright:

  • No advanced experimentation: no custom metric queries, no holdouts, no multi-armed bandits
  • No advanced statistics: no CUPED variance reduction, no Bonferroni correction, no switchback or non-inferiority tests
  • No automated rollouts or environment-based targeting: ramping and gating by environment is a Pro feature
  • No monitoring or flag-impact analysis: no alerts on a gate’s downstream metric movement
  • No stale-gate management or code references: nothing flags a gate nobody has touched in months
  • No user profiles or behavioral cohorts: no saved segments built from past behavior
  • No SSO, RBAC, or SCIM provisioning: access control is all-or-nothing
  • No change-review or approval workflows: anyone with a seat can flip a gate
  • No warehouse-native deployment or data export API: Developer’s analytics stay inside Statsig
  • Support is community Slack only, with no SLA

Analytics retention is capped at one year on Developer versus unlimited on Pro, and session replay drops from 100,000 a month on Pro to 50,000 on Developer. Those two are lower caps on the same feature. Everything else on the list above is missing from Developer entirely.

Two teams, two different Statsig bills

A five-person team running one flag and one experiment. One gate sits flat at 100% (no metered events). One 50/50 experiment runs against 40,000 monthly users with Metric Lifts on, firing one exposure per user. The team logs about three custom events per user a month (signup, checkout, a feature-usage ping): 40,000 exposures plus 120,000 custom events, 160,000 events total. That’s a twelfth of the 2 million free cap. Total: $0/mo, comfortably on Developer, with an actual experiment running.

A growth team running five concurrent experiments. 300,000 monthly users, each exposed to roughly two of the five experiments on average (600,000 exposure events), logging about 20 custom events a user across funnels and feature usage (6 million custom events). That’s 6.6 million metered events a month, past the 2 million free cap and past the 5 million included on Pro.

  • Base Pro plan: $150
  • Overage: 1.6 million events × $0.05/1,000 = $80
  • Total: about $230/mo

Turning Metric Lifts on across five experiments at once is what drives that jump, not the extra users. Every active experiment multiplies exposures the moment you start measuring it instead of just shipping it.

Where a flat rate changes the math

dif prices the same axis Statsig does, events, but on one flat rate instead of a tiered jump. dif Cloud pricing is $50/mo for 1 million events, then $0.02 per 1,000 after that, with unlimited seats and no feature gate between tiers. There’s no $0-to-$150 cliff and no SSO or change-review wall between plans, because there’s only one plan.

That comparison only covers the hosted analytics layer. The bigger difference is upstream of billing: dif has no dashboard to run gates from at all. Every flag and experiment is a Markdown file in your repo, assigned locally with no network call and no exposure event required just to use the flag. Feature flags in your repo covers that model.

Statsig’s advanced statistics (CUPED, sequential testing, multi-armed bandits) and warehouse-native deployment are real strengths dif doesn’t try to match. If your team’s experimentation program needs that depth, Statsig Pro or Enterprise is the right tool for it, the same way LaunchDarkly wins on instant rollback. dif’s bet is that most teams running a handful of flags and experiments would rather keep them in git than learn a metering model.

FAQ

Is Statsig really free? Yes, for the Developer tier: no credit card, 2 million metered events a month, 50,000 session replays, unlimited seats, and unlimited flag checks. It skips advanced experimentation, SSO, and warehouse-native deployment, which only appear on Pro or Enterprise.

What counts as an event in Statsig’s pricing? An exposure (checking a gate, experiment, or dynamic config with Metric Lifts on, or a partial rollout) or a custom event logged through the SDK, a warehouse import, or a data collector. Checks against a flag held flat at 0% or 100% with Metric Lifts off don’t count.

Does Statsig charge for feature flags? Not for the check itself. Flag and config checks are unlimited on every tier. You pay once you turn on Metric Lifts to measure a rollout, or once a user is exposed to an actual experiment.

How much does Statsig Pro cost? $150/mo, billed monthly in advance, with 5 million events included and $0.05 per 1,000 events beyond that. It’s the tier that adds unlimited analytics retention, advanced statistics, and change-review workflows.

What does Statsig’s free tier not include? Advanced experimentation (custom queries, holdouts, multi-armed bandits), advanced statistics (CUPED, Bonferroni correction, switchback tests), automated rollouts, monitoring and flag-impact analysis, stale-gate management, SSO and RBAC, and warehouse-native deployment. Analytics retention is also capped at one year, versus unlimited on Pro.

Is there a cheaper alternative to Statsig? dif’s CLI and SDK are free and open source, with no event meter at all for using a flag, since assignment runs locally. dif Cloud is a flat $50/mo for 1 million events if you want hosted analytics. It doesn’t match Statsig’s advanced statistics or warehouse-native deployment, so the right choice depends on whether you need that depth.

The number before you commit

Statsig pricing has no per-seat charge and no cost for checking a flag, which makes the free tier look more generous than it is. Metric Lifts and experiment exposures are the actual gate. Turn those on across more than a couple of tests and you clear 2 million events fast, landing on Pro’s $150/mo plus whatever you run past 5 million. Count your actual exposure and custom-event volume, not just your user count, before you assume Developer covers you.

If a flat per-event rate and flags in your repo matter more than Statsig’s deeper statistics, dif Cloud runs $50/mo regardless of plan tier, and the CLI is free to try with no account: npm install -g @dif.sh/cli, then dif init. The docs cover the full install.