The best A/B testing tools in 2027

Looking for an A/B testing tool to find out, with numbers to back it up, which version of a page converts best? This ranking compares ten tools on what matters at the moment of purchase: the types of tests offered, the robustness of the statistics, the way the tool handles your data and what you will pay. The profiles reflect what each vendor shows on its site in October 2026; after them come a comparison table and five points to check before you sign.

The 10 best A/B testing software in 2027

1. Wingify (AB Tasty and VWO): to test, personalize and roll out features in a single platform, on a French-language site.
2. Adobe Target: for companies that want to test and personalize web and mobile experiences in connection with Adobe Analytics.
3. Optimizely: for marketing and digital teams that want testing inside a platform that also manages content and AI agents.
4. Convert: for a tool focused on testing, with public prices and a 15-day free trial.
5. Statsig: for product teams that want tests, feature flags and product analytics, with a free plan.
6. LaunchDarkly: for developers who test every code change and every AI agent in production.
7. GrowthBook: for data teams that want an open source tool plugged into their data warehouse.
8. Webyn: to hand analysis, variant creation and personalization over to AI agents, on a French-language site.
9. Varify.io: for a fixed price published in euros and results measured in Google Analytics 4.
10. Monetate: for large brands, including in regulated industries, that want testing and personalization together.

Why does the choice of an A/B testing tool matter in 2027?

Google Optimize has shut down. According to the Google Analytics help center, Google Optimize and Optimize 360 have been unavailable since September 30, 2023, and any tests still running on that date were ended. Google says it is investing in third-party A/B testing integrations for Google Analytics 4 and names three partner providers: AB Tasty, Optimizely and VWO. To test with Google Analytics 4, Google therefore points you to third-party tools.

A badly set up test can hurt your search rankings. In its guidelines on A/B tests, Google asks you not to show different URLs to its crawler and to users, a practice it classifies among the cloaking techniques that break its spam policies. It recommends the rel="canonical" attribute on the variants of a page and a temporary redirect (302) to send a visitor to a variant. A test that runs longer than necessary can, in its view, look like an attempt to deceive search engines.

Test cookies raise the question of consent. The CNIL, France's data protection authority, exempts audience measurement trackers from consent under precise conditions: a purpose limited to measuring the audience, solely on behalf of the site publisher, and anonymous statistics only. That page does not mention A/B testing. Some offers fall outside this exemption, in particular when the provider reuses the data for its own purposes. Before choosing a tool, then, ask the vendor what it does with your visitors' data.

How did we rank these tools?

To score a testing tool, we used only what its vendor publishes: in October 2026, we read, tool by tool, the homepage, a features page and the pricing grid. No paid testing, no partnerships.

1. Types of tests. Classic A/B test, split URL (a variant hosted at another address), multivariate test, multi-armed bandit test (the tool sends more traffic to the winning variant), server-side tests and feature tests.

2. Robustness of results. Statistical methods offered, sequential testing, sample ratio mismatch detection (when the actual traffic split drifts from the planned one), guardrail metrics and automatic rollback.

3. Data handling. Where results are measured (in the tool, in your analytics tool or in your data warehouse), where data is stored, and what the vendor says about cookies and the GDPR.

4. An offer that is easy to read from France. Site language, displayed price, free trial.

Customer counts, claimed conversion lifts and review ratings were left out.

The 10 best A/B testing software: the profiles

1. Wingify (AB Tasty and VWO)

Wingify brings AB Tasty and VWO together under one name. Its homepage announces that the two vendors "have joined forces" and links to a page about the merger; the former abtasty.com and vwo.com sites redirect to wingify.com. The platform combines experimentation, personalization, behavior analytics, feature management, commerce and customer engagement.

Its AI, Wingz, suggests hypotheses, sets up tests and delivers the result. For the web, the pricing page lists A/B/n, split and multivariate tests, a visual editor and a code editor, and unlimited experiments, variations and metrics from the entry plan. For product teams, the feature management module puts a flag (feature flag, a switch that turns a feature on for a share of users) on each new release: exposure starts at 1%, rises gradually to 100%, and the feature can be switched off instantly if something goes wrong.

Best for: product, engineering and e-commerce teams that want to decide and act in one place, from testing to personalization and feature rollout, on a French-language site.

Strengths: data centers in the United States, the European Union and India, so you choose where your data is stored; sample ratio mismatch detection and guardrail metrics from the Pro plan; an MCP server (a standard that lets an AI assistant operate the tool) to manage feature flags from your coding agent.

Limitations: no fixed-price plan, the pricing page announces a custom proposal. Multivariate and cross-domain tests start with the Pro plan; self-hosting, single sign-on (SSO) and API access are reserved for the Enterprise plan. Wingz Advanced, which adds autopilot, AI workflows and synthetic A/B testing, is a paid option.

Pricing: on quote, with a proposal built around your goals; free trial offered.

2. Adobe Target

Adobe Target optimizes and personalizes the main digital interactions of a website or an app, with a stated goal: improving conversions, retention and revenue. Its experimentation page covers four families: A/B/n tests, multivariate tests, multi-armed bandit tests and optimization for mobile devices. Tests run client-side, server-side or both at once.

An estimation tool calculates the traffic a multivariate test needs to be conclusive. Experimentation Accelerator, an app of Adobe Journey Optimizer, analyzes past and ongoing tests in Target and makes AI-generated test recommendations.

Best for: companies that want to test and personalize their websites and mobile apps, and connect these tests to Adobe Analytics.

Strengths: product pages and pricing page in French; a two-way server-side integration between Adobe Target and Adobe Analytics to deepen the analysis of tests and personalization; full factorial multivariate tests, which examine every possible combination.

Limitations: no published price, Adobe talks about custom configuration and licensing options, and the volume of digital properties equipped is part of the calculation. Several highlighted features rely on other Adobe products: Adobe Analytics for in-depth analysis, Adobe Journey Optimizer for Experimentation Accelerator.

Pricing: on quote (custom pricing).

3. Optimizely

Optimizely presents itself as an AI platform for marketing, which brings together content management, experimentation, AI agents, personalization, analytics, feature management and commerce. Its homepage tells visitors who only know its testing tool that the platform provides the tools to create and optimize every digital experience.

The Agentic Experimentation module covers A/B, multi-page, feature and server-side tests, with an in-house statistics engine (Stats Engine), real-time personalization, audience targeting and warehouse-connected analytics to measure impact.

Best for: marketing and digital teams that want to create content, test, personalize and prove business impact in the same platform.

Strengths: a continuous workflow, from brief to live experience, without switching tools; analytics plugged into your data warehouse; an Agent Experience component to track and optimize how large language models use your content.

Limitations: site in English only. The pricing page presents experimentation as one product among ten (content management, agents, analytics, data, commerce...), and each offer is put together individually with the vendor.

Pricing: on quote.

4. Convert

Convert sells a tool dedicated to testing, Convert Experiences, which combines A/B tests, split URL, price tests, server-side experiments, one-to-one personalization and AI flows, complemented by multivariate and multi-page tests. For server-side tests, six SDKs (libraries to integrate into your code: PHP, Python, Ruby, iOS, Android and JavaScript) let you test business logic, APIs, prices and feature flags.

The analysis module is included in every plan, at no extra charge: heatmaps of clicks, taps and scrolling, and session recordings filtered on signs of frustration (Convert Signals). On the statistics side, the tool offers frequentist and Bayesian models and sequential testing. The Convert Intelligence assistant and an MCP server let you create and analyze tests with AI.

Best for: privacy-minded companies that want a complete testing tool, with prices you can look up on the site.

Strengths: 15-day trial, with no credit card required; the option to delay assigning a visitor to a variant until their explicit consent is obtained; integrations, API access and an MCP server on every plan.

Limitations: the site, pricing page included, exists only in English. The price follows the number of monthly tested users (MTU), with billed overage that can be turned off. The entry plan is limited to 5 active projects and 5 active experiments per project, and multivariate and multi-page tests start with the Pro plan. No plan includes development help.

Pricing: excl. VAT, prices published in US dollars on the site.

5. Statsig

Statsig gathers more than five products in one platform, including experimentation, feature flags, product analytics, session replays and web analytics. Its homepage states that Statsig is part of the Amplitude family.

The statistics engine handles A/B/n and multivariate tests, Bayesian and frequentist approaches, and advanced techniques such as CUPED, which shortens a test and gains precision thanks to historical data, or sequential testing, which keeps you from reading results too early. Two deployments exist: Statsig Cloud, where the vendor manages the infrastructure, and Warehouse Native, which runs the statistics engine in your own data warehouse and is only offered with the Enterprise plan.

Best for: product and engineering teams that want tests, flags and product analytics in one tool, from the solo developer to the large organization.

Strengths: a free Developer plan with no credit card, with 2 million events per month, 50,000 session replays and unlimited seats; a visual editor to create experiments without code; more than 30 SDKs.

Limitations: site in English only. Beyond the 5 million events included in the Pro plan, each thousand events is billed. Deployment in your data warehouse, single sign-on and priority support are reserved for the Enterprise plan, on quote.

Pricing: prices published in dollars on the site; free Developer plan.

6. LaunchDarkly

LaunchDarkly describes itself as a runtime control layer: the tool manages code and AI agents in production with feature flags, progressive rollouts and automatic rollbacks. Experimentation sits alongside observability and AgentControl, which governs the behavior of AI agents.

For testing, the pricing page lists A/B/n tests on flags, Bayesian and frequentist analysis, sequential testing, multi-armed bandits, holdouts (a group kept aside to measure the cumulative effect) and integration with data warehouses such as Snowflake or BigQuery. A flag switches everywhere in under 200 milliseconds, with no redeploy, from MCP, Git, the command line, the API or the interface.

Best for: development teams that ship code and AI agents to production and want to test every change.

Strengths: a free-forever Developer plan, with no credit card, with 100,000 monthly active experimentation users included; no per-seat pricing, the whole team can be invited and the price follows usage; more than 25 SDKs for front end, back end and mobile.

Limitations: site in English only, written for developers, and the site does not mention a visual editor to change a page without code. Release monitoring, guardrail metrics and automatic rollback belong to Guardian, a paid option of the Enterprise plan.

Pricing: prices published in dollars on the site; free Developer plan, then pay-as-you-go with the Foundation plan, after a 14-day trial.

7. GrowthBook

GrowthBook is an open source platform for experimentation, feature flags and product analytics, plugged into your data warehouse: metrics are defined in SQL on your own data. The tool comes in a cloud version and a self-hosted version, with the same features, on your cloud or on your servers.

Any flag becomes an A/B test in one click. Rollouts are automated with ramp-up schedules, alerts on guardrail metrics and approval steps, and a rollback triggers automatically when a guardrail metric degrades. From Claude Code, Cursor or VS Code, a developer can put a block of code behind a flag and create an experiment.

Best for: product, data and engineering teams that want experimentation and flags plugged into their data warehouse, or hosted in-house.

Strengths: a free Starter plan for up to 3 users, with unlimited flags, experiments and traffic; a free self-hosted open source version; more than 24 lightweight SDKs for client, server, mobile and servers close to visitors (edge).

Limitations: site in English only. The Pro plan is paid per seat and stops at 30 users and 3 projects. Ramp-up schedules, exportable audit logs, single sign-on (SSO) and automatic account provisioning (SCIM) belong to the Enterprise plan, on quote.

Pricing: prices published in dollars on the site; Starter plan and open source version free.

8. Webyn

Webyn entrusts conversion optimization to three AI agents that work on your data in real time. The Insight Agent analyzes behaviors, journeys and friction points to spot the underperforming pages, segments and journeys. The Coding Agent generates and deploys the variants to test, without code, and the Personalization Agent adjusts content, messages and recommendations according to each visitor's behavior, source and intent.

From the homepage, a free scanner analyzes your site's URL and returns 4 AI recommendations.

Best for: teams that want to launch tests and personalization without tying up their product or development teams.

Strengths: site in French; a dedicated CRO Angel (CRO: conversion rate optimization) from the first plan; dynamic traffic allocation and QA across several devices and browsers included in that same plan.

Limitations: no published price, the three plans (Boost, Scale, Elevate) are on quote, chosen according to your traffic and what is at stake for you. Personalization starts with the Scale plan, and heatmaps only appear with the Elevate plan.

Pricing: on quote.

9. Varify.io

Varify.io evaluates your tests with your own analytics data: results rely on your Web Analytics data, imported into its reports, with a Google Analytics 4 integration from the first plan. The vendor presents its platform as developed in Germany.

The tool offers A/B tests, split URL, multi-page tests, personalization and full page redesigns, with a visual editor and a code editor. Its AI assistant turns an idea described in plain language into an experiment ready to launch, and an AI CRO audit suggests prioritized test ideas. The tool works with MS Clarity, Hotjar and Contentsquare.

Best for: small teams new to A/B testing and agencies that want a fixed price, with no traffic-based billing when tests are measured in Google Analytics 4.

Strengths: a French-language site and a price grid in euros; unlimited traffic at the same price; a platform advertised as cookieless and GDPR-compliant.

Limitations: the first plan is limited to one website and 5 active experiments. AI features start with the Pro plan and are paid in AI credits, or with your own key. Unlimited traffic applies to tests measured with Google Analytics 4 or BigQuery: with another analytics tool, tests go through Varify Tracking, configured according to your number of tested users.

Pricing: 15-day free trial, then the Growth plan at €199 per month billed annually (€2,388 per year) or €249 monthly, the Pro plan at €249 per month annually (€2,988 per year) or €349 monthly, the Enterprise plan from €799 per month; VAT treatment not stated, price checked on October 10, 2026.

10. Monetate

Monetate defines itself as an experience optimization platform, which combines AI, personalization, recommendations and experimentation in a single solution. It is organized in two suites: Symphony for personalization, Maestro for experimentation. The former sitespect.com site redirects to the Maestro page, and the Monetate homepage announces the acquisition of Simon AI.

Maestro covers A/B, A/B/n and multivariate tests, dynamic tests and feature experimentation, client-side or through the API, with an architecture designed to avoid flicker on the tested page. Forte, its secure component, targets regulated industries such as finance and healthcare, logged-in areas and checkout funnels, with possible deployment in the cloud, on a private network or on premises.

Best for: large brands and digital businesses, including in regulated industries, that want personalization, recommendations and testing together.

Strengths: personalization, recommendations and testing in a single solution; Maestro advertised as compliant with the GDPR, the CCPA (California's privacy law) and the PCI standard (card payment security), and ready for HIPAA (US rules on health data); a Concierge managed service, where the vendor's team handles design, development and execution.

Limitations: site in English only and no published price, each plan is custom-built. The site mentions neither a free trial nor a self-service offer.

Pricing: on quote.

The 10 A/B testing tools compared

Prices and features checked on October 10, 2026 on each vendor's site. Varify.io is the only one to show euros: in place of an entry price, the last column lists what pushes the bill up.

Software	For whom	Tests offered	French version	What drives the bill up
Wingify	Product, engineering and e-commerce teams	A/B/n, split, multivariate, bandit, feature flags	Yes (pricing page in English)	Custom proposal, added products, options (Wingz Advanced, add-on modules)
Adobe Target	Companies, on web and mobile	A/B/n, multivariate, multi-armed bandit, server-side	Yes	Options chosen, volume of digital properties, delivery channels
Optimizely	Marketing and digital teams	A/B, multi-page, features, server-side	No	Products chosen among ten, custom-built offer
Convert	Privacy-minded companies	A/B, split URL, multivariate, pricing, server-side	No	Tested users per month, overage, plan (active projects and experiments)
Statsig	Product and engineering teams	A/B/n, multivariate, bandit, holdouts, sequential	No	Events beyond the included quota, warehouse deployment (Enterprise)
LaunchDarkly	Development teams	A/B/n on flags and configurations, bandit, holdouts	No	Service connections, client-side active users, AI executions, Guardian option
GrowthBook	Product, data and engineering teams	A/B, split URL, bandit, holdouts, flags	No	Number of seats, managed warehouse events, requests to the delivery network (CDN)
Webyn	Teams that test without tying up developers	AI-generated variants, personalization	Yes	Plan (Boost, Scale, Elevate) according to your traffic and what is at stake for you
Varify.io	Small teams and agencies	A/B, split URL, multi-page, personalization	Yes	Plan (active experiments, domains), AI credits; unlimited traffic with GA4
Monetate	Large brands and digital businesses	A/B/n, multivariate, dynamic, features	No	Features chosen, custom-built plan

Other tools worth a look

Software	Its specialty, according to the vendor
Contentsquare (Hotjar)	Experience analytics with heatmaps and session recordings; the former hotjar.com site redirects to a Hotjar page on Contentsquare.
Crazy Egg	Heatmaps, recordings, surveys and A/B tests to see what a site's visitors do.
Mouseflow	Behavior analytics with session recordings, heatmaps and visitor feedback.
Lucky Orange	Heatmaps, conversion funnels and session recordings, tied together by an AI that tells you what to fix first.
Personyze	No-code web personalization and A/B tests for marketers, with a French-language site.
Mida	Lightweight A/B testing platform for websites, with no developer needed, and a free plan.
Omniconvert	A/B tests and conversion optimization (Explore), customer reviews and NPS (Pulse), and Nexus, AI agents for Shopify stores.
Datadog Experiments	A/B tests plugged into the data warehouse; the geteppo.com homepage announces the acquisition of Eppo by Datadog.
ABsmartly	A/B testing platform built by the team behind Booking.com's experimentation engine, for product teams and data teams.
CustomFit.ai	Personalization, A/B tests and conversion optimization for Shopify brands and direct-to-consumer (D2C) brands, with no engineers.
Intelligems	Tests on prices, content, checkout and shipping for Shopify stores.

If you would rather hand your tests to a team, see the best CRO agencies in France, which specialize in conversion rate optimization. To make use of the data your tests produce, also see the best data and analytics agencies.

What should you check before choosing?

1. What unit is the price calculated on? The units vary: tested users per month at Convert, events at Statsig, seats at GrowthBook, service connections and active users at LaunchDarkly, a fixed price at Varify.io when your tests are measured in Google Analytics 4. Estimate your volume over twelve months.

2. Where are results calculated? Some tools measure in their own interface, others read your data: Google Analytics 4 for Varify.io, your data warehouse for GrowthBook or for the Warehouse Native offer from Statsig. Choose the tool that calculates from the number your teams already trust.

3. Does the test damage display or search rankings? Ask how the tool avoids flicker, that moment when the visitor glimpses the old version before the variant. Also check that it lets you apply Google's guidelines: rel="canonical" on variants, a temporary redirect (302) for split URL tests, and quick removal of test elements once the test is over.

4. Which statistical safeguards are included in your plan? Sequential testing, sample ratio mismatch detection, guardrail metrics: at Wingify or Convert, some of these features are reserved for higher plans. Check the exact plan that contains them.

5. What happens to your visitors' data? Ask where the data is hosted, whether the vendor reuses it for its own purposes, whether the tool can wait for consent before assigning a visitor to a variant, and whether it works without cookies. For your specific case, have the configuration validated by the person in charge of compliance.

Where does this ranking stop?

No test ran on a real page. Our profiles describe each tool based on its vendor's site, consulted in October 2026, without measuring its effect on a conversion. The Kameleoon site could not be read at that date: the tool is therefore neither ranked nor described here, which says nothing about its quality.

Your traffic matters more than the rank. A store testing its product pages and a team shipping features every week do not need the same tool. A reliable test also needs enough visitors: Google points out that its duration depends on your conversion rate and your site's traffic. Rely on the Best for lines more than on the profile number.

It follows a moving market. Among the tools mentioned here, AB Tasty and VWO have joined forces under the Wingify name, Datadog bought Eppo and Monetate bought Simon AI. We will review this ranking regularly; until then, check prices and features with the vendor before signing anything.

How do you attract enough visitors to test once the tool is chosen?

An A/B testing tool improves what visitors already on your site do. It does not bring new visitors from Google, and it does not get you cited by ChatGPT when a buyer asks which solution to choose. Yet, according to Google, the time a reliable test needs depends in particular on your site's traffic.

Sorank takes care of that part. After an audit of your site, the tool picks out your keywords, writes and then publishes articles designed for Google as well as for AIs, and follows, query by query, the AI answers that mention you. How it works is laid out on our GEO software page. Want to talk about it for your own site? Book a call.

Conclusion

Three profiles, three picks: Wingify for a product or e-commerce team that wants to test, personalize and roll out features on a single platform, with a French-language site; GrowthBook for a product and data team that wants an open source tool, plugged into its data warehouse, with a free plan to start; Varify.io for a small team or an agency that wants a fixed price published in euros and results read in Google Analytics 4.

Before signing, run a first test on a page that gets enough traffic and compare the tool's numbers with those of your analytics tool.