Why traditional A/B testing is broken

Most companies treat A/B testing like a special event. Someone has an idea, the team spends two weeks building a variant, they run it for a month, interpret the results over a meeting, and then… go back to whatever they were doing before.

The average company runs 2-3 A/B tests per year on their landing page. At that rate, you’d need a decade to find your optimal conversion rate.

The bottleneck isn’t the tool

Optimizely, VWO, Google Optimize (RIP): they all make it easy to run tests. The bottleneck is everything around the test:

• Coming up with hypotheses: requires someone to look at data, think about what might work, and propose a change
• Building variants: even with visual editors, someone has to design and implement each variant
• Waiting for significance: most sites don’t have enough traffic to reach significance quickly
• Interpreting results: was it the headline? The button colour? The layout? Hard to isolate
• Deciding what to test next: back to square one

This entire loop is manual. It depends on humans having time, attention, and expertise. For most teams, that means it doesn’t happen.

What if the whole loop was automated?

Imagine an AI that:

1. Looks at your analytics and identifies what’s underperforming
2. Generates hypotheses based on patterns it sees across thousands of sites
3. Creates variants and deploys them automatically
4. Measures results with statistical rigour
5. Promotes winners without waiting for a meeting

That’s not science fiction. That’s what Darwin does.

The key insight is that A/B testing is fundamentally an evolutionary process. You have a population (your visitors), mutations (variants), selection pressure (conversion rate), and survival of the fittest (promotion). Evolution doesn’t need a committee to approve each mutation. It runs continuously, and the best adaptations win.

The maths of continuous testing

If you run one experiment per week instead of one per quarter, you get 13x more learning per year. If each winning experiment improves conversion by just 2%, that compounds:

• After 10 wins: +22% conversion
• After 20 wins: +49% conversion
• After 50 wins: +170% conversion

The difference between “we test occasionally” and “we test continuously” is the difference between linear and exponential improvement.

The future is autonomous

We built Darwin because we believe landing page optimisation should work like evolution: continuous, autonomous, and relentless. You shouldn’t need a CRO team to have a great conversion rate. You should just need one script tag.

Join the Darwin waitlist to be among the first to try it.