LaunchDarkly vs Statsig vs PostHog Feature Flags: Real Feature Flag Cost at 100,000 MAU in 2026
The short answer: these three don't meter the same thing at all, and the gap is dramatic in a direction easy to get backwards. LaunchDarkly bills directly by client-side Monthly Active Users — at 100,000 MAU, that's roughly $833–$1,050/month depending on which of two conflicting rates on LaunchDarkly's own pricing page applies, plus a per-service-connection fee on top. Statsig does not meter ordinary flag and config checks at all — they're unlimited on every tier, including the free Developer plan — so a basic rollout-and-targeting use case with no experimentation enabled can genuinely cost $0 on Statsig regardless of MAU. Statsig's metered "Events" line only activates once a team turns on experimentation or Metric Lift analysis, which generates exposure and metric-logging events; at an ILLUSTRATIVE heavy-experimentation workload (20 concurrent experiments plus custom metric logging, averaging 50 metered events per user per month), that same 100,000-MAU product costs $150/month on Statsig's Pro tier. PostHog, at an ILLUSTRATIVE 20 flag-check requests per user per month, costs roughly $100/month. The common thread across all three: LaunchDarkly's per-MAU meter doesn't care what a user does, while Statsig's and PostHog's usage-based meters only charge for the specific activity that generates a billable event — and for Statsig specifically, a plain flag check is not that activity.
The billing-unit mismatch, up front
LaunchDarkly bills client-side Monthly Active Users — a count of distinct users, not evaluation volume — meaning a product with many registered-but-rarely-active users pays the same per-user rate as one where every user triggers flag checks constantly. Statsig bills Metered Events, a specific, narrower category than "every flag evaluation": ordinary flag and config checks are explicitly unlimited and unmetered on every Statsig tier, and the metered-event counter only increments when experimentation or Metric Lift features are active and generate exposure or metric-logging events. PostHog bills feature-flag API requests specifically, with a declining per-request rate at higher volume tiers, separate from its other product lines (analytics events, session replay) which meter independently. None of these three converts cleanly into "cost per 100,000 MAU" without first establishing what the product actually does with flags, not just how many users it has.
What changed in 2026
LaunchDarkly stopped charging per seat entirely, moving to unlimited seats on every tier — a real, confirmed structural change from its historical per-developer-seat model, now replaced by the service-connection and client-side-MAU meters described below.
What each vendor bills
LaunchDarkly (OFFICIAL, launchdarkly.com/pricing, with an unresolved internal conflict). Developer: free, for individual developers and small teams. Foundation (paid): $10 per service connection per month (each server-side SDK instance connected to an environment) plus $8.33 per 1,000 client-side MAU, billed annually — this is the rate shown on LaunchDarkly's own pricing-page plan card. A separate comparison table on that same pricing page lists $10 per 1,000 MAU instead of $8.33 — a genuine, unresolved internal inconsistency on LaunchDarkly's own site; confirm directly which currently applies before budgeting. Enterprise and Guardian: custom.
Statsig (OFFICIAL, statsig.com/pricing). Developer: free, unlimited seats, 2 million Metered Events/month included, and unlimited flag and config checks regardless of volume — ordinary feature-flag evaluations are not a billable unit on Statsig at any tier. Pro: $150/month, 5 million Metered Events included (also with unlimited flag/config checks), then $0.05 per additional 1,000 events. Metered Events specifically cover exposure events generated when experimentation or Metric Lift analysis is enabled, plus logged custom events and metrics — not raw flag evaluations. Statsig deduplicates exposures: a single user repeatedly hitting the same experiment within its exposure window generates one counted exposure, not one per check. Statsig was acquired by OpenAI in 2026, a development that does not appear to have changed its public pricing structure but is relevant context for teams evaluating its long-term product direction.
PostHog Feature Flags (OFFICIAL, posthog.com/pricing, as part of PostHog's broader usage-based platform). Free: 1,000,000 feature-flag requests/month. Beyond that, a declining tiered rate: $0.0001 per request from 1–2 million, $0.000045 per request from 2–10 million, $0.000025 per request from 10–50 million, and $0.00001 per request beyond 50 million. Because PostHog meters flags as one line within a broader platform that also bills analytics events and session replay independently, a team already using PostHog for analytics may find flags effectively marginal-cost-only, while a team adopting PostHog for flags alone is buying into a larger product suite than a flags-only tool would require.
Two Statsig scenarios, because "100,000 MAU" alone doesn't determine its cost
Scenario A: basic rollout and targeting, no experimentation. A product using Statsig purely to gate features by user segment, with no A/B tests or Metric Lift analysis running. Because ordinary flag checks generate no Metered Events at all, this workload produces zero billable events regardless of MAU or check frequency, and sits comfortably on the free Developer tier. Cost: $0/month.
Scenario B: experimentation and Metric-Lift-heavy. An ILLUSTRATIVE AI SaaS running 20 concurrent experiments and logging custom metrics for Metric Lift analysis, averaging 50 metered events per user per month (a combination of deduplicated experiment exposures and metric-logging events — not flag checks). At 100,000 MAU, that's 5,000,000 metered events, landing exactly within the Pro tier's included allowance. Cost: $150/month (flat, no overage at this specific volume).
Cost at 100,000 MAU
Formula: LaunchDarkly = (MAU ÷ 1,000 × per-1,000-MAU rate) + (service connections × $10); Statsig = $0 if no experimentation/Metric-Lift activity generates metered events, or Developer/Pro tier pricing applied to actual metered-event volume if it does; PostHog requires an ILLUSTRATIVE requests-per-user assumption, since it doesn't bill MAU directly.
| Platform | Basis | Monthly cost at 100,000 MAU |
|---|---|---|
| Statsig, Scenario A (flags only, no experimentation) | Unlimited flag checks, zero metered events generated | $0 |
| PostHog Feature Flags | ILLUSTRATIVE 20 requests/user/month = 2,000,000 requests; 1M free, 1M billable at the 1–2M tier rate | ~$100 |
| Statsig, Scenario B (heavy experimentation) | ILLUSTRATIVE 50 metered events/user/month = 5,000,000 events, within Pro's included allowance | $150 |
| LaunchDarkly, at the $8.33/1,000-MAU rate | 100,000 ÷ 1,000 × $8.33 + 5 service connections × $10 | $883 |
| LaunchDarkly, at the conflicting $10/1,000-MAU rate | 100,000 ÷ 1,000 × $10 + 5 service connections × $10 | $1,050 |
At this specific illustrative usage pattern, LaunchDarkly costs anywhere from roughly 6 times more than Statsig's heavy-experimentation scenario to effectively infinitely more than Statsig's flags-only scenario — a gap driven by billing philosophy (per-user versus per-triggered-event, and specifically Statsig's choice not to meter plain flag checks at all), not by any difference in underlying infrastructure cost between the three platforms.
Why the per-user versus per-triggered-activity distinction matters so much
Formula: LaunchDarkly cost is invariant to how flags are actually used; Statsig's cost is zero until experimentation or Metric Lift features generate metered events, then scales with that specific activity. A product doing pure feature-gating with no experimentation runs free indefinitely on Statsig regardless of user count, while the identical product pays LaunchDarkly's full per-MAU rate for the exact same usage pattern, since LaunchDarkly doesn't distinguish between a simple rollout check and a complex experiment evaluation. The moment a team turns on Statsig's experimentation or Metric Lift features, cost appears and scales with exposure and metric-logging volume — this article's Scenario B illustrates one way that volume could reach Pro-tier pricing, but a lighter experimentation program (fewer concurrent experiments, less custom metric logging) could just as easily stay within the free 2-million-event Developer allowance even with experimentation turned on.
Sensitivity
- Whether experimentation or Metric Lift is enabled at all, on Statsig specifically. The single largest lever by far; it is the difference between $0 and a metered bill, independent of MAU.
- Metered events per user per month, once experimentation is active. Determines whether Statsig's Pro tier's included allowance covers the workload or overage applies at $0.05/1,000 events beyond it.
- Which LaunchDarkly MAU rate is current. An unresolved conflict between $8.33 and $10 per 1,000 MAU on LaunchDarkly's own pricing page — a roughly 20% swing on this article's headline LaunchDarkly figure.
- Service connection count, on LaunchDarkly. Each server-side SDK instance connected to an environment adds $10/month, a real cost for architectures with many services or environments.
Budgeting traps
- Assuming every flag evaluation on Statsig is a billable Metered Event. It is not — ordinary flag and config checks are unlimited and free on every tier; only experimentation exposures and logged metrics count toward the meter.
- Citing LaunchDarkly's $8.33 rate without checking the conflicting $10 figure on the same pricing page. Confirm directly which currently applies before budgeting.
- Ignoring LaunchDarkly's per-service-connection fee as a rounding error. At $10 per connection, a microservices architecture with many server-side SDK instances can accumulate this cost meaningfully.
- Assuming Statsig's cost scales with MAU at all for a flags-only use case. It does not; cost only appears once experimentation or Metric Lift activity is turned on.
What to ask before you buy
Confirm whether your Statsig usage will involve experimentation or Metric Lift analysis, not just feature flags, since that distinction determines whether your bill is $0 or usage-based — MAU count alone does not predict Statsig's cost the way it does LaunchDarkly's. Ask LaunchDarkly directly which of its two conflicting published MAU rates ($8.33 or $10 per 1,000) currently applies, since the discrepancy appears on LaunchDarkly's own pricing page without public resolution.