Coframe Completes $25 Million Series A Round: Website Optimization Now Centered on Continuous Experimentation

Website growth teams are never short on ideas. Rewording a headline, moving a case study to the top of the page, or pairing different ads with different landing pages—all of these are worth trying. What really takes time, however, is turning each idea into a version that can be launched, measured, and rolled back if something goes wrong.

On November 9, 2026, Coframe announced that its cumulative funding had reached $35 million, including a newly disclosed Series A round of $25 million jointly led by Khosla Ventures and Hanabi Capital. Investor Antler confirmed its participation on the same day. The announcement also mentioned that certain leaders from OpenAI, Anthropic, and Google DeepMind participated in the investment, but did not specify that all three companies were institutional investors.Company Financing AnnouncementThese ranges are explained.

From the setup page to running a set of experiments

Coframe's services cover conversion rate optimization, landing pages, and ad-related user experiences. According to the process described on its official website, the company develops test plans, creates variants, runs experiments, analyzes results, and then implements the winning solution on the website. The product can be integrated into existing sites using JavaScript snippets, and the company emphasizes allowing clients to review content before it goes live.Product Description on the Official WebsiteSince the human approval process remains in place, “autonomous optimization” should not be interpreted as unrestricted modification of the website.

From a product positioning perspective, it aims to address the collaboration costs involved in design, development, quality assurance, and data analysis. For a site with established, stable traffic, this value is relatively easy to understand: in a given quarter, the team can conduct more experiments based on clear hypotheses, rather than having to queue every minor change for development.

What can we learn from the Replit case study?

Antler's Investment StatementIt states that Coframe ran experiments on Replit’s homepage, pricing page, enterprise page, and paid landing page, and reported a 410% increase in enterprise funnel conversions and a doubling of demo requests. These figures come from the investors’ presentation; the page does not provide sufficiently complete details regarding the experiment’s sample size, time frame, or statistical methodology. Therefore, while they are useful for understanding the general direction of the case study, they do not guarantee that other clients will achieve the same level of growth.

It’s especially important to distinguish between “a 410% increase in conversion rate” and “an increase of 410 percentage points.” The former describes a relative change; you cannot determine the absolute growth solely based on the percentage. A single form submission in a company’s sales funnel is also different from a final contract signing. When reviewing growth case studies, you need to know at least the starting point, the definition of conversion, the composition of traffic, and the final business results.

How Should a Well-Designed Pilot Program Be Implemented?

Taking the demo booking page of a software company as an example, you can start by selecting a stable customer acquisition channel and clarifying that the goal is qualified bookings rather than all clicks. Next, formulate specific hypotheses—such as whether industry case studies can alleviate visitors’ concerns—and then create different versions of the page. Only after the brand, legal, and sales teams have reviewed and approved the promises and wording should you begin directing traffic to users.

During testing, you should also monitor form errors, page load times, the percentage of low-quality leads, and subsequent conversion rates. If a new page makes it easier to submit forms but results in sales teams receiving a large number of invalid inquiries, looking solely at the number of appointments booked could lead to overly optimistic conclusions. This pilot process is a recommendation for product usage; it is not a guarantee of compliance with Coframe’s published standards.

Having AI quickly generate more options shifts the problem in a different direction: Does the team have enough traffic to distinguish between genuine improvements and random fluctuations? If there are very few visitors and too many versions are being tested simultaneously, it may actually take a long time to reach robust conclusions. While the speed of experimentation can be increased, the time required to accumulate evidence cannot be saved out of thin air.

Personalization Also Needs Boundaries

Showing different pages to different visitors can sometimes reduce the effort required to find information, but it is essential to ensure that core commitments remain consistent. Key information—such as product prices, refund policies, and capability limitations—should not be obscured in the pursuit of optimization goals. Businesses also need to know which visitor signals the system uses, how long they are retained, and how consent and opt-out requirements are enforced.

Antler is increasingly bullish on the web experience for AI agents. This is an investment judgment: as more agents compare products on behalf of users, web pages may need to present prices, terms, and features more clearly. However, an increase in bot traffic does not automatically mean that all traffic leads to transactions, nor can it replace the analysis of actual customer behavior.

Coframe’s value lies in turning the process of formulating hypotheses, launching tests, and analyzing results into a continuous workflow. The ultimate savings for a company depend on the combined effects of changes in net revenue, lead quality, experiment interpretability, and brand risk. For growth teams, the best starting point remains an experiment with clearly defined boundaries, sufficient data, and the ability to stop at any time.