On September 30, 2023, Google sunset Optimize — the free testing tool a huge swath of the web ran on. The teams that relied on it had to choose: pay for a platform, or stop testing. What buyers tell us about the choice is consistent — and the most revealing part isn't which tool they picked, but how few tests they were running in the first place.
Where the Optimize refugees landed
Three tiers absorbed the migration — and most teams sort themselves by program maturity and how much engineering support they have, not by brand.
The migration sorted teams by maturity — not everyone needed the tier they thought they did.
How buyers rate the platforms
The striking pattern: the lower-cost challengers match or outscore the enterprise incumbent. Praise clusters on ease of use and testing depth; criticism clusters on cost, complexity, and support.
| Vendor | Avg rating | What buyers say |
|---|---|---|
| 7.6 | Strong testing depth; buyers flag it can run "too heavy" on the page and is complex to QA. | |
| 7.3 | The value pick — "similar functionality to Optimizely, more budget-friendly"; support is the soft spot. | |
| not rated | Liked for personalization and a developer-friendly approach; too few buyers in our interviews to publish a rating. | |
| 7.3 | The enterprise default; powerful but "the most expensive in the market," and called "compartmentalized." |
Path #1 — pay up the enterprise ladder
The default for big programsOptimizely & Adobe Target
Optimizely is where large programs land by default — full-stack experimentation with the breadth enterprises need. Buyers praise the A/B testing itself, but the score sits in the middle for a reason: it's "the most expensive one in the market," and several describe the platform as "compartmentalized," with a "sense of fragmentation" and rough onboarding.
The tell, in more than one interview: buyers noticing "less investment" in the core web-experimentation product even as the price holds — the classic incumbent complaint.
Path #2 — trade down to a value tool
Most of the function, less of the billVWO, Convert & Kameleoon
The most common landing spot for mid-market teams. Buyers choose VWO explicitly against Optimizely — "similar functionality" at a price that "makes it a competitive choice" — and value how easily non-technical users can run tests without engineering.
The trade-off they name is support and polish: slow response times and the occasional glitch on tasks that "should have been straightforward." For programs that test steadily but not at enterprise volume, buyers report it's a trade worth making. The head-to-head is its own read: Optimizely vs. VWO — what buyers say.
Path #3 — move testing into code
The frontierFeature flags & server-side
For teams with engineering muscle, experimentation is merging with feature management. Statsig, LaunchDarkly, and GrowthBook run tests server-side, in the deploy pipeline, rather than through a marketer's visual editor — the same flag that ships a feature measures it. It's the path product- and growth-engineering teams increasingly describe, and the one the visual-editor incumbents are racing to answer.
The counter-current: the tool was rarely the problem
The most uncomfortable pattern in the interviews isn't which platform — it's how little testing actually happens. Buyers describe a handful of tests a quarter, programs without the traffic to reach significance, and tools bought at enterprise scale for SMB-scale usage. Some, after Optimize, simply went without for a while and barely noticed. Before the platform question sits a harder one: does your program run enough tests, on enough traffic, to justify any of these?
What this means for your program
Match the tool to the program, not the brand. If you run a high-velocity practice with engineering support, the enterprise platforms — or a code-native stack — earn their keep. If you run a handful of marketer-built tests, a value tool likely delivers the same wins for a fraction of the cost; buyers told us so directly. The expensive mistake the interviews surface isn't picking the "wrong" tool — it's buying enterprise experimentation for a program that isn't running enough tests to amortize it. The honest first question is about your test velocity and traffic, not the vendor shortlist. For the hard-won lessons from buyers who've already been through it, see what experimentation-tool buyers wish they'd known; and for the 2026 switching triggers that have nothing to do with the Optimize exit — including experimentation moving to developer-owned platforms — see why teams switch A/B testing tools.
Common questions
What happened to Google Optimize?
Google sunset it on September 30, 2023. Teams that ran free A/B testing on it had to choose a paid platform or go without — the inciting event behind much of the movement in this category.
Optimizely vs. VWO — how do buyers choose?
Optimizely is the enterprise default but the most-cited on cost and complexity ("the most expensive in the market"); VWO is the value alternative chosen for "similar functionality" at a friendlier price, trading away some support depth. Program scale and engineering support decide it.
Are feature-flag tools replacing A/B testing platforms?
For teams with engineering muscle, increasingly yes — experimentation is converging with feature management (Statsig, LaunchDarkly, GrowthBook), moving tests server-side into the deploy pipeline rather than a visual editor. Marketing-led teams still favor the visual platforms.
Which experimentation tools do buyers rate highest?
In Alium's corpus through July 2026: AB Tasty 7.6/10, VWO 7.3/10, and Optimizely 7.3/10, with Kameleoon rated similarly by too small a group to publish — notably, the lower-cost challengers match or outscore the enterprise incumbent.
This is the aggregate. Your stack is specific.
Do a 15-minute interview about your own stack and get this analysis personalized — what peers rate the experimentation tools you run, who's trading down or moving testing into code, and whether your program is sized for the platform you're paying for.
Get my personalized briefNo password needed · your interview is anonymized before it ever informs a page like this one.