Visibility in this ranking can be paid for. Payment gives a vendor more visibility, higher up on the review pages; it never adds a vendor, and it never changes a word of the review. The largest vendors in A/B testing tools
cannot hold places 1 to 3. Places 1 to 3 go to challengers, vendors outside the best-known brands in a category,
wherever the category has them; the established names follow from place 4 in their ranked order. How it works: placement disclosure ·
editorial process.
The 13 tools, reviewed
#1 GrowthBook
Open-source experimentation and flags on your warehouse · United States · growthbook.io
Open source Warehouse-native Per-seat pricing
GrowthBook is the only product here where the traffic never feeds the invoice: unlimited experiments, flags and traffic on the free Starter plan, then $40 per seat once a team passes three people. Results are computed by queries against your own data warehouse, so the numbers can be audited and rerun. Most of the repository is MIT-licensed and can be self-hosted.
- Pricing: Per seat / month. Starter is free (3 users, 1 project); Pro is $40 per seat a month for up to 30 users; Enterprise is quoted. Free plan: open source.
- Best for: Engineering-led product teams
- Strengths: A/B and multivariate tests; Feature flags; Bayesian, frequentist and sequential engines.
Where it falls short
A warehouse is the assumed starting point, and teams without one must use the managed warehouse option or pipe events in. The visual editor is newer than those of Convert or VWO and sits on the Pro plan.
Wrong for
A marketing team with no analyst and no developer, which will struggle to wire up data sources and goals.
Visit GrowthBook →
#2 Convert
A/B testing for marketers with plain published pricing · United States · convert.com
Visual editor Published pricing Free migration
Convert is the rare vendor in the marketing-side group that prints its prices: Growth starts at $299 a month on annual billing for 100,000 tested users, and every plan includes unlimited experiments in one product with no separate modules. The Growth tier already carries targeting, heatmaps, a QA wizard and a sample ratio mismatch check.
- Pricing: Per tested user / month. Growth from $299 a month billed annually ($399 monthly) for 100,000 tested users a month. Free plan: free trial.
- Best for: Marketing teams leaving VWO
- Strengths: A/B and split URL tests; Multivariate testing (Pro); Full stack and feature flags (Pro).
Where it falls short
Multivariate testing, multi-page tests, feature flags and SSO begin at the Pro plan, so the cheap tier is limited to A/B and redirect tests. Tested-user counts reset monthly and jump in steps as traffic grows.
Wrong for
A product organization that wants experiments on its own warehouse metrics and server code first, or a company of several million monthly users that needs custom contract terms from day one.
Visit Convert →
#3 Kameleoon
Web, feature testing and personalization on one platform · France · kameleoon.com
French vendor Web and server Personalization
Kameleoon documents web experimentation, feature experimentation and personalization as one platform with a shared user manual. Its SDK list runs from JavaScript and React to iOS, Android, React Native, Flutter, Node.js, Python, Java and .NET, which lets a marketing group and an engineering group use the same vendor.
- Pricing: Quote only. Pricing on request; the vendor's pricing page was behind a bot check and not read. Free plan: none.
- Best for: Mixed marketing and product teams
- Strengths: Web experimentation; Feature experimentation; Prompt-based experiment creation.
Where it falls short
Pricing, traffic limits and plan structure were not visible to us, so budgeting needs a quote.
Wrong for
A small site that wants to start on a free plan or pay by card the same day.
Visit Kameleoon →
#4 PostHog
Experiments bundled with analytics, replay and flags · United States · posthog.com
Open source Usage-based All-in-one
PostHog treats an experiment as a feature flag with goals attached, so one SDK serves flags, analytics and test assignment. Both Bayesian and frequentist engines are offered, and a replay of any variant is one click from the results. The free allowance renews monthly and includes 1 million flag requests, which covers many early-stage products.
- Pricing: Usage-based. Free monthly allowance (1 million flag requests, experiments billed with flags), then usage-based billing by product; paid tiers need a card. Free plan: free plan.
- Best for: Engineering teams wanting one suite
- Strengths: Experiments; Feature flags; Bayesian and frequentist engines.
Where it falls short
The experiment interface assumes comfort with events, properties and cohorts, so a non-technical marketer will need help. Billing is spread across products, which makes a single forecast harder.
Wrong for
A conversion-rate team that edits landing pages and wants a point-and-click editor with targeting.
Visit PostHog →
#5 VWO
Web testing, behavior analytics and personalization from Wingify · India · vwo.com
Wingify Visual editor Insights
VWO's plan page lists a visual and code editor, unlimited experiments, variations and metrics on Growth, then guardrail metrics, sample ratio mismatch detection and multivariate tests on Pro, and self-hosting, SSO and an approval workflow on Enterprise. Behavior analytics, surveys and feature management sit around testing in the same account.
- Pricing: Quote only. Growth, Pro and Enterprise plans are listed by feature on the pricing page with no prices; each ends in a demo request. Free plan: free trial.
- Best for: Teams that want testing plus insights
- Strengths: A/B, split URL and multivariate tests; Visual and code editor; Rollout campaigns.
Where it falls short
No plan prices are shown, and the merged company's product map is still changing, so ask what is sold under which login. Multivariate testing, guardrail metrics and cross-domain tests require Pro, and self-hosting is Enterprise only.
Wrong for
A buyer who needs to see a price before a sales call, or a team bothered by a vendor mid-integration.
Visit VWO →
#6 AB Tasty
Experimentation and personalization, now inside Wingify · France · abtasty.com
French origin Wingify Personalization
AB Tasty began in Paris and still lists its own login next to the Wingify login, so existing customers keep their account while the combined company decides what comes next. The Wingify platform pages list web experimentation, personalization, feature management and analytics under the merged brand.
- Pricing: Quote only. Pricing on request; the pricing page was behind a bot check and not read. Free plan: none.
- Best for: European brands with web-first needs
- Strengths: Web experimentation; Personalization; Feature management.
Where it falls short
We could not read its pricing or plan limits. The ownership change means roadmap and packaging may shift, and overlap with VWO raises the question of which product stays.
Wrong for
A new buyer who wants a settled product line and published prices, or a team that needs warehouse-native analysis.
Visit AB Tasty →
#7 Statsig
Experiments, flags and analytics with heavy statistics · United States · statsig.com
Free tier Statistics OpenAI deal
Statsig packs the most statistical tooling in this ranking at its price: Bayesian and frequentist modes, CUPED, winsorization, multiple-comparison corrections, holdouts and bandits. The free Developer plan has unlimited seats and 2 million events a month, and Pro starts at $150. Warehouse-native deployment is an Enterprise option. Statsig's own blog says it is joining OpenAI, with closing subject to regulatory approval.
- Pricing: Usage-based. Developer plan is free (2M events a month). Free plan: free plan.
- Best for: Data-literate product teams
- Strengths: A/B/n and multivariate tests; Feature flags; CUPED.
Where it falls short
Event-based billing can climb with a busy product, and approvals and warehouse deployment are Enterprise features. The marketing-style page editor is thinner than Convert's.
Wrong for
A buyer who needs a vendor with no pending change of control, or a marketer who wants a mature visual editor.
Visit Statsig →
#8 Amplitude
Experiments built on a product analytics base · United States · amplitude.com
Analytics suite Web and feature tests Event-based
Amplitude's experimentation sits next to its analytics, session replay and feature management, so a product manager can go from a funnel drop to a test in the same tool. Its page says experiments can be designed through a guided interface without engineering help, with data quality checks and duration estimates. Free and Plus plans exist, though Free lists experiments as limited.
- Pricing: Usage-based. Free plan includes limited experiments. Free plan: free plan.
- Best for: Existing Amplitude analytics customers
- Strengths: Feature experimentation; Web experimentation; Feature management.
Where it falls short
The Free plan limits experiments, the best experiment features need the sales-led tiers, and billing is event-based with separate session replay and agent-session meters.
Wrong for
A team that does not use Amplitude analytics and only wants experiments.
Visit Amplitude →
#9 Harness Feature Management & Experimentation
Flags, release monitoring and experiments in a pipeline · United States · harness.io
Pipeline-native Release safety AI configs
Harness frames experimentation as the measurement step after a flag rollout: a flag is created, policies and approvals apply, a gradual rollout watches latency and errors, and an experiment then reports lift. Its page also describes AI configs and AI experiments for swapping models and prompts at runtime. That pipeline view is its strength.
- Pricing: Quote only. Sold as a module in Harness plans; the DevOps plan lists Feature Management & Experimentation; no standalone price was seen. Free plan: none.
- Best for: Engineering orgs on Harness
- Strengths: Feature flags; Experimentation; Release monitoring.
Where it falls short
The experiment tooling has no visual page editor for marketing tests, and the plan structure bundles it with other Harness modules instead of a simple price. We saw no free tier.
Wrong for
A conversion team testing landing pages, or a small startup without a Harness contract.
Visit Harness Feature Management & Experimentation →
#10 Datadog Experiments
Warehouse-native experiments inside Datadog · United States · datadoghq.com
Ex-Eppo Warehouse-native Observability link
Eppo's own site now reads Datadog Experiments and says Eppo has been acquired by Datadog. The product page promises sample ratio mismatch detection, variance reduction, sequential testing and the ability to watch latency, errors and crashes next to experiment results. Metrics are computed in your own data warehouse with the methodology visible.
- Pricing: Quote only. No price was read on the product page; Datadog publishes a separate pricing page that did not render as plain text. Free plan: free trial.
- Best for: Datadog customers with data teams
- Strengths: Experiments; Sample ratio mismatch detection; Variance reduction.
Where it falls short
There is no marketing editor, so page tests need engineering. Pricing was not visible to us.
Wrong for
A marketing team that needs to change page elements without developers, or a company with no Datadog footprint.
Visit Datadog Experiments →
#11 LaunchDarkly
Feature flag platform with experiments included · United States · launchdarkly.com
Feature flags Usage meters Developer-first
LaunchDarkly's free Developer plan includes unlimited seats and unlimited flags with 100,000 experimentation monthly active users, which lets a team test features the moment flags exist. The paid Foundation plan has no platform or seat fees but bills per server-side service connection and per client-side MAU, and it adds session replay and AI run tracking.
- Pricing: Usage-based. Developer plan free with 100K experimentation MAU a month. Free plan: free plan.
- Best for: Engineering orgs with flags in place
- Strengths: A/B tests and experiments; Unlimited feature flags; 30 SDKs.
Where it falls short
Meters for connections, client-side MAU and AI runs combine into a bill that is hard to forecast. There is no page editor for marketers, and statistics are less deep than at GrowthBook or Statsig.
Wrong for
A marketing group running landing page tests, or a consumer product with millions of monthly users on a small budget.
Visit LaunchDarkly →
#12 Optimizely
Experimentation inside a wider digital experience platform · United States · optimizely.com
Enterprise suite Stats Engine Warehouse-native
Optimizely's page describes agentic experimentation covering A/B, multi-page, feature and server-side tests, a proprietary Stats Engine, audience targeting and warehouse-native analytics. It also sells a CMS, a content marketing platform and an agent platform, so experimentation is one piece of a packaged digital suite. It is listed as a major vendor and cannot hold a top place.
- Pricing: Quote only. Every plan is individually packaged and quoted; no prices are published. Free plan: none.
- Best for: Enterprises buying a content and testing suite
- Strengths: A/B and multi-page tests; Feature and server-side testing; Stats Engine.
Where it falls short
No plan or price is published, and the packaging is built around the wider Optimizely suite. Results use a proprietary statistics engine that is harder to audit than open methods.
Wrong for
A small or mid-sized team that wants to start on a card and a free plan.
Visit Optimizely →
#13 Adobe Target
Testing and personalization inside Adobe Experience Cloud · United States · business.adobe.com
Adobe stack Personalization MCP server
Adobe's documentation describes Target as the Experience Cloud solution for tailoring and personalizing digital experiences on web, mobile, apps and social, with audiences, implementation and administration topics, and an MCP server in public beta that exposes experimentation and reporting operations. It is a major vendor and is placed accordingly.
- Pricing: Quote only. Pricing on request; no price was visible on the pages read. Free plan: none.
- Best for: Adobe-centered enterprises
- Strengths: A/B testing; Audiences; Personalization.
Where it falls short
Pricing is quote-only and tied to the Adobe contract. Setup details were not visible to us beyond the documentation.
Wrong for
A team without an Adobe contract, or one that wants to be testing within days.
Visit Adobe Target →
Questions and answers
What is the best A/B testing tool in 2026?
GrowthBook ranks first of 13, best for engineering-led product teams. Convert is second (marketing teams leaving VWO) and Kameleoon third (mixed marketing and product teams). Places 1 to 3 go to challengers, vendors outside the best-known brands, wherever the category has them.
How much does an A/B testing tool cost?
The spread is wide. GrowthBook and PostHog have free plans, Statsig's Pro plan is $150 a month, GrowthBook Pro is $40 per seat, and Convert's Growth plan starts at $299 a month billed annually.
Does an A/B testing script slow down my website?
Client-side tools load JavaScript before the page renders, which can cause flicker or a measurable delay if the snippet loads late. Vendors mitigate this with an anti-flicker setting or by loading the script synchronously in the head.
What is the difference between client-side and server-side testing?
Client-side testing changes the page in the visitor's browser through a script, which a marketer can do in an editor. Server-side testing decides the variant in your application code and sends the finished page or API response.
How many visitors do I need to run a test?
It depends on the baseline conversion rate and the smallest lift you care about. A page converting at 3% needs roughly tens of thousands of visitors per variant to detect a 10% relative lift with usual settings.
Is there a free alternative to Google Optimize?
Google ended Optimize in 2023 and offered no direct replacement.
Should I choose Bayesian or frequentist statistics?
Both work if used as designed. Frequentist tests give a fixed sample size and are strict about stopping early. Bayesian results give a probability that a variant is better, which many non-statisticians read more easily but which can encourage peeking.
Can I run A/B tests without a developer?
Yes, for page-level changes.
What happens to my tests if my vendor is acquired?
It varies. Eppo became Datadog Experiments, VWO and AB Tasty announced a merger in January 2026, and Statsig announced it is joining OpenAI.
Can I move to another A/B testing tool later?
Running tests can be recreated, but history is harder. Export assignment logs, raw events and the final reports of every past experiment, because a new tool cannot rebuild old results from nothing.
Do A/B testing tools comply with GDPR?
Compliance depends on configuration. These tools set cookies or local storage to keep a visitor in the same variant, which usually needs consent in the EU.
