Read the tests you already run correctly
Omniconvert Explore reports significance by segment, with revenue per visitor and AOV in every result.
Placeholder values below. Replace them with your own test data.
See how Explore reads results by segment ↓
Simple arithmetic on the numbers you enter, direction only. It does not test significance, and it is not a benchmark.
Built on data from 2,500+ eCommerce brands
In shortRead your A/B tests correctly. Omniconvert Explore reports significance by segment, shows revenue per visitor and AOV in every result, and layers customer value on top of raw test data.
How safely do you read a test result today?
Tick what is true today. Be honest, nobody is watching.
A flat blended result can hide a clear, significant win
A test that looks flat overall can be a win for mobile visitors, paid traffic or returning customers, and a loss elsewhere. Reading only the top line gets it killed.
Illustrative example, not a benchmark. Ship it to everyone and Segment B may lose out. Kill it and Segment A misses the gain.
How Explore makes a test result trustworthy
The difference between running tests and reading them right comes down to three things.
Explore reports significance by segment, not just the blended result
- A test that looks flat overall can be a clear, significant win for mobile, paid or returning visitors
- It can lose elsewhere, and a blended number averages the two away
- You neither ship a false win nor discard a variant that works for the audience that matters
Your team still: chooses which segments matter to the business.
Explore shows revenue per visitor and AOV in every result
- Conversion rate can move in the opposite direction to revenue
- Revenue per visitor and average order value sit beside it on every variant
- A conversion lift that quietly lowers order value is visible before it ships
Your team still: decides whether the trade-off is worth shipping. The revenue-per-visitor page covers this in depth.
Omniconvert Commander layers customer value onto your test data
- A lift is not equally valuable if it comes from one-time discount hunters
- RFM and customer-value context shows who actually converted
- You can tell a lift in valuable customers from traffic that never returns
Your team still: decides how much a lift is worth once you know who it came from.
Reading your tests, with and without Explore
| Task | Without | With Explore |
|---|---|---|
| Reading significance | Often the blended result only | Per variant and per segment |
| Revenue view | Conversion rate on its own | Revenue per visitor and AOV beside it on every variant |
| Flat tests | Can be discarded on the top line | Can be checked for segment wins |
| Who converted | A lift with no customer context | RFM and customer value from Omniconvert Commander layered on |
| Calling winners | GA4 shows what happened | Significance computed for each variant and segment |
Avon read the result by segment and market, not on one blended number
Avon read a multi-country test at the segment and market level rather than on a single blended number, then rolled out the winning variant across markets.
Omniconvert, Avon multi-country testing case study
Two things worth knowing upfront
Isn't this the same as Scale Experiments?
No. Scale Experiments is about running more tests; this is about reading the ones you already run correctly. You can run a high volume of tests and still make the wrong call on each one if you only ever look at the blended number.
Don't we already have Google Analytics or GA4 for this?
GA4 tells you what happened on the site. It does not natively compute test significance or tie a result to segment-level customer value, so it can show a conversion lift without telling you whether it came from customers worth keeping. This adds that layer.
Common questions
How do I know if an A/B test result is statistically significant?
Statistical significance means the difference between your variant and control is unlikely to be down to chance. In practice most teams use a 95 percent confidence level, which corresponds to a p-value below 0.05. Reaching it reliably requires a planned sample size and running the test to its intended duration rather than stopping on the first promising day.
What sample size do I need for an A/B test?
It depends on your baseline conversion rate, the smallest effect you want to be able to detect, and the confidence level you require. Lower baselines and smaller effects need more traffic. Use a sample-size calculator before launching so you know the duration up front instead of deciding when to stop mid-test.
Why should I read A/B test results by segment?
A blended result can average away opposite effects: a variant might win strongly for mobile or returning visitors and lose for everyone else, netting out to flat. Reading results by segment surfaces those cases, so you keep what works for the audience that matters and avoid shipping a win that only exists on paper.
What is the difference between statistical and practical significance?
Statistical significance tells you an effect is probably real rather than noise. Practical significance tells you whether it is large enough to matter for revenue. A change can be statistically significant but too small to be worth shipping, so read both together rather than treating significance as a green light on its own.
Can GA4 tell me if my A/B test won?
GA4 reports what happened on your site, but it does not natively run significance testing or connect a test outcome to segment-level customer value. A dedicated testing tool computes significance for each variant and segment and ties the result to who converted, which GA4 does not do out of the box.
How does Omniconvert Explore report statistical significance?
Omniconvert Explore reports significance for each variant and for each segment, alongside revenue per visitor and average order value, and layers Omniconvert Commander customer-value context on top, so you can judge not just whether a result is real but whether it came from customers worth keeping.
See what your test results are actually telling you
30 minutes. We'll look at how your current tests are being read, not a demo store.