A split test, or A/B test, is the only honest way to know whether a change to a page helped. You show two versions to different visitors at the same time and count what each one does. This guide walks through testing one step of a funnel in Vonza Funnels: what to test first, how to set it up, how to read it, and when to stop.
What a split test is, and is not
In Vonza Funnels a split test is described as "Show two versions of a step's page to different visitors and keep the one that converts better." A few things follow from that:
- It tests one step's page, not the whole funnel at once.
- Both versions run at the same time, so a good week or a bad ad does not favour one.
- Each visitor keeps seeing the same version, so nobody sees the price change on refresh.
- It counts a metric that fits the step: an opt-in on an opt-in step, an order on a checkout step, and a visit to the next step on a sales page or a free-form page.
Split tests exist on funnel steps. A landing page built in the Vonza website builder, a website page or a standalone checkout page does not have them, so if you want to test a page, build it as a funnel step.
What to test first
Start where the numbers are. Open the funnel's Stats and look at the step-by-step view, which shows how many people reached each step and where they dropped off. The step with the most visitors and the biggest drop is the one to test, because a small improvement there carries through every later step.
Then test changes in this order, one at a time:
- The headline. The biggest single lever on most pages.
- The offer itself. On an opt-in step, the freebie; on a sales step, what is included.
- The layout. Any step can switch design with Change design; a split test lets you try a second design without guessing.
- The button label. Smaller effect, but cheap to try.
- The price card. Which plan shows first, or whether an annual option appears.
Do not test two things at once. If the headline and the picture both change and the variation wins, you do not know which change did it.
Setting up the test
From the canvas, open the step and start a split test. The app walks you through:
- Name the versions. The control is the page as it is; the variation is the copy you will change. Names are up to 40 characters; "Control" and "Shorter headline" is enough.
- Edit the variation. Make the one change.
- Set the traffic share. Each version gets a whole-number share between 1 and 99 per cent. For a fair test, 50/50 is the default choice. Send less to the variation only when a wrong guess would be expensive, for example on a checkout step.
- Start. The variation goes live when you publish the funnel. The canvas marks the step with an A/B chip so you can see which steps are under test.
The split test is written for a second version, which is the right number for almost every creator. One variation, one change, one clear answer.
Reading the result
The test counts the step's metric for each version: opt-ins, orders, or visits to the next step. Resist reading it daily. Three rules keep you honest:
- Wait for a full week at least. People behave differently on a Sunday evening and a Tuesday morning.
- Count conversions, not percentages. A version at 12 per cent with eight conversions is not beating a version at 9 per cent with six. Both need many more before the gap means anything.
- Expect small differences. A headline change rarely doubles a rate. A reliable few points is a good result.
If you want a number for "enough", the A/B test calculator takes the visitors and conversions of both versions and tells you whether the difference is likely to be real. Use it before you declare a winner.
Finishing the test
Three buttons end or pause a test:
- Stop. Everyone sees the control until you resume. Use it if the variation has a problem you need to fix.
- Resume. Carries on where it left off.
- Pick winner. Every visitor sees the chosen version from now on and the test ends; the other version's page is removed. Pick the winner only after the rules above are satisfied.
After a winner, write down what you changed and what happened. A short log of tests, even in a notes file, is the most useful marketing document you will own after six months.
A worked example
A coach has a lead magnet funnel with an opt-in step and a thank-you step. Stats show 1,000 unique visitors a month to the opt-in page and a low rate of sign-ups. The headline reads "Free coaching resources". The variation reads "The 5 questions to ask before you raise your coaching prices". Traffic 50/50. After two weeks the variation has clearly more opt-ins from about the same number of visitors; the calculator says the difference is unlikely to be chance. Pick winner. The next test is the freebie itself: the same headline, with the PDF replaced by a 10-minute video.
The figures in this example are made up to show the method; they are not results.
Tests that are not worth running
- A test on a step with very few visitors. It will not finish. Fix the traffic first.
- A test of a colour. Unless the button is invisible, the effect is too small to measure.
- A test where both versions are weak. Write the best page you can first, then test against it.
- A test you will not act on. If you would keep the control regardless, do not test.
Where split tests fit in the funnel
Split tests improve one page. The funnel's other tools cover the rest: the Tracking tab for your pixel and Google tag, so ad platforms learn from the winners (see funnel tracking pixels); automations for the people who did not convert; and Stats for the whole picture. The sales funnel page shows how they fit together, and the funnel conversion calculator helps you see what a few points at one step do to the total.
Open Funnels, find the step with the biggest drop-off and start your first test. New to Vonza? Start a 14-day free trial from the sales funnel page.
