Most companies still decide what goes on their homepage the same way they did in 1995: an argument in a meeting, where the loudest voice or the highest-paid opinion wins. AB Tasty spent sixteen years selling the antidote. Ship both versions of the button, the headline, the checkout flow. Let real visitors vote with their clicks. Keep the winner. Repeat forever.
That is the whole idea behind the Paris company, and it is deceptively boring - which is exactly why it works. Every brand with a website will always want to know whether this version converts better than that one. AB Tasty turned that permanent, unglamorous question into a software platform now used by more than 1,100 brands, from Disney and L'Oreal to McDonald's, Sephora and USA Today. In January 2026 it did the most on-brand thing imaginable for a testing company: it ran the ultimate experiment on itself and merged with its biggest rival.
What it actually doesGuessing, replaced by evidence
At its core, AB Tasty is a digital experience optimization platform. That phrase hides four practical jobs. First, experimentation: A/B and multivariate tests on a website or app, built in a visual editor so a marketer can change a page without filing an engineering ticket. Second, feature management: developers wrap new features in flags and release them to 1% of users, then 5%, then everyone - and roll back instantly if something breaks. Third, personalization: showing different visitors different experiences based on who they are and how they behave. Fourth, recommendations and search: the AI that decides which products to surface on an e-commerce page.
Bundling those four into one platform is the strategic bet. A retailer can test a new product page, roll it out safely, personalize it by audience, and tune the recommendations underneath it - without stitching together four separate vendors.
OriginsIt started as something else entirely
AB Tasty did not begin as AB Tasty. In 2009 two childhood schoolmates, Alix de Sagazan and Remi Aubert, started a web analytics consultancy in Paris called Liwio. They spent their days telling clients what was wrong with their websites. The pivot came from an obvious frustration: it is one thing to diagnose a problem, another to let clients fix and measure it themselves. Around 2013 they relaunched as a self-serve A/B testing product. The consultancy became a platform.
The founders kept an unusual arrangement for over a decade - running the company as co-CEOs, a setup that fails more often than it works. Their internal culture leaned into a single line that doubles as the company's entire worldview.
We never lose. We either win or learn. AB Tasty internal mantra
It sounds like a poster. But it is also a hiring filter and a product philosophy: if every launch is an experiment, a failed test is not a failure, it is data. Employees are called “Magic Makers,” there is a “Culture Crew,” and the company runs a podcast named, with zero subtlety, the 1000 Experiments Club.
The weird productA website that reads the room
The most unusual thing AB Tasty ships is EmotionsAI, launched in 2023 out of its acquisition of a startup called Dotaki. The premise is genuinely strange: by watching how you scroll, how long you hesitate, and how your mouse moves, it tries to infer your emotional state - are you anxious, decisive, browsing, comparing - and then serves a version of the page built for that mood. Whether you find that clever or unsettling, it is a real product doing a thing most of the market does not attempt.
What you can copy
The mechanic, not the software. The reusable idea here is cheap and universal: for any decision where you have two defensible options, ship both to a slice of real users and let behavior break the tie. Most teams could run their first meaningful A/B test this week - on an email subject line, a pricing page, a sign-up button - without buying anything.
The condition it needs. It only works with enough traffic to reach statistical significance. On a page with 40 visitors a month, a test tells you nothing. This is why AB Tasty sells to brands, not hobby blogs.
The growth engineBuy the roadmap, don't build it
Between 2022 and 2025, AB Tasty made four acquisitions in four years, and the pattern is instructive. It bought Epoq (Germany, 2022) for AI recommendations and site search; Dotaki (2023) for the emotions modeling behind EmotionsAI; Potions (2024) for e-commerce personalization; and Wandz.ai (2025) for real-time adaptive experiences. Rather than build every capability in-house, it bought the missing pieces and folded them into one suite.
The other 2025 move was defensive. In November the company launched Evi, an agentic AI assistant that runs an A/B test from hypothesis to build to analysis. Early adopters reported roughly 73% less setup time and about 25% more campaigns. The telling detail: AB Tasty bundled Evi into every contract at no extra cost. When a wave of AI-native startups is coming for your category, giving away the AI layer is how you keep customers from leaving to find it elsewhere.
The businessWho pays, and how much it takes
The model is straightforward B2B SaaS: annual subscriptions priced by traffic, by which modules a customer turns on, and by support tier. Growth comes from cross-selling - a brand that starts with web testing gets sold feature management, then personalization, then recommendations. Standalone annual recurring revenue was estimated around $40M before the merger, on a team of roughly 340 to 400 people across eight countries.
Funding was comparatively modest for the reach. After a Series A in 2014 and a Series B, the notable round was a $40M Series C in 2020 led by Credit Mutuel Innovation, with Korelya Capital, Omnes, Partech and XAnge. That is a lean cap table for a company selling to Disney and LVMH - a sign it grew more on revenue than on venture fuel.
This combination gives us the opportunity to pursue the same ambition we've always had, but at a much greater scale and significantly expanded resources. Alix de Sagazan, on the 2026 merger
The plot twistWhen rivals merge instead of fight
On January 20, 2026, AB Tasty combined with VWO (Wingify), an Indian experimentation company, under the Singapore private equity firm Everstone Capital. The two had spent years as direct competitors. The merged entity reported over $100M in combined annual recurring revenue, 4,000+ customers, roughly 800 employees, and 11 offices. Analysts pegged the deal's value in the $400M to $500M range, at four to five times combined revenue.
In the new structure, VWO co-founder Sparsh Gupta became CEO. Alix de Sagazan took the Chief Revenue Officer role, and Remi Aubert became Chief Customer & Strategy Officer - founders who traded the crown for reach. When two of the larger independents in a category stop competing and combine, it usually says the market is consolidating and scale now matters more than another year of head-to-head feature wars.
The fieldWhere it sits
AB Tasty's rivals are a crowded set: Optimizely, Adobe Target, Kameleoon, Dynamic Yield, LaunchDarkly on the feature-flag side, and Convert, plus newer entrants like Datadog's Eppo and Amplitude Experiment. The rough map below reflects general market positioning, not audited share.
Its edge is less about any single feature and more about the bundle: experimentation, feature management, personalization, and merchandising under one roof, sold to enterprise brands that would rather manage one contract than five. The 2026 merger doubled down on that positioning - breadth and scale over point-solution depth.
Where the pitch breaks down
Experimentation platforms are not magic. They need traffic volume, a real testing culture, and the discipline to act on results rather than relaunch the same debate with a chart attached. A team that runs one test and ignores the outcome gets nothing. And the tools themselves are converging - which is precisely why the market is merging rather than multiplying.
Still, the underlying business rests on a problem nobody outgrows. As long as brands have websites, they will want to know whether this works better than that. AB Tasty built a company on refusing to guess - and then, fittingly, tested that conviction at the largest scale it could find.