Statsig and Amplitude: What the 2026 Handover Means for Your Experiments

Amplitude announced on May 5, 2026 that it is taking over the Statsig brand and customer base, while the Statsig engineering team remains at OpenAI following OpenAI's acquisition of the company in September 2025. Amplitude committed to maintain and develop the current Statsig platform across cloud and data warehouse, including support for existing customers, and said it will begin building a more integrated roadmap for the two platforms together. No sunset date, forced migration, or price change has been announced. If you run Statsig today, nothing breaks right now, but the ownership and the roadmap both changed, and that is worth planning around. Last updated July 2026.

This is the second ownership change in under a year for a platform many teams put at the center of their release process. That is unusual enough to justify a proper look rather than a shrug, so here is what actually happened, what it plausibly means, and what a sensible team does about it this quarter.

What happened to Statsig?

Two separate events, often mashed together in coverage. In September 2025, OpenAI acquired Statsig for a reported $1.1 billion and named founder Vijaye Raji its CTO of Applications. At the time, OpenAI said Statsig would keep operating independently from its Seattle-area office and continue serving existing customers. Then in May 2026, Amplitude announced it would assume the Statsig brand and customer base, while the team stayed at OpenAI.

The part worth sitting with is the third row. Amplitude gets the code, the brand, and the customers. OpenAI keeps the people. That combination is rare, and it is the source of the criticism that circulated after the announcement: a platform inherits its judgment from the people who built it, and rebuilding that judgment in a new organization takes time. That is a risk assessment, not a fact about the product, and Amplitude did explicitly commit to maintaining and developing the platform. Treat it as a reason to keep your options open, not a reason to panic.

Do I need to migrate off Statsig?

Not urgently, and probably not this quarter. Nothing has been sunset, support was explicitly committed to, and a rushed migration of experiment infrastructure is expensive and error-prone. The better response is to reduce your switching cost so that a future move, if it ever becomes necessary, is a two-week project instead of a two-quarter one. That is cheap insurance and it pays off even if you never leave.

Three concrete steps. Keep your event schema documented and portable, so it is not encoded only inside one vendor's UI. Keep experiment and flag definitions in version control rather than clicking them into a dashboard and forgetting the reasoning. And export historical experiment results on a schedule, because past results are the asset you will miss most if access ever changes. If your event data lives in a warehouse, it is also worth being able to trace where each metric actually comes from, since knowing which upstream table feeds a given metric is the difference between a migration you can plan and one you have to reverse-engineer.

What is the best Statsig alternative?

It depends which part of Statsig you actually use. The platform bundled feature flags, experimentation, and product analytics, and most teams lean heavily on one of the three.

If flags are the dependency, GrowthBook is the pragmatic replacement and costs nothing self-hosted, though you pay in engineering time. Our comparison of a GrowthBook alternative covers where that trade lands. If the draw was the integrated analytics, PostHog covers the most ground in one self-serve product, and the PostHog alternative breakdown lays out what it does and does not replace. And if you are now evaluating Amplitude itself as the destination, the Amplitude alternative comparison explains its MTU pricing shape before you commit to it.

Will Amplitude raise prices for Statsig customers?

Nothing has been announced, and it would be irresponsible to claim otherwise. What is knowable is the pricing shape on each side. Statsig meters events: flag checks are unlimited and free at every tier, while analytics, experiment metrics, and session replays draw down metered event volume, with a free tier, a Pro plan around $150 a month, and custom Enterprise. Amplitude prices on Monthly Tracked Users, with a free Starter plan to 10,000 MTUs and 2 million events a month, Plus around $49 a month billed annually to 300,000 MTUs, and sales-quoted Growth and Enterprise tiers above that.

Those are different meters, and any eventual convergence of the two product lines has to reconcile them. That is not a prediction of an increase; it is a reason to know your own event and MTU volumes now, so that whenever renewal comes you can evaluate a quote in an afternoon instead of guessing.

What this changes about how you run experiments

Honestly, very little at the mechanical level. What it should change is where you put your attention. Teams routinely spend months evaluating experimentation platforms and then run four underpowered tests a year, which is the wrong end of the problem. Vendor churn is a good prompt to check whether the experimentation program was earning its cost in the first place.

The math is unforgiving. A reliable test wants roughly 1,000 conversions per variation per month to detect a realistic effect, which on a page converting at 2 percent means about 50,000 visitors per variation. Plenty of teams carrying an experimentation license do not clear that bar on their best page. If that is you, the platform was never the bottleneck, and switching platforms will not fix it. Our guide to A/B test sample size walks through the arithmetic for your own numbers.

The gap no experimentation platform fills

Statsig, GrowthBook, PostHog, Amplitude, and every other tool in this category answer one question well: which of these versions performed better. None of them answer the question that comes first, which is what to try. You still have to look at the page, decide the headline is vague, decide the CTA is buried below three paragraphs, decide the form asks for a phone number nobody wants to give, and then write the replacement. That work is where most conversion gains actually come from, and it is entirely manual in every platform listed above.

That is the step Converto handles. Paste a live URL and it reads the page the way a skeptical first-time visitor would, scores copy, layout, CTAs, page speed, and AI-readability, ranks the leaks by likely impact, and generates ready-to-ship variants for the weak parts. Nothing to install, no events to instrument, no warehouse, self-serve from $19 a month. It does not replace your experimentation stack. It produces the challenger that stack is waiting for, which means when your traffic does support a test, you arrive with a real hypothesis instead of a blank editor. Suggestions are data-informed, not guarantees.

A sensible plan for the next 90 days

Keep running Statsig. Document your event schema and move flag and experiment definitions into version control if they are not there already. Export historical experiment results once, then quarterly. Note your event volume and MTU count so you can read any future quote quickly. Watch for two specific signals that would change the calculus: a published end-of-life schedule, or a merged Statsig and Amplitude product with a single pricing model. Neither exists today.

Then spend the time you would have spent on vendor anxiety on the page itself. Run your highest-traffic landing page through a landing page audit, fix what the audit finds, and measure the result with whatever platform you end up on. The tooling question resolves itself in time. The page is losing money today.

The takeaway

Amplitude took over the Statsig brand and customers on May 5, 2026 and committed to maintaining the platform; the team stayed at OpenAI; nothing is shutting down. Do not migrate in a hurry, do reduce your switching cost, and know your volumes before your next renewal. And remember that no version of this decision changes the part that actually moves conversion rate: someone still has to write a better page than the one you have.