Panel rating · 6 judges · How to read the stars

Category median

Sovereignty: 2 of 6 dimensions proven

0–5 in half steps. 5 means the rubric's top anchor is met on the evidence.

by GrowthBook, Inc. · www.growthbook.io

Compare with Statsig → Report an error on this page Is this your product? →

Link to this page

GrowthBook: AI-assisted analysis on whats-best.ai GrowthBook: AI-assisted analysis on whats-best.ai (dark)

Add ?theme=dark to the image address for dark backgrounds. The badge shows the category and the month of the latest evaluation, and EU-Made where that is the product’s provenance. It never shows a score.

GrowthBook at a glance

Vendor
GrowthBook, Inc.
Headquarters
Not stated on the vendor’s pages as of 16 Nov 2026
Data hosting
Not stated on the vendor’s pages as of 16 Nov 2026
Free plan
Yes, per the vendor’s own page source, captured 19 Oct 2026
Panel verdict
The AI panel rates GrowthBook 1.5/5, the mean of 6 judges. GrowthBook, from GrowthBook, Inc. how scores are made
As of
Vendor pages captured 22 Sep 2026 to 16 Nov 2026; evaluated by the panel 17 Nov 2026. all sources

GrowthBook alternatives, European ones first →

Read this page as one judge. Each weighs the same scores by what they care about.

The panel's verdict

GrowthBook, from GrowthBook, Inc. of Anaheim, California, is a warehouse-native platform for experimentation, feature flags and product analytics. It is strongest on statistical rigour and data and integrations, where scores reach 8-9: both Bayesian and frequentist engines are documented with stated priors, CUPED and Holm-Bonferroni or Benjamini-Hochberg corrections, and metrics are SQL computed in your own Snowflake, BigQuery, Databricks or Redshift, and the vendor states no PII ever leaves your data warehouse. Sovereignty is weakest at 2: the privacy notice states personal information is transferred to and processed in the United States, and self-hosting is the only stated way to keep data off the vendor's infrastructure. The bench splits sharply — three judges scored 0 on every criterion while the other three scored between 2 and 9, though the rationales behind the zeros read favourably — and consent and tracking splits at 5-7 over the default that persists the gbuuid cookie on page load, before consent. Pricing transparency scores 7: Starter is free up to 3 users, Pro is $40 / seat / month up to 30 users, overages published.

Report an error

Speaks for it

  • Both Bayesian and frequentist engines are documented with stated priors, CUPED, sequential testing and a choice of Holm-Bonferroni or Benjamini-Hochberg corrections
  • Metrics are defined in SQL and computed in your own Snowflake, BigQuery, Databricks or Redshift, and the vendor states no PII ever leaves your data warehouse
  • The Cloudflare edge proxy rewrites HTML before it ships to cut flicker, and the default blocking API call has two documented fixes — a KV push model via webhooks or a just-in-time cache
  • Starter is free up to 3 users with unlimited traffic, experiments and feature flags, and Pro's usage overages are published — 2M events/mo (then $30/M), 2M/mo (then $10/M) CDN requests, 20GB/mo (then $1/GB) bandwidth
  • The core is open source under MIT with a self-hosting option — same product on your own cloud or on-premises, no data or PII leaving your system for analysis — though some directories sit under a separate Enterprise license

Report an error

Held against it

  • Sovereignty scores sit at 2 — the contracting entity is GrowthBook, Inc. of Anaheim, California, and the privacy notice states personal information is transferred to and processed in the United States
  • We found no public information on a data processing agreement, a subprocessor list or an EU region for GrowthBook Cloud, leaving self-hosting as the only stated alternative
  • By default the gbuuid cookie is persisted by the injected front-end SDK on page load, before any consent answer, and we found no public information on consent-management-platform integrations, the cookie's lifetime or IP anonymisation
  • Sequential testing sits behind Pro and Enterprise, multiple-comparison corrections apply only to the frequentist engine, and adjusted p-values are not exportable via the API
  • Enterprise pricing is Custom, and we found no public information on VAT treatment or minimum contract term

Report an error

Best for

  • You want experiment results computed in your own data warehouse with visible SQL, so any uplift can be re-checked against your own data
  • You deliver at the edge and want visual and redirect experiments rewritten into HTML before it ships, with a documented KV cache to remove the default blocking call
  • You want per-seat pricing with unlimited traffic and experiments — Starter free up to 3 users, Pro at $40 / seat / month up to 30 users
  • You can run your own infrastructure, since self-hosting — same product, no data or PII leaving your system for analysis — is the documented path to keeping visitor data under your control

Report an error

Avoid if

  • You must keep EU visitor data on EU infrastructure as a cloud customer — the captured pages show a US contracting entity, US processing and self-hosting as the only stated alternative
  • Your page-speed budget cannot absorb the default blocking edge API call per request and you lack the engineering to set up the documented KV push or just-in-time cache
  • You need the default setup to persist no identifier before consent — the gbuuid cookie is persisted on page load unless you change the documented configuration

Report an error

The scores

Panel median The median of the judges’ scores per criterion. Each label opens its reasoning below.

Experiment types & delivery

panel disagrees Show reasoning
How this is scored

What can be tested and where: client-side changes through an editor, server-side and feature experiments through SDKs, multivariate and multi-page tests, and personalisation — judged on what the documentation shows rather than on the feature grid.

0 — Not an experimentation tool; the product cannot split traffic or serve variants (analytics, heatmaps or session recording only).

3 — Client-side A/B tests built in a visual editor, with a traffic split and targeting by URL or device.

5 — A/B, split-URL and multivariate tests through a visual or code editor, audience targeting on behaviour and attributes, and either rule-based personalisation or server-side experiments through a documented SDK.

8 — Client-side and server-side experiments through documented SDKs for common languages, multi-page tests, feature-flag experiments sharing one audience and metric model with web tests, mutually exclusive experiment groups, holdouts, and personalisation that can itself be tested against a control.

10 — One experimentation programme across every surface: web, app, server and edge from the same platform, experiment interactions managed, a documented experiment lifecycle from hypothesis to archived result, and a library of past results a team can search.

Report an error

The Growth Lead

The AI Visual Editor lets my team build front-end tests without a developer, and URL redirects, multivariate, sequential and bandit tests are all documented alongside feature-flag experiments that share the same targeting attributes and metric model. Server-side and edge delivery run through 24+ SDKs, mutually exclusive tests work through namespaces, and holdouts and experiment templates are documented. We found no public information on multi-page tests or on personalisation measured against its own control, which keeps this a level below the very top. 6 6 7 11 16

Report an error

The E-Commerce Manager

A genuine experimentation platform with the delivery spread I need: A/B tests through feature flags, the AI visual editor, URL redirects and custom assignment, delivered client-side, server-side through 24-plus documented SDKs, and at the edge on Cloudflare Workers where visual and redirect experiments run without flicker. Multivariate tests, namespaces for mutually exclusive experiments, holdouts and bandits are all documented. The captured pages show no personalisation rules that can themselves be tested against a control and no multi-page tests, so I score just below the top band rather than guess. 6 6 12 7 2

Report an error

The Product Engineer

The delivery model is exactly what I want to see: experiments run server-side, client-side and at the edge through 24+ documented SDKs, with experiment rules on feature flags sharing the warehouse metric model, namespaces for mutual exclusion, holdouts, bandits and URL redirects. Edge delivery is a documented first-class path, not a bolt-on. We found no public information on personalisation that can itself be tested against a control, or a searchable library of past results. 6 6 12 7 16

Report an error

The CRO Consultant

Client-side, server-side and edge experiments are all documented through 24+ SDKs, feature-flag experiments share one audience and metric model with web tests, and namespaces, holdouts, experiment templates, power analysis and bandits round out the programme. We found no public information on personalisation that can itself be tested against a control, or multi-page tests, which keeps this one step short of a complete programme across every surface. 6 6 12 2

Report an error

The Data Protection Officer

Client-side, server-side and edge experiments are all documented, delivered through feature flags, an AI Visual Editor, URL redirects or the team's own assignment, with 24+ SDKs, namespaces for mutually exclusive experiments, holdouts and experiment templates. We found no public information on rule-based personalisation or multi-page tests, which keeps this short of the top band. 6 6 7 16

Report an error

The Skeptic

Client-side visual editor, URL redirects, feature-flag experiments, server-side inline experiments and an edge proxy are all documented, with namespaces for mutually exclusive tests, holdouts on paid plans, multivariate and bandit tests, and 24+ SDKs across web, mobile, server and edge. I found no public information on multi-page tests or on personalisation that can itself be tested against a control, so it falls just short of the next level. 6 6 7 16

Report an error

Statistical method & guardrails

panel disagrees Show reasoning
How this is scored

Which statistics decide the winner and what protects the customer from misreading them. Scored on what the vendor documents: the method by name, how peeking and multiple comparisons are handled, and whether sample ratio mismatch is detected.

0 — No statistical evaluation; the product shows raw counts per variant and leaves the decision to the user.

3 — A results report per variant with the method named (frequentist or Bayesian) and a significance or probability-to-beat figure.

5 — A documented method with confidence or credible intervals shown per variant, and a sample-size or test-duration calculator or stated guidance on when a result may be read.

8 — Sequential testing or another documented protection against peeking, plus correction for multiple metrics or variants, sample ratio mismatch detection, variance reduction such as CUPED, and guardrail metrics that can stop a harmful test.

10 — The statistics are auditable: methodology published in enough detail to reproduce a result, the choice of method explained per use case, raw per-visitor data available for independent re-analysis, and the interface refuses to present an underpowered result as a conclusion.

Report an error

The Growth Lead

Both a Bayesian and a frequentist engine are documented by name, with priors, default settings and the two-sample t-test spelled out, plus sequential testing in the paid tiers, Holm-Bonferroni and Benjamini-Hochberg corrections, CUPED, sample ratio mismatch detection and guardrail metrics with Slack alerts. Because the implementation is open source and every metric query shows its SQL against our own warehouse, I can defend any uplift to finance line by line. Minimum data checks that stop us reading a 5-vs-2 result as a win, plus suspicious-uplift detection, round out the guardrails; adjusted p-values are documented as available only in the UI or CSV. 8 9 2 7 7

Report an error

The E-Commerce Manager

Both engines are documented by name — Bayesian by default with an uninformative prior and an optional Normal prior at mean 0, standard deviation 0.3, frequentist two-sample t-tests — alongside sequential testing, Holm-Bonferroni and Benjamini-Hochberg corrections with the correction family spelled out, CUPED, post-stratification, sample ratio mismatch detection, guardrail metrics with alerts, and minimum-data checks that stop you reading five-versus-two conversions as a winner. Because results are computed in your own warehouse with visible SQL and the implementation is open source, I can re-check a result in my own sales data before I ship a winner; I stop one short of full marks because we found no public information on the interface refusing to present an underpowered result as a conclusion. 8 9 2 7 6

Report an error

The Product Engineer

Both engines are documented in depth — Bayesian with a proper-prior toggle, frequentist two-sample t-tests — plus sequential testing, CUPED, SRM detection, guardrail metrics, and a choice of Holm-Bonferroni or Benjamini-Hochberg corrections with the correction family spelled out. The implementation is open source under MIT and results are reproducible and re-analysable on your own warehouse data. Caveats: sequential testing is gated to Pro and Enterprise, corrections apply only to the frequentist engine, and adjusted p-values are not exportable via the API. 7 8 9 2

Report an error

The CRO Consultant

Both engines are documented to the level a client's data team can audit: the default Bayesian prior (improper, uninformative, with an optional Normal prior of mean 0 and standard deviation 0.3), the frequentist two-sample t-test on relative percent change, plus sequential testing, CUPED, SRM detection, guardrail metrics and a choice of Holm-Bonferroni or Benjamini-Hochberg corrections. The methodology is published in detail and the implementation is open source, so results are reproducible; the one gap is that adjusted p-values are available in the UI and CSV downloads but not through the API. 8 9 2 7

Report an error

The Data Protection Officer

Both engines are documented — Bayesian with a stated default prior and frequentist with two-sample t-tests — alongside sequential testing, CUPED, sample ratio mismatch detection, guardrail metrics with alerts, and a choice of no correction, Holm-Bonferroni or Benjamini-Hochberg. The methodology is published in enough detail to follow, minimum-data checks guard against early reads, and the implementation is open source. 8 9 2

Report an error

The Skeptic

This is the rare case where the probability-to-beat figure ships with its assumptions written down: the default Bayesian engine uses an improper uninformative prior with an optional Normal proper prior (mean 0, standard deviation 0.3), the frequentist engine is a two-sample t-test on relative change, and multiple-comparison handling is named — Holm-Bonferroni and Benjamini-Hochberg with the test family defined — alongside sample ratio mismatch detection, CUPED, sequential testing, minimum data thresholds and guardrails with alerts and auto-rollback. The statistics engine is open source and metrics are SQL against the customer's warehouse, so results are reproducible and re-analysable. I stop short of the very top because sequential testing and the power calculator sit behind paid tiers, corrections exist only in the frequentist engine, and adjusted p-values are not exportable through the API. 8 9 2 7

Report an error

Snippet performance & flicker

panel disagrees Show reasoning
How this is scored

The cost the client-side snippet imposes on the page it tests: blocking load, flicker of original content, script weight and the effect on Core Web Vitals — scored on what the vendor measures and publishes, not on "lightning fast".

0 — A synchronous, render-blocking snippet stated as the only way to install the product, with no flicker handling.

3 — An asynchronous or anti-flicker snippet with a timeout, described in the installation documentation.

5 — Asynchronous loading and flicker handling explained with their trade-off, the snippet delivered from a CDN, and its size or loading behaviour documented.

8 — Published performance figures including the impact on Core Web Vitals, a self-hosting or first-party-domain option for the script, per-project bundles containing only active experiments, and a server-side or edge alternative for flicker-sensitive tests.

10 — Performance is a stated commitment: measured overhead published and maintained, flicker eliminated by edge or server-side rendering as a documented path, and tooling that shows the customer what their own configuration costs the page.

Report an error

The Growth Lead

The Cloudflare edge app is documented to run visual and redirect experiments before HTML ships, "without redraw flicker" or delay, and the default blocking API call per request is openly documented with two ways to eliminate it: a webhook push model to an edge key-value store or a just-in-time payload cache. SDK connection endpoints are served from a global CDN, and self-hosting with your own CDN or the proxy server is offered. We found no published figures on Core Web Vitals impact or measured script weight. 11 12 12 14 2

Report an error

The E-Commerce Manager

Unusually honest documentation of the cost: by default the edge app makes a blocking call to fetch feature values on every request and delays page delivery, and the pages show two documented ways to remove it — pushing the payload into Cloudflare KV via webhooks or a just-in-time cache — plus edge evaluation of visual and redirect experiments so flicker is cut before the HTML ships. Self-hosting, a self-hosted CDN layer and a caching proxy are all documented, which covers my first-party needs; we found no public information on measured Core Web Vitals impact or per-bundle script weight, so I cannot verify what my product page actually pays. 11 12 11 14 2

Report an error

The Product Engineer

The documentation is unusually honest: the edge app's default blocking payload fetch is stated outright, along with two ways to eliminate it — a push model to Cloudflare KV via SDK webhooks or a just-in-time cache — and injected bootstrapped SDKs that need no extra network requests. Visual and redirect experiments run at the edge before HTML ships, which is the flicker answer I actually want. We found no public information on measured Core Web Vitals impact or per-project bundles with only active experiments. 11 12 14 2

Report an error

The CRO Consultant

The edge documentation is unusually candid that the default Edge App makes a blocking API call on each request that delays page delivery, and then shows the escape routes: a webhook push to Cloudflare KV or a just-in-time payload cache to eliminate it, plus edge evaluation that runs visual and redirect experiments without redraw flicker before HTML ships. Running your own CDN or the GrowthBook Proxy is offered. We found no published figures for snippet weight or the impact on Core Web Vitals. 11 12 14 2

Report an error

The Data Protection Officer

The edge app is documented to rewrite HTML before it reaches the browser so visual and redirect experiments run without flicker, and the default blocking call to the API on each request is stated openly, with a Cloudflare KV push model or just-in-time cache to remove it plus a self-hosted CDN or proxy option. We found no public information on published Core Web Vitals impact figures or per-project bundles containing only active experiments. 11 12 11 2

Report an error

The Skeptic

Flicker is attacked at the architecture level — the edge proxy evaluates flags and rewrites HTML before it ships, with visual and redirect experiments documented as running without redraw flicker — and the documentation is unusually candid that the default edge path makes a blocking call per request, giving two documented ways to eliminate it (webhook push to a key-value store or a just-in-time cache), plus a self-hosted CDN and proxy option. I found no published measurements of script weight or Core Web Vitals impact, so the "50% smaller SDKs" claim cannot be checked against numbers. 11 12 12 2

Report an error

Analytics, data export & integrations

panel disagrees Show reasoning
How this is scored

Getting results and raw data out: integration with analytics and tag management, export of visitor-level results, warehouse-native analysis, and an API — because an experiment result that cannot be checked in the customer's own data is a claim, not a finding.

0 — A closed dashboard: the vendor states that results cannot be exported or sent to another tool.

3 — Results exportable as a CSV file or report, and an integration with at least one named analytics tool.

5 — Integrations with common analytics and tag managers (Google Analytics, Adobe Analytics, Google Tag Manager or similar), export of results, and a documented API or webhooks.

8 — Visitor-level raw data export or streaming to a data warehouse, warehouse-native analysis on the customer's own metrics, CDP integration for audiences, and an API for managing experiments and reading results with stated limits.

10 — The platform treats the customer's warehouse as the source of truth: metrics defined once and computed there, full historical experiment data exportable in open formats, and a versioned API a team can build its own programme tooling on.

Report an error

The Growth Lead

This is warehouse-native: Snowflake, BigQuery, Databricks and Redshift are queried directly, metrics are SQL I can read, and no PII ever leaves our warehouse for analysis — exactly what I need when finance asks to see the raw numbers behind an uplift. A full REST API covers experiments, results, features, metrics and segments on v1 and v2 paths, results export as CSV, and webhooks plus Segment, GA4 and Google Tag Manager tracking integrations are documented. What holds it back slightly: adjusted p-values are documented as available only in the UI or CSV, and we found no stated API rate limits. 15 5 14 9 11

Report an error

The E-Commerce Manager

The customer's warehouse is the source of truth: metrics are defined in SQL and computed in Snowflake, BigQuery, Databricks or Redshift with every query visible, no PII leaves your system, and a full REST API covers creating experiments and pulling results into your own reporting, alongside webhooks, dbt sync, GA4, Google Tag Manager, Segment and a Shopify integration. One stated limit: adjusted p-values are available in the UI and CSV downloads but not through the API, and we found no public information on stated API rate limits. 14 15 7 13 9 2

Report an error

The Product Engineer

Warehouse-native done right: metrics defined once in SQL and computed in Snowflake, BigQuery, Databricks or Redshift with every query visible, no PII leaving the customer's warehouse, and a full REST API with versioned endpoints for creating experiments and pulling results — enough to build internal tooling on. Segment, GA4 and Google Tag Manager tracking are wired in by default. We found no public information on stated API limits. 5 14 11 16

Report an error

The CRO Consultant

This is the architecture I look for: metrics are defined in SQL and computed in the client's own Snowflake, BigQuery, Databricks or Redshift, with visible SQL and reproducible results, so an experiment claim can always be checked in the client's own data rather than trusted. A full REST API covers creating experiments and reading results, alongside webhooks, dbt sync, Slack, and default tracking into Segment.io, GA4 and Google Tag Manager. We found no public information on stated API rate limits. 15 14 7 11

Report an error

The Data Protection Officer

Analysis is warehouse-native: metrics are written in SQL and computed against the customer's own Snowflake, BigQuery, Databricks or Redshift data, with visible queries and reproducible results, and a full REST API covers creating experiments, reading results and managing features, alongside webhooks, an MCP server and Segment.io, GA4 and Google Tag Manager tracking hooks. Adjusted p-values are available in the UI and downloadable CSVs but not through the API, the one stated export limit. 14 16 7 9

Report an error

The Skeptic

Warehouse-native is an architecture, not a slogan: metrics are SQL over Snowflake, BigQuery, Databricks or Redshift with visible queries, the vendor states no PII ever leaves the customer's warehouse, and there is a REST API spanning experiments, results, features, segments and fact tables plus webhooks, an MCP server and default tracking into Segment.io, GA4 and Google Tag Manager, with results exportable as CSV. I found no public information on stated API rate limits or on a customer-data-platform audience integration, which keeps the top level out of reach. 5 9 14 16

Report an error

European sovereignty panel opinion

Show reasoning
How this is scored

Where visitor data is processed and stored and who the contracting entity is. Independently sourced by the sovereignty pipeline; weighted higher here than in categories that hold only the customer's own data, because the script runs on every visitor to the customer's site and their behaviour is what the platform records.

0 — Vendor, contracting entity and processing of visitor data all shown to be outside the EU, with no transfer safeguard stated.

3 — A named contracting entity and a privacy policy or data processing agreement; EU data residency offered as an option or add-on, or the hosting location not stated.

5 — EU hosting or EU processing of visitor data stated as standard, together with an EU contracting entity or a published subprocessor list, and a data processing agreement offered.

8 — EU contracting entity, EU hosting on named infrastructure including the delivery of the snippet, a subprocessor list published with locations, and a data processing agreement covering the visitor data the script collects.

10 — Sovereign end to end and evidenced: vendor, entity, hosting, snippet delivery and every subprocessor European, certification published, and no visitor data reaching a non-EU party at any point.

Report an error

The Growth Lead

The contracting entity is GrowthBook, Inc. of Anaheim, California; the privacy notice states that personal information is transferred to and processed in the United States, the cloud runs on AWS with no EU region named in the captured pages, and the terms state the website is provided for use only by persons located in the United States. We found no public information on a data processing agreement, a subprocessor list or transfer safeguards. The documented escape is self-hosting — the same product on our own European infrastructure, air-gapped if we wish, with no data or PII leaving our system — but as a cloud buyer we would contract with a US entity under California law. 3 4 15 1

Report an error

The E-Commerce Manager

GrowthBook, Inc. is a California entity and the privacy notice states personal information is transferred to and processed in the United States, with the cloud service hosted on AWS; we found no public information on a data processing agreement, an EU region option, or a transfer safeguard for the cloud service. The one documented path that keeps EU visitor data in Europe is self-hosting — "No data or PII leaves your system" — which for a shop serving German visitors is the mitigation I would be buying, and it costs me engineering to run. 3 4 15

Report an error

The Product Engineer

The contracting entity is GrowthBook, Inc. of Anaheim, California, and the privacy notice states personal information is "transferred to and processed in the United States"; we found no public information on an EU region for the cloud product or on a data processing agreement. SOC 2 Type II and ISO 27001 are published, and self-hosting — "Same features, same product, 100% on your cloud or on prem" — is the documented alternative for keeping data off the vendor's infrastructure. 3 5 1 15

Report an error

The CRO Consultant

GrowthBook Cloud is hosted on AWS in the United States and the privacy notice states that personal information is transferred to and processed in the United States; the contracting entity is GrowthBook, Inc. of Anaheim, California, and I found no public data processing agreement, subprocessor list or EU hosting option for the cloud offering — the terms even restrict use to persons located in the United States. Self-hosting, including air-gapped on-premises where no data or PII leaves the system, is a genuinely sovereign path, but the customer is doing that work rather than the vendor. 3 15 4 4

Report an error

The Data Protection Officer

The contracting entity is GrowthBook, Inc. of Anaheim, California; the privacy notice states that personal information is held with third-party providers located in the United States and that using the service is consent to transfer to and processing in the United States, and the terms state the site is provided for use only by persons located in the United States. We found no public information on an EU hosting option for the cloud service, a data processing agreement, a published subprocessor list naming the CDN, or any transfer safeguard; self-hosting — the same product on your own cloud or on-premises, with no data or PII leaving your system for analysis — is the only documented way to keep visitor data under European control. 3 4 1 15

Report an error

The Skeptic

The contracting entity is GrowthBook, Inc. of Anaheim, California, cloud hosting is GrowthBook on AWS, and the privacy notice states personal information is held and processed in the United States; I found no public information on a data processing agreement, a subprocessor list with locations, or an EU hosting option, and the AI features name US providers (Anthropic, Google, OpenAI, xAI). Full self-hosting with the statement that no data or PII leaves the customer's system is the only sovereignty-relevant path the captured pages show, which lifts it slightly off the floor. 3 5 15

Report an error

Pricing transparency

panel disagrees Show reasoning
How this is scored

A category priced by traffic — monthly tracked users, visitors or impressions — where the tier a site lands in depends on numbers the buyer has to estimate. Whether a buyer can compute the real annual cost including traffic limits, overage, server-side or personalisation modules and seats — from public pages alone.

0 — No public prices at all; every tier is a sales conversation.

3 — A public "from" price, a free tier or one priced tier, while the other tiers or the traffic metric behind the price are not stated.

5 — Prices public per tier with the billing period and the traffic metric (tracked users, visitors or impressions) and its limits stated; modules such as server-side SDKs or personalisation, or overage, may still be "contact sales".

8 — Every tier and module priced publicly with the traffic metric defined, limits, overage rates, seats, minimum term and VAT treatment stated.

10 — Complete price computability: annual invoice derivable for a given traffic volume, set of modules and team size, with overage and every add-on published.

Report an error

The Growth Lead

Starter is free with unlimited traffic and up to 3 users, Pro is "$40 / seat / month" with up to 30 users and 3 projects, and overages are published: "2M events/mo (then $30/M)" for the managed warehouse, "2M/mo (then $10/M)" CDN requests and "20GB/mo (then $1/GB)" bandwidth. That lets me derive a real annual figure for a given team size and event volume, which is more than most vendors give me. Enterprise is custom, and we found no public information on minimum term or VAT treatment. 2 2 2 2

Report an error

The E-Commerce Manager

Pricing is per seat, not per visitor: Starter is free with unlimited traffic and experiments, Pro is "$40 / seat / month" up to 30 users, and the usage-based parts publish limits and overage — managed warehouse 2M events per month then "$30/M", CDN requests 2M per month then "$10/M", bandwidth 20GB then "$1/GB". For a shop whose traffic triples in December, a seat price with unlimited experiments is a gift; I hold it one band down because Enterprise is custom and we found no public information on VAT treatment, minimum term, or what the "5x" AI usage limits in Pro actually mean. 2 2 2 2

Report an error

The Product Engineer

Starter is free for up to 3 users and 1 project, Pro is $40 per seat per month up to 30 users and 3 projects, and the usage modules are itemised with overage rates — managed warehouse events with $30 per million beyond the allowance, CDN requests at $10 per million and bandwidth at $1 per gigabyte. Enterprise remains custom, and we found no public information on VAT treatment or minimum term. 2 2 2

Report an error

The CRO Consultant

Pricing is per seat rather than per visitor — Starter free with 3 users and 1 project, Pro at $40 / seat / month up to 30 users and 3 projects, Enterprise custom — with the metered limits published so I can model a bill before contract: 2M managed-warehouse events per month then $30 per million on Pro, 2M CDN requests then $10 per million, and 20GB bandwidth then $1 per GB, even though traffic itself is advertised as unlimited. We found no public information on VAT treatment or minimum contract term, and Enterprise pricing is a sales conversation. 2 2 2

Report an error

The Data Protection Officer

Starter is free with limits stated (up to 3 users, 1 project, unlimited traffic), Pro is $40 / seat / month with up to 30 users and 3 projects, and overage is published: CDN requests 2M/mo then $10/M, CDN bandwidth 20GB/mo then $1/GB, managed warehouse 2M events/mo then $30/M. Enterprise is Custom, and we found no public information on minimum term or VAT treatment. 2 2 2

Report an error

The Skeptic

Starter is free for up to 3 users, Pro is "$40 / seat / month" up to 30 users, feature flags, experiments and traffic are unlimited on every plan, and the usage limits and overage rates are published — managed warehouse events at "2M events/mo (then $30/M)", CDN requests "(then $10/M)" and CDN bandwidth "(then $1/GB)". Enterprise is "Custom" and I found no public information on VAT treatment or minimum terms, so a complete annual invoice is not derivable for every configuration. 2 2 2

Report an error

European sovereignty — proven facts

2 of 6 dimensions proven

Built only from facts shown on the vendor's own pages. A dimension we could not prove is left open, not scored as zero.

Hosting location Not determined unverified
Why unverified?

No server region is named for the AWS hosting and no EU region option is offered for GrowthBook Cloud; self-hosting is the only stated alternative for keeping data off GrowthBook's infrastructure.

— uncited Report an error
Hosting options Self-hostable 1/1 pts 5 Report an error
Open source Open source (permissive) unverified
Why unverified?

Open-core model: several directories are governed by the separate commercial GrowthBook Enterprise License, so part of the product is not open source.

1/1 pts 16 Report an error
Data processing agreement Not determined — uncited Report an error
Transfer impact assessment Not determined — uncited Report an error

Questions about GrowthBook

Is GrowthBook European?

No. whats-best.ai classifies GrowthBook as Rest of world.

Where does GrowthBook store data?

Not stated on the vendor’s pages as of 16 Nov 2026.

Does GrowthBook have a free plan?

Yes. GrowthBook has a free plan, per the vendor’s page captured 19 Oct 2026.

What are the best alternatives to GrowthBook?

As of 19 Nov 2026, the panel rates Convert Experiences (3.5/5), Varify.io (3/5) and Webtrends Optimize (3/5) highest among the other 45 Conversion Optimization products evaluated. Ratings that look equal are rounded to half stars; the order uses the unrounded rating, then the name. The highest-rated European alternative is Varify.io (EU-Made, Germany). Ratings are the AI panel’s opinion, formed from public evidence: how scores are made. The full list, with European products first: GrowthBook alternatives.

Where this could be wrong

What we left out

A claim that does not survive our checks costs us the claim, not the page. This is what was taken off this one.

Sources (16)

The pages every claim on this page was read from — each one checked, dated, and kept verifiable.

  1. 1 Vendor homepage www.growthbook.io Checked 16 Nov 2026 +4 earlier captures: 9 Nov 2026, 2 Nov 2026, 26 Oct 2026, 19 Oct 2026 Details →
  2. 2 Pricing page www.growthbook.io Checked 16 Nov 2026 +3 earlier captures: 2 Nov 2026, 19 Oct 2026, 22 Sep 2026 Details →
  3. 3 Privacy policy www.growthbook.io Checked 16 Nov 2026 Details →
  4. 4 Terms of service www.growthbook.io Checked 16 Nov 2026 +1 earlier capture: 19 Oct 2026 Details →
  5. 5 Security / trust page www.growthbook.io Checked 16 Nov 2026 Details →
  6. 6 Experiment types & delivery — found from sitemap docs.growthbook.io Checked 16 Nov 2026 +1 earlier capture: 1 Oct 2026 Details →
  7. 7 Experiment types & delivery — found from sitemap www.growthbook.io Checked 16 Nov 2026 +3 earlier captures: 9 Nov 2026, 2 Nov 2026, 19 Oct 2026 Details →
  8. 8 Statistical method & guardrails — found from sitemap docs.growthbook.io Checked 16 Nov 2026 Details →
  9. 9 Statistical method & guardrails — found from sitemap docs.growthbook.io Checked 16 Nov 2026 Details →
  10. 10 Consent & visitor tracking — found from sitemap www.growthbook.io Checked 16 Nov 2026 Details →
  11. 11 Snippet performance & flicker — found from sitemap docs.growthbook.io Checked 16 Nov 2026 +1 earlier capture: 1 Oct 2026 Details →
  12. 12 Snippet performance & flicker — found from sitemap docs.growthbook.io Checked 16 Nov 2026 +1 earlier capture: 1 Oct 2026 Details →
  13. 13 Analytics, data export & integrations — found from sitemap docs.growthbook.io Checked 16 Nov 2026 Details →
  14. 14 Analytics, data export & integrations — found from sitemap docs.growthbook.io Checked 16 Nov 2026 Details →
  15. 15 Self-hosting / on-premises installation docs — found from the homepage www.growthbook.io Checked 16 Nov 2026 Details →
  16. 16 Code repository (open-source licence) — found from the homepage github.com Checked 16 Nov 2026 +1 earlier capture: 9 Nov 2026 Details →