Brand Statsig Moves from OpenAI to Amplitude: All Hail Experimentation Capability!

This is team Convert’s take on what the shift means.

Disclaimer: Convert Experiences is an A/B testing and 1:1 personalization platform that’s been in the experimentation space for 16 years.

5th May 2026, the Official Release Drops

Amplitude’s Spencer Skates broke the internet with the declaration. Amplitude is entering into a strategic partnership with Statsig. The platform we’ve come to know as OpenAI’s acquisition.

It resulted in speculation. Yes! But the announcement is both expected and unique.

Expected: Pundits have been opining that OpenAI was looking to build its experimentation muscle with the Statsig deal. It was never about the AI behemoth entering the SaaS space.

Unique: OpenAI could have let the platform decay. Without fanfare. Through neglect. We’re choosing to view this handover as accountability on their part where the customers who depend on Statsig to grow their business will now have a home (and a partner) devoted to delivering experimentation capabilities via the SaaS model.

Throwback: Why OpenAI Bought Statsig

On September 2, 2025, OpenAI announced it had acquired Statsig in an all stock deal valued at $1.1 billion, one of its largest acquisitions to date.

Statsig founder and CEO Vijaye Raji was named CTO of Applications under former Instacart CEO Fidji Simo, with responsibility for product engineering across ChatGPT and Codex. OpenAI’s public position was that Statsig would “continue operating independently and serving its customer base out of its Seattle office.”

That was the status report for eight months.

Sequoia, in a widely circulated note on the deal, said: AI generates infinite variations, but knowing which one works is the harder question.

The same analysis cast Statsig as the runtime control plane OpenAI needed to close the loop between AI generated code and what actually worked in production.

Statsig’s own announcement aligned somewhat with this read. The single largest trend in software since the company’s founding had been the rise of AI, and joining OpenAI was the most direct way for the team to participate in it.

Now: The Statsig – Amplitude Deal

On May 5, 2026, Amplitude announced it would take on Statsig’s brand and customers, maintain the existing platform across cloud and warehouse deployments, and develop an integrated roadmap with the team that remained at OpenAI.

The September promise of operational independence for Statsig had been retired.

For Statsig customers, the practical changes are straightforward: ownership shifts to Amplitude, the long term roadmap converges with Amplitude’s, and the next renewal cycle determines pricing and contract terms.

Amplitude has committed to maintaining the existing platform and supporting current customers through the transition, working alongside the OpenAI employed Statsig team during the handover period.

Several questions though stay unanswered in the public announcement.

How a warehouse native architecture gets reconciled with Amplitude’s event stream product analytics roots. How deeply the two roadmaps actually merge once integration moves from press release to engineering work. How pricing transitions look at renewal.

Statsig customers should treat that next contract negotiation as the real test of what the handover means in practice.

What Practitioners are Saying

Reactions inside the experimentation community split along three lines.

Ben Labay put the structural read: “experimentation isn’t a category to acquire. It’s a muscle to absorb.”

Earlier consolidations Eppo into Datadog, Split into Harness, Optimizely into Episerver were category rollups in the conventional sense: bring the platform inside a larger stack, sell more seats. The Statsig deal works differently. The team and the customer base went to different buyers because OpenAI wanted the capability for itself and someone else needed to inherit the SaaS business that came attached.

Simon Jackson read the same event as a competitive signal. AI has cheapened code generation enough that the bottleneck has moved from shipping to learning. The teams that win in an AI native era, in his words, will be “the ones learning the fastest” rather than the ones with the most engineers shipping the most features.

Dennis van der Heijden, our co founder at Convert, took a grounded view.

The handover is more thoughtful than the headlines suggest. OpenAI could have absorbed the team and let the platform wither a familiar fate for SaaS businesses bought by larger companies that don’t really want to run them.

Instead, the brand, customers, and operational continuity went to a buyer for whom serving thousands of experimentation customers is the actual business.

The three takes don’t contradict each other. The deal is unusual in shape, meaningful for what it implies about where competitive advantage lives in an AI native market, and, for Statsig customers specifically, handled with more care than it might have been.

What’s Next for Experimentation? 

Zoom out, and the Statsig handover sits in the middle of a much larger restructuring of the experimentation category.

In the past 18 months alone:

• Datadog acquired Eppo for a reported $220 million (May 2025)
• Harness acquired Split (May 2024, undisclosed but Split had raised >$100M)
• Webflow acquired Intellimize (2024, eight figure range)
• Braze acquired OfferFit for $325 million (2025)
• Monetate acquired SiteSpect (2025, via a $75M loan)
• Everstone Capital took a majority stake in Wingify/VWO (~$200M, January 2025)
• Everstone then merged VWO with AB Tasty in January 2026, creating a $100M+ ARR platform with 4,000+ customers

Two parallel forces are doing most of the work.

Ryan Lucht, who lived through the Eppo–Datadog deal from inside, described both: a private equity roll up play targeting cashflow positive web testing tools, and an absorption play in which larger categories analytics, observability, customer engagement, AI applications pull experimentation in as connective tissue.

The web only experimentation TAM, by Convert’s reading of the market, sits at roughly $1B with growth slowing to ~10%.

A hard number to defend a high multiple exit on. Three viable paths emerge from there: go upmarket and consolidate (the Everstone VWO AB Tasty path), embed into a bigger category (the Datadog Eppo, Harness Split path), or sell the team for the capability the buyer wants internally (the OpenAI–Statsig path).

The Statsig deal is the third path.

It also splits something the first two paths kept together: the team went to one acquirer, the customers to another. The platform survives, but inside an AI analytics suite rather than as a standalone business.

There is no precedent for that shape in the experimentation category.

But one has been set, and more may follow suit.

Where Convert is Headed!

For years, the experimentation industry has felt close to reaching its total addressable market (TAM). Experimentation teams with enough data science knowledge and traffic to handle the correct methods and process, all divided over the same shortlist of vendors, same shortlist of buyers.

Statsig’s revenue concentration was widely discussed inside the industry a small number of large customers driving the bulk of the line.

The category was full. The market wasn’t growing.

We believe AI changes that math. If hypothesis generation, variant development, QA, and analysis can each be simplified and so compressed in time by an order of magnitude and each is being compressed right now, including inside Convert’s own roadmap then the cost of running a real experiment falls through the floor.

What used to take a sprint takes a morning. Time to launch becomes the actual unlock, ahead of statistical sophistication.

Ninety five percent confidence has been gospel for two decades. It made sense when each test cost weeks of engineering and you only had budget for one bet a quarter.

But if a small team can ship and validate ten experiments in the time it used to take to ship one, running at 80% confidence with smaller, safer rollouts stops being reckless and starts a practice of direction over certainty.

More learning per euro spent.

The natural counter to this line of thinking that AI’s infinite variant supply demands more statistical rigor, not less is defensible, but it tends to apply to programs whose binding constraint is the luxury of pure signal quality.

Most teams below enterprise scale have the opposite problem.

The bold bet underneath all of this: there is a much larger market of teams who gave up on experimentation because the math never worked at their scale. They reverted to coin flips and opinions.

The ecommerce shop without millions of sessions. The B2B team that “doesn’t have enough traffic.” The startup that ran two tests, saw flat outcomes, and stopped.

These teams never qualified for someone else’s TAM in the first place.

AI abilities turn them into viable customers.

Customers who work, learn, scale on a neutral experimentation layer that is simple, accessible, and made for speed to unlock their 1st generation of business and strategic wins.

So, one more consolidates.

Convert stays standing. Independent. Mid market. Betting that the next decade of experimentation belongs to teams who learn to test cheaply, often, and at higher velocity.

 Conglomerates tend to have their own growth path. We expect them to do well on it.

Our mission and vision are steeped in the grassroot problems of SMEs.