This is what a Twinpage result looks like.
Exposures and conversions per version, and a verdict that only says "likely better" when the data supports it. Plus two checks: is the traffic split right, and is there enough data yet.
Join the waitlist: https://t.co/yxlN6R26Sc
Every reliable test follows the same four steps: form a hypothesis, build the variants, split the traffic, and read the result once there is enough data.
Try it free: https://t.co/MgYpi4ijp1
#ABTesting#CRO
A friend just ran the first Twinpage test on a site I don't control.
Worked: the script went in once and the install check turned green.
Didn't: the goal step wasn't obvious. That's next on my list.
Lesson: setup is where people decide to trust a tool.
https://t.co/yxlN6R2EHK
A/B/n testing compares an original with several variants at once, for example three headlines against the current one, with traffic shared equally.
Try it free: https://t.co/arqoNcLMPL
#ABnTesting#ABTesting
Reliable statistics come first on our roadmap. The editor, integrations and alerts follow.
Join the waitlist: https://t.co/YVnYoMAPDM
#Experimentation#GrowthMarketing
It also waits for a minimum sample before it says anything, and flags a test whose traffic split looks wrong.
That means a slower "yes" sometimes. I'd rather that than a confident wrong answer.
Join the waitlist: https://t.co/yxlN6R2EHK
#ABTesting
The first thing I built for Twinpage wasn't the editor. It was the statistics.
Most people stop an A/B test the first day it turns green. Checking daily with a normal significance test pushes false positives well above 5 %.
Twinpage's test stays valid however often you look.
Many of the rules in our Claude Code setup never caught a single bug.
So every 10 features, a prune audits each rule and removes the ones that never fired. The agents got better at following the rules that were left.
What 153 features taught us, in five lessons:
Morphness is open source.
It's the Claude Code harness behind 153 features of Twinpage: a lead agent plans, a builder builds, QA checks criterion, a retro adds rules, and a prune deletes the ones that never fire.
Installs in 3 commands. If it's useful, a star helps others find it
Stopping an A/B test at the first green reading is one of the most common causes of false winners.
Join the waitlist: https://t.co/FuJXfkWbiG
#ABTesting#Experimentation
Many people I know can't tell what appeals to their website visitors. They aren't marketers or developers, and A/B testing tools are expensive and hard to start.
I hit the same wall, and so did two friends. So I'm building Twinpage.
Join the waitlist: https://t.co/2rZjxyBmbP
A/B testing compares an original page with one changed version. Visitors are split between the two, and the version that performs better on your goal wins.
Join the waitlist: https://t.co/rVlRuDvv9O
#ABTesting#CRO