Statsig Alternatives
Seven ways out of <em>Statsig</em>, priced on their real billing unit, and when staying still makes sense.
GrowthBook is the best Statsig alternative for most product teams: the same flags and experiments, an open-source core, and a price per seat ($40 a month on Pro) instead of a meter on your events. PostHog is the pick if you used Statsig as an all-in-one bundle with analytics and replays, LaunchDarkly if releases and approvals matter most, and Datadog Experiments (formerly Eppo) if Datadog already watches your stack.
Flagsmith, Unleash and ConfigCat cover teams that mostly used Statsig for feature flags. The reason so many teams are looking: Statsig changed owners twice in under a year, from OpenAI in 2025 to Amplitude in 2026, while the people who built it stayed at OpenAI. Below, what each tool costs on its own unit, the billing traps the other comparisons skip, and the cases where keeping Statsig is the right call.
The ranking is editorial: a sponsored spot is always flagged on its card, and no affiliate link changes a score.
Why teams are leaving Statsig
Statsig built one of the strongest experimentation platforms on the market: feature gates, A/B tests with a serious statistics engine, product analytics and session replay on the same events, with a warehouse-native option for data teams. The field reports we read rarely fault the product itself. They leave over who owns it, where the roadmap goes, and how the event meter grows. Here are the reasons, each with the fact behind it.
Two owners in under a year
A roadmap now written inside Amplitude
An event meter that grows with traffic
The free plan stops measuring at 2 million events
Governance and data access sit on Enterprise
Customers report friction since the sale
GrowthBook
GrowthBook is the closest match to what Statsig did best: feature flags and experiments in one tool, with a statistics engine that offers Bayesian, frequentist and sequential testing, CUPED variance reduction and post-stratification. The difference is where the data sits. GrowthBook is warehouse-native: it reads your metrics where they already live (Snowflake, BigQuery, Databricks, Redshift, ClickHouse and others) with read-only access, and every SQL query behind a result is visible. If you have no warehouse yet, a managed one is included. More than 4,000 companies use it, and its core code is open source, so a version you deploy stays yours to run.
The pricing logic is the opposite of Statsig's. GrowthBook charges per seat, the people who log in to run experiments and manage flags, not per event or per end user. The cloud Starter plan is free for up to 3 users and one project, with unlimited flags, experiments and traffic. Pro costs $40 per seat per month, for up to 30 users and 3 projects, and adds sequential testing, CUPED, sticky bucketing, URL split tests, multi-arm bandits and the AI visual editor. A team of 10 pays $400 a month in seats; traffic only adds to it past the included CDN usage, 2 million requests and 20 GB a month on Pro, then $10 per extra million requests and $1 per GB. Self-hosting is free with unlimited users, but the free Open Source edition leaves out sequential testing, CUPED and bandits, which come with a self-hosted Enterprise license. The meter to watch is the managed warehouse: 2 million events a month on Pro, then $30 per extra million, so 10 million events add $240 a month. Teams leaving Statsig also get a dedicated migration kit: gates become feature flags, past Statsig experiments transfer, and an AI assistant rewrites JavaScript, TypeScript and React code today.
- Seat pricing: flag evaluations are not billed
- Warehouse-native analysis with visible SQL
- Migration kit built for Statsig customers
- Free self-hosted edition with unlimited users
- ✓Bayesian, frequentist, sequential and CUPED on Pro
- ✓Past Statsig experiments can be imported
- ✓Your event history can stay in your own warehouse
- ✗SAML SSO, SCIM and approval workflows need Enterprise
- ✗Pro stops at 30 users and 3 projects
- ✗Managed warehouse billed past 2M events on Pro
| Criterion | GrowthBook | Statsig |
|---|---|---|
| Billing unit | Seats ($40/seat/mo on Pro) | Metered events |
| Free plan | 3 users, unlimited traffic | 2M events a month |
| Self-hosting | Free edition, unlimited users | Cloud; warehouse native on Enterprise |
Take GrowthBook if experiments are the reason you bought Statsig and you have, or want, a data warehouse. If you also want product analytics and session replay in the same tool without SQL, PostHog fits better.
PostHog
PostHog covers the same ground as Statsig's bundle: feature flags, experiments, product analytics, web analytics and session replay share one SDK and one dataset, so you go from a test result to the replays of the users behind it. It adds error tracking, surveys, logs and a data warehouse. On ownership, it is the opposite story: PostHog says it has significant revenue, is default alive and does not rely on investors to grow, and its code is MIT-licensed outside its enterprise directory. Its experiment engine runs Bayesian or frequentist statistics, with CUPED variance reduction available.
Pricing is per product and per unit, with a free allowance that resets every month: 1 million analytics events, 5,000 recordings and 1 million feature flag requests. Experiments are billed with feature flags. Past 1 million, a flag request costs $0.0001 up to 2 million, then $0.000045 up to 10 million, and less beyond. At 5 million flag requests a month you pay about $235 ($100 for the second million, $135 for the next three). The trap is in the unit. A request is a call that evaluates your flags: client SDKs make one when they load and when a user is identified, and server-side local evaluation bills each poll, 10 requests when definitions changed. One server polling at default settings makes about 86,400 requests a month; an older SDK without ETag support is billed 10 requests per poll, about 864,000. SSO, audit logs and custom roles come in paid platform packages. Set a billing limit per product: on the free plan, extra flag calls get a quota-limited response instead of a bill. If PostHog's meters are what you are fleeing next, our PostHog alternatives compare the tools on the other side.
- Flags, experiments, analytics and replay in one SDK
- Free allowance every month, no card
- Billing limit per product
- No minimum spend or annual contract
- ✓Closest bundle to Statsig, from an independent company
- ✓Unlimited team members on every plan
- ✓CUPED and frequentist or Bayesian results
- ✗Flag requests billed, unlike Statsig's unlimited checks
- ✗Local evaluation polls count as requests
- ✗Self-hosted deployments officially unsupported
| Criterion | PostHog | Statsig |
|---|---|---|
| Flag checks | 1M requests free, then from $0.0001 | Unlimited on every plan |
| Analytics | 1M events free a month | 2M metered events free |
| Owner | Independent company | Amplitude since 2026 |
Take PostHog if you want Statsig's all-in-one product from an independent vendor with a free allowance. If your data team wants results computed in its own warehouse, GrowthBook or Datadog Experiments suit it better.
LaunchDarkly
LaunchDarkly is the reference for feature management, and the tool many teams compared Statsig against in the first place: targeting by attribute and segment, percentage rollouts, release guardrails and approvals. It has grown into experimentation (A/B tests and multi-armed bandits on every plan, the free one included), observability and AI configs.
The free Developer plan has unlimited seats and flags, 1 project and 3 environments. Foundation is pay as you go, and flags bill on two units. A service connection is one instance of one server-side SDK connected to one environment, so every pod, replica and staging environment counts: 5 are free, then $10 each billed yearly or $12 monthly. Client-side MAU are unique context keys evaluated in a month: 1,000 are free, then $8.33 per 1,000 billed yearly or $10 monthly. LaunchDarkly's own example, 10 connections and 2,000 MAU, costs $58.33 a month billed yearly. A web app with 50,000 monthly users costs $408.17 a month in MAU alone, billed yearly, or $490 monthly. The trap: if you change an existing context key, LaunchDarkly counts the new key as an additional MAU, so an anonymous visitor who logs in under a new key can be billed twice. Session replays, errors, traces and AI runs are metered on their own past their free amounts. SSO (the lower plans have 2FA only), SCIM, custom roles, audit logging and mandatory approvals are Enterprise, on quote.
- Mature targeting, approvals and release guardrails
- Experiments and bandits from the free plan
- Unlimited seats on every plan
- Flag pricing shown with a worked example
- ✓Mature targeting, approvals and flag history
- ✓No per-seat fee
- ✓Annual billing cuts the unit price
- ✗MAU and service connections are hard to forecast
- ✗A changed context key counts as a new MAU
- ✗SSO, SCIM and audit logs on Enterprise only
| Criterion | LaunchDarkly | Statsig |
|---|---|---|
| Billing unit | Service connections and client MAU (flags) | Metered events |
| Free plan | 5 connections, 1K MAU | 2M events a month |
| Experiments | A/B tests and bandits from Developer | A/B tests free; advanced methods on Pro |
Take LaunchDarkly if releases and governance matter more to you than the statistics engine. If experiments drive your roadmap and your traffic is large, GrowthBook's seat price will cost less.
Datadog Experiments (Eppo)
Eppo was a warehouse-native experimentation platform. Datadog acquired it in 2025 and now sells it as Datadog Experiments, next to Datadog Feature Flags, Product Analytics and real user monitoring. The engine supports sequential, fixed-sample and Bayesian methods with CUPED++ variance reduction, and a warehouse-native mode computes results on your own business metrics. You do not need Product Analytics to use it, and you can keep another flag system. The real advantage over Statsig is the link to observability: an experiment can show that a variant raised latency or errors, not only conversion, if you also run Datadog RUM or Product Analytics, two separate products.
The unit is unusual: Datadog charges per launched experiment, from $450 per experiment per month billed annually, or $575 on demand, with unlimited seats, metrics and dimensional analyses. An experiment only becomes billable once it has launched, reached at least 500 subjects and collected 3 days of data, so a test that never starts costs nothing. Each experiment includes 1 million monthly flag configuration requests (MFCRs). Running 4 experiments at once costs $1,800 a month billed annually, $21,600 a year, or $2,300 a month on demand. Feature Flags alone are free under 1 million MFCRs, then $55 per million billed annually ($100 on demand) up to 10 million. Our guide to switching away from Datadog covers the other side of that platform choice.
- Warehouse-native mode with CUPED++
- Experiments tied to latency and error data
- Unlimited seats and metrics
- Billing starts only after 500 subjects and 3 days
- ✓Sequential, Bayesian and CUPED++ methods
- ✓Works with your own flag system
- ✓Public price per experiment
- ✗Costly for teams running many experiments
- ✗Deepest value only inside Datadog
- ✗Annual billing for the lower rate
| Criterion | Datadog Experiments (Eppo) | Statsig |
|---|---|---|
| Billing unit | Launched experiments | Metered events |
| Entry price | $450 per experiment/mo billed yearly | Pro $150/mo |
| Warehouse native | Included mode | Enterprise only |
Take Datadog Experiments if Datadog already monitors your stack and you run a handful of experiments that must be right. If you run dozens of tests a month, GrowthBook's seats will cost far less.
Flagsmith
Flagsmith is a feature flag and remote config platform whose core is open source under the BSD-3-Clause license, a permissive license that lets you modify and redistribute it. You can run it as SaaS, in a private cloud it manages, or on your own servers, air-gapped included on Enterprise. It added native experimentation, still in beta: a multivariate flag serves the variants, events go to a warehouse (the managed Flagsmith warehouse, or your own ClickHouse or Databricks), and a built-in Bayesian engine shows lift, credible intervals and win probability. Statsig's engine goes further, with CUPED, holdouts and switchback tests, so treat Flagsmith as a flag tool that can also test. Read our full list of Flagsmith alternatives for its limits.
Pricing is per API request, counted across all projects. A single-page React app makes about 1 request when a user loads it; a server-rendered site makes about 1 per page view. Free covers 50,000 requests a month, 100,000 users in experiments, 1 team member and 1 project. Start-Up costs $45 a month, or $40 billed yearly, for 1 million requests, 500,000 users in experiments and 3 team members. Scale-Up costs $300 a month, or $250 billed yearly, with 5 members included and $60 per extra member up to 20, plus SAML SSO and audit logs. The trap is overage: past your quota on Start-Up you pay $7 per 100,000 requests, so a month at 1.5 million requests adds $35, unless it is your first overrun: within 200% of the limit, that one gets a 30-day grace period.
- BSD-3-Clause core, self-hosted or air-gapped
- Native Bayesian experiments, in beta
- Public prices with a yearly discount
- 30-day grace period on a first overrun
- ✓Permissive BSD-3-Clause license
- ✓Cheap entry for small teams
- ✓Flags keep serving past the limit
- ✗Experiment engine thinner than Statsig's
- ✗Free plan has 1 team member
- ✗Overage at $7 per 100K requests on Start-Up
| Criterion | Flagsmith | Statsig |
|---|---|---|
| Billing unit | API requests and team members | Metered events |
| Self-hosting | Open source, BSD-3-Clause core | Cloud; warehouse native on Enterprise |
| Entry paid plan | $45/mo ($40 billed yearly) | $150/mo |
Take Flagsmith if self-hosting under a permissive license matters more than advanced statistics. If you need CUPED and sequential tests, GrowthBook covers both on Pro, or self-hosted with an Enterprise license.
Unleash
Unleash is an open-source feature management platform built around progressive delivery: gradual rollouts, targeting, change requests with four-eyes approvals, stale-flag tracking and impact metrics that can stop a rollout. Server-side SDKs fetch the configuration, cache it in memory and evaluate flags locally, so evaluations create no request and no user data leaves your servers. Its experiment side is the weak point next to Statsig: Unleash splits users into variants, but to attribute results you send impression events to an analytics tool, which computes the significance.
The price is per seat. Pay-As-You-Go costs $75 per seat per month, billed monthly by card, with SSO, SCIM, role-based access and audit logs included, and unlimited flags, projects, monthly active users and service connections. Pay-As-You-Go on Unleash Cloud includes 53 million API requests a month, then $5 per million; at default settings one server-side connection uses about 216,000 requests a month, so the allowance covers about 245 of them. Frontend SDKs, though, make requests for each user session, so a client-heavy app can pass it. Self-hosting the paid edition carries a 5-seat minimum, $375 a month. The free Open Source edition is self-hosted only, with 1 project, 2 environments, up to 5,000 flags and no SSO or change requests. Custom annual contracts add 99.99% uptime and premium support, on quote.
- Unlimited flags, MAU and service connections
- SSO and approvals on the self-serve plan
- Local evaluation, no request per check
- Free self-hosted Open Source edition
- ✓Server-side traffic does not change the bill
- ✓Governance without an Enterprise contract
- ✓Cloud in the EU or the US
- ✗Experiment results need an outside analytics tool
- ✗$75 per seat adds up for large teams
- ✗Open Source edition limited to 1 project and 2 environments
| Criterion | Unleash | Statsig |
|---|---|---|
| Billing unit | Seats ($75/seat/mo) | Metered events |
| SSO | Included on Pay-As-You-Go | Enterprise only |
| Experiment analysis | Variants sent to your analytics tool | Built-in stats engine |
Take Unleash if most of your Statsig usage was rollouts and kill switches, not experiment readouts. If you need a built-in stats engine, pick GrowthBook or PostHog instead.
ConfigCat
ConfigCat is built around flags and remote configuration: targeting rules, segments and percentage options, with 19 SDKs listed, from JavaScript and Python to Unity and Unreal Engine. For an A/B test, it logs flag evaluations into Google Analytics, Amplitude, Mixpanel or Segment, which compute the result. That is a step down from Statsig for experimentation, and a deliberate choice: if Statsig's analytics sat unused, you stop paying for them.
Its unit is the config JSON download, the file your app fetches from ConfigCat's CDN. Between two downloads you can evaluate flags as many times as you like, and you control how often the SDK polls. Seats, MAU and service connections are unlimited on every plan, and the Forever Free plan has the same feature set as the paid plans. Free gives 10 flags per product across 2 products, 2 environments, 5 million downloads and 20 GB of network traffic a month. Pro costs $110 a month billed yearly ($1,320 a year) or $120 monthly, for 100 flags per product across 3 products, 3 environments, 25 million downloads and 100 GB. Smart, with unlimited flags, costs $325 billed yearly. Two caps force an upgrade: flags per product and monthly downloads. ConfigCat's own sizing example, a web app whose 15,000 users each spend 5 hours a month in it at the default 60-second polling, makes 4.5 million downloads a month, close to the free cap. Past a limit your app keeps running, and ConfigCat contacts you to upgrade or buy extra quota. Early-stage startups can apply for 50% off.
- Flat monthly price, unlimited seats and MAU
- Same features on the free plan
- Startup program at 50% off
- 19 SDKs, game engines included
- ✓No overage billing: the plan sets the bill
- ✓Evaluations between downloads are free
- ✓Same feature set on the free plan, with lower quotas
- ✗A/B results read in GA4, Amplitude or Mixpanel
- ✗Free plan capped at 10 flags per product
- ✗Unlimited flags only from Smart at $325/mo billed yearly
| Criterion | ConfigCat | Statsig |
|---|---|---|
| Billing unit | Flat plan, capped by flags and downloads | Metered events |
| Experiments | Results computed in GA4, Amplitude or Mixpanel | Built-in stats engine |
| Entry paid plan | $110/mo billed yearly | $150/mo |
Take ConfigCat if you only need toggles and want a fixed bill. If you still run A/B tests and want the result in the same tool, GrowthBook or PostHog keep that loop.
7 Statsig alternatives compared
Each score is the plain average of five criteria chosen for teams leaving Statsig: experiment analysis, flags and rollouts, price as you grow, independence and exit (open source, owner, data you keep), and ease of moving from Statsig. Prices were read on each pricing page in 2026. The edge column says what each tool gives you that Statsig, under its new owner, does not.
| Best for | Edge over Statsig | Free plan | Team size | Visit | ||||
|---|---|---|---|---|---|---|---|---|
| 1 | GrowthBook | Best overall Statsig replacement | Seat price, open-source core, Statsig importer | 4.5/5 | Free (3 users, cloud); Pro $40/seat/mo | ✓ | Data-led product teams | Visit → |
| 2 | PostHog | Best all-in-one replacement | Same bundle, independent vendor, free allowance | 4.3/5 | Free monthly allowance; then usage-based, from $0.0001 per flag request | ✓ | Startups and product teams | Visit → |
| 3 | LaunchDarkly | Best for release management at scale | Deepest release governance | 4.1/5 | Free Developer plan; Foundation from $10 per service connection (billed yearly) | ✓ | Large engineering orgs | Visit → |
| 4 | Datadog Experiments (Eppo) | Best for teams already on Datadog | Experiments tied to observability data | 4.0/5 | From $450 per launched experiment/mo billed yearly ($575 on demand) | — | Datadog customers | Visit → |
| 5 | Flagsmith | Best for self-hosting flags anywhere | BSD-3-Clause self-hosting, air-gapped option | 3.9/5 | Free (50K requests/mo); Start-Up $45/mo ($40/mo billed yearly) | ✓ | Teams that self-host | Visit → |
| 6 | Unleash | Best for flag-heavy teams on a seat price | SSO and approvals on a self-serve seat plan | 3.8/5 | Pay-As-You-Go $75/seat/mo (billed monthly); free self-hosted Open Source | ✓ | Platform teams | Visit → |
| 7 | ConfigCat | Best for simple flags on a flat bill | Flat bill, unlimited seats and MAU | 3.5/5 | Free (10 flags per product); Pro $110/mo billed yearly ($120 monthly) | ✓ | Small teams | Visit → |
Is your tool missing from this page? Email us: hello@hackceleration.com
You're leaving Statsig because... take this
$40 per seat on Pro; flag evaluations run locally and are not billed.
You used flags, tests, analytics and replay togetherPostHogSame bundle, 1M flag requests and 1M events free each month.
Safe releases and approvals matter mostLaunchDarklyFree Developer plan, then $10 per service connection billed yearly.
Datadog already monitors your stackDatadog ExperimentsFrom $450 per launched experiment a month billed yearly.
You must host flags on your own serversFlagsmithBSD-3-Clause core; air-gapped installs on Enterprise.
You run hundreds of flags and few experimentsUnleash$75 per seat, unlimited flags, MAU and service connections.
You only need on/off switchesConfigCatFree for 10 flags per product, Pro $110 a month billed yearly.
Your tests are marketing tests on web pagesVWOA visual editor for marketers; Optimizely is the enterprise option, on quote.
Stay on Statsig, or switch?
Staying makes sense if…
- You fit inside the free Developer plan: under 2 million events a month, with unlimited flag checks and 50,000 session replays.
- You already use Amplitude for analytics: the two products are being connected so experiment outcomes appear in Amplitude, and Amplitude says customers of both will use them together without having to choose.
- You rely on advanced methods Statsig lists, such as CUPED, holdouts, switchback and non-inferiority tests, and no data scientist has time to rebuild them elsewhere.
- You signed or renewed an Enterprise contract recently, so a migration now costs more than waiting to see the roadmap.
Switch if…
- Your renewal is coming and procurement wants a vendor whose owner and roadmap have not changed twice since 2025.
- Your Pro overage regularly doubles the $150 base, which happens at 8 million events a month.
- You must self-host for compliance, which GrowthBook, Flagsmith and Unleash allow with an open-source edition.
- You need SSO for a security review and cannot sign an Enterprise contract just for that: Unleash includes it at $75 per seat.
How to choose a Statsig alternative
Statsig sits in three categories at once, and its alternatives usually sit in one or two. Feature flag platforms (LaunchDarkly, Unleash, Flagsmith, ConfigCat) let you ship code switched off and turn it on for 1% of users, a segment or everyone, without a deploy. Experimentation platforms (GrowthBook, Datadog Experiments) split users into variants, measure a metric for each group and tell you whether the difference is real or noise. Product analytics tools count what users do, in funnels, retention curves and replays. Statsig, like PostHog, does all three on the same events. Before you compare prices, write down which of the three you actually used: a team that only toggled features pays for an engine it never opened.
Under the hood they all work the same way. Your app calls an SDK, the SDK evaluates the flag rules (locally from a cached copy, or through a network call), assigns the user to a variant, and logs an exposure. The platform then joins exposures with your metric events and runs a statistical test. A fourth family, client-side web testing for marketers (Optimizely, VWO), changes pages in the browser through a visual editor; it replaces Statsig's marketing experiments, not its product flags, and our Optimizely alternatives and heatmap and A/B testing tools ranking cover it. Our best feature flag tools ranking goes deeper on the first category, and our product analytics ranking on the third.
What to demand from any replacement
A statistics engine you can explain and, ideally, reproduce (Bayesian or frequentist, with sequential testing so you can peek without fooling yourself), variance reduction such as CUPED for small effects, guardrail metrics that stop a rollout that hurts revenue, a sample ratio mismatch check, local flag evaluation so a check adds no network call, SDKs for every language you ship, approvals and audit logs, and a way to get raw exposure data out. If you came from Statsig's warehouse-native mode, require the same on the new tool, or your data team loses its single source of truth.
Know the unit behind each price
These tools bill on different units, and the unit decides your bill more than the list price. A metered event (Statsig) is a logged event, an ingested metric or an exposure on a flag with Metric Lifts enabled; 0% and 100% rollouts are not charged. A flag request (PostHog, Flagsmith, Unleash) is a call from an SDK to fetch or evaluate flags: one per app load on a single-page app, one per page view on a server-rendered site, or one per poll for a server. A client-side MAU (LaunchDarkly) is a unique user key evaluated in a month. A service connection (LaunchDarkly) is one server SDK instance connected to one environment, so 20 pods in production and staging are 20 connections. A seat (GrowthBook, Unleash) is a person who logs in. A launched experiment (Datadog) is a test past 500 subjects and 3 days. A config download (ConfigCat) is one fetch of the flag file, however many evaluations follow. One user with ten page views is one MAU, but can be ten requests and twenty events.
Price three scenarios, not one
Take today's volume, then double it, then add a launch spike. At 10 million events a month, Statsig Pro costs $400; GrowthBook's price for 10 people stays $400 in seats, plus $240 if you use its managed warehouse and any CDN overage past 2 million requests; PostHog depends on flag requests, about $235 at 5 million; LaunchDarkly at 50,000 monthly users costs $408.17 billed yearly before service connections. The ranking of these tools flips with your shape of traffic, which is why a single headline price misleads.
Moving off Statsig without breaking production
Inventory first: list live gates, dynamic configs and running experiments, and archive the dead ones. Let running experiments finish, and export their results before the switch. Wrap flag calls behind OpenFeature or a thin internal function, so the vendor becomes a provider you can swap. Move flags service by service, keeping Statsig as a fallback, because a flag that disappears returns its default value in production. Run an A/A test on the new tool for a week: if it finds a winner between two identical variants, your bucketing or tracking is wrong. Teams that also leave Amplitude can read our Amplitude alternatives, and teams weighing a return to LaunchDarkly our guide to switching away from LaunchDarkly, which lists its costs from the other side.
- Statistics engine with sequential testing and CUPED
- Local flag evaluation in every SDK you use
- Billing unit you can forecast on your own traffic
- SSO and approvals on the plan you can afford
- Raw exposure export or warehouse-native mode
- OpenFeature provider to keep the next exit cheap
Statsig alternatives: your questions
Is Statsig shutting down?
No. Amplitude committed to maintain and develop the Statsig platform across cloud and warehouse deployments, including support for existing customers. Its stated plan, published in 2026, was to connect the two products at the data layer, so Amplitude events and cohorts work inside Statsig and Statsig results appear in Amplitude, with customers of both able to use them together by the end of the third quarter of 2026. What changes is the long-term direction, not the service today.Can I keep my code if I switch vendors again later?
Yes, if you route flag calls through OpenFeature, a vendor-neutral feature flag API run as an incubating project of the Cloud Native Computing Foundation. Your code calls OpenFeature, and a provider connects it to Statsig, GrowthBook, LaunchDarkly or another vendor. The Statsig providers listed for .NET, Go and Java servers are community-maintained, so check their activity before you depend on them.Which Statsig alternatives can I self-host, and under which license?
Three in this list on an open-source edition, under different licenses. GrowthBook's code is MIT outside its enterprise directories, Flagsmith's platform is mostly BSD-3-Clause, and Unleash's server is under the AGPL-3.0, which requires you to share changes if you offer a modified version over a network. PostHog's code is MIT outside its ee directory, but self-hosted deployments are officially unsupported, so treat that route as unsupported. ConfigCat's Dedicated plan can also run on-premise, at $4,500 a month billed yearly.What is CUPED, and do these alternatives support it?
CUPED is a variance reduction method: it uses each user's behavior before the experiment to remove noise, so a test reaches a reliable result with fewer users or less time. Statsig lists it among its advanced techniques. GrowthBook includes it on Pro, PostHog lets you enable it per experiment, and Datadog Experiments ships an extended version it calls CUPED++.Do feature flag checks slow down my app?
Not when the SDK evaluates locally. Server-side SDKs from Unleash fetch the flag configuration, cache it in memory and evaluate flags locally, and Unleash says a service can evaluate a flag a million times per second without an extra request. GrowthBook says its SDKs download a payload at startup and evaluate locally, adding sub-millisecond latency. Client-side SDKs still make a network call when they load.What happened to Split.io?
Harness bought it. Harness completed the acquisition of Split Software in 2024 and now sells it as Feature Management & Experimentation, a module listed on the Harness pricing page next to CI, CD and infrastructure as code. It makes the most sense for teams already running Harness pipelines; outside that ecosystem, you are signing up for a platform, not a flag tool.How do I estimate my volume before picking a new tool?
Start from your Statsig console: in Project Settings, the Usage & Billing tab shows your metered events. Split that number into exposures and logged events, then count monthly users, server instances and the people who need a login. For PostHog, the vendor's own advice is to install it, set a $0 billing limit and read the projection after a few days.What is warehouse-native experimentation?
It means the platform computes experiment results inside your own data warehouse instead of copying your events into its servers. GrowthBook connects to Snowflake, BigQuery, Redshift, Databricks and ClickHouse with read-only access, so every query is visible and a metric can be added after a test starts. On Statsig, warehouse-native deployment requires an Enterprise contract.