A/B testing
Test the changes that matter, and stop calling noise a win.
A/B testing — variant list with conversion rate, sample size and confidence
Product screenshot — to be swapped in
The operating brief
Traffic is only half of the equation. A site converting at one per cent needs twice the traffic of a site converting at two to earn the same result, and improving the second number is usually cheaper. A/B testing in Helix is how the agents earn conversion rate rather than assuming it.
The hard part of testing is not running the test, it is not fooling yourself. Small sites generate small samples, and small samples produce impressive-looking differences that vanish next month. This module is built around that problem.
Capabilities
What the a/b testing agent does
01
Page and element tests
Headlines, hero copy, form length, call-to-action wording, proof placement and pricing presentation, split at the visitor level with sticky assignment across sessions.
02
Ad asset tests
Headline and description variants proposed from your best-performing organic content, tested inside the ad platform and retired on evidence rather than opinion.
03
Sample size set before you start
Helix calculates the traffic needed to detect a meaningful lift given your current conversion rate, and tells you upfront if the test will take four months. That is often the most useful output.
04
Honest significance reporting
Confidence intervals rather than a green tick, no peeking-based early calls, and a clear label when a result is inconclusive. Inconclusive is a legitimate outcome.
05
Winners rolled out, losers reverted
Approve a winner and the change ships to the live page or campaign with a full log. Losing variants are removed cleanly, and the result is kept so the same idea is not retested next quarter.
How it works
From connected to running
01
Pick a target
Usually the page with the most traffic and the worst conversion rate.
02
Size the test
Helix estimates duration from your real traffic before anything runs.
03
Build variants
The agent proposes variants that follow your brand profile.
04
Run
Traffic splits evenly with stable assignment; results update daily.
05
Decide
Ship the winner, revert the rest, and keep the finding on record.
When not to test
If a page gets forty visits a month, testing it is theatre. At that volume you would need a very large effect and many months to say anything defensible, and the honest move is to make the obvious improvement and measure the trend instead.
Helix says this plainly rather than running a test to look busy. Low-traffic sites get recommendations and before-and-after tracking; testing is recommended when the traffic can support it.
What a test cannot tell you
A test compares two options on one page for one audience during one period. It does not tell you whether a third option would beat both, whether the lift holds through a seasonal change, or whether the extra conversions were the customers you actually want.
Results are reported with those limits attached, and lifts are tracked against downstream revenue where the CRM connection makes that possible — a variant that produces more leads and fewer deals is not a win.
Plan availability
Add-on for Starter and Growth; included on Agency.
Starter $199 · Growth $469 · Agency $1069 per month, or $159 / $372 / $852 a month billed yearly (2 months free).
Compare plans →Works alongside
Paid ads autopilot
Google Ads optimised every day against your rules, with a full log of every change.
Explore Paid ads →Leads & attribution
Follow the journey from first click to a scored prospect, and connect real revenue back to the work that caused it.
Explore Leads & attribution →Start with a free scan.
No credit card, no call. You will see what the agent sees.