A/B testing.

Show two versions of your widget to shoppers and keep the one that sells more. Build version B in the same editor you use for your widget, and Daima counts real orders for both versions and tells you in plain words when one is clearly ahead.

Last updated · November 7, 2026

How it works

1. Build version B. Change anything in the same editor you already know: layout, colors, text, the sticky bar.
2. Shoppers are split. Some see your widget as it is now (version A), the rest see B. Each shopper keeps seeing the same version.
3. Daima counts real orders. When one version clearly sells more, Daima tells you. One click makes it your widget.

A test only affects shoppers who are looking at your product page. Existing subscribers and their renewals are not touched. A test never changes your store on its own — nothing is applied until you click.

What you can test

Almost any widget setting. Variant B is edited in the same sections as the widget editor, so you can try a different layout (Tabs, Cards or Compact list), corner style, colors, text and labels, plans and savings, social proof, motion and mobile options, the sticky add-to-cart bar, and variant options. Settings that differ from A are marked with a dot in the section list, and each changed setting shows what A has. You can reset a single setting back to A.

Two things can’t be tested: hiding theme selectors and custom CSS.

While a test runs, B is your live widget settings plus the changes you made for B. If you edit your widget during a test, the widget editor shows a banner: “An A/B test is running — changes here also change version A.”

Starting a test

1. Open A/B Testing in the admin nav and click New test. If you’re not sure where to begin, the page offers four ideas to start with: Cards vs Tabs, Pre-select Subscribe, Sticky add-to-cart bar and Button color. Click Start with this on one to open the new-test page with B already filled in.

2. Give the test a Test name, up to 80 characters. It’s only for you.
3. Under Who sees B, choose 50 / 50, 70 / 30 or 90 / 10. The first number is the share of shoppers who see A. A 50 / 50 split is the fastest way to a clear answer; 70 / 30 and 90 / 10 are safer but take longer.
4. Change settings under Variant B settings. The right side shows your widget (A) and B side by side, with One-time / Subscribe and Desktop / Mobile views, and a count of changes from A.

5. Click Start test. The button stays disabled until B is different from A.

How shoppers are split

The first time a shopper’s browser sees the widget during a test, Daima picks A or B for them using your split, and remembers the choice in that browser. They see the same version every time they come back on that browser. Someone visiting from a different phone or computer may be given a different version.

Orders are counted for the version the shopper saw. The order has to come from a shopper who saw the widget on a product page; orders that never went through a page showing the widget aren’t counted.

Reading the results

The A/B tests page shows your running test with a plain-language verdict, the day the test is on, and a side-by-side tile for each of four numbers. Click See results for the full page:

• Visitors — shoppers who saw the widget
• Orders — orders that came through each version
• Conversion rate — orders divided by visitors
• Subscriptions started — orders that included a subscription
• Chose Subscribe — the share of choices that were Subscribe rather than one-time
• Revenue per visitor — order revenue (before shipping and tax) divided by visitors, in your store’s main currency

You also get a daily chart of the conversion rate (it appears after the test has run for two days) and a list of what’s different in B.

When is a winner clear?

Daima gives one of four verdicts:

• Too early to tell. Each version needs at least 100 visitors and the two together at least 20 orders before Daima compares them. Past that, it keeps saying this until it’s sure enough, and estimates how many more visitors and days that may take.
• B is winning. Daima is at least 95% sure B’s higher conversion rate isn’t luck. For example: “B is winning — we’re 96% sure.”
• A is winning. Daima is at least 95% sure B sold less.
• No clear difference. The test has run at least 21 days and Daima is still under 80% sure. Keep the version you like better, or try a bigger change such as a different layout.

The results page shows how sure Daima is on a bar with a mark at 95%.

Making the decision

You decide when a test ends. Daima never stops a test or changes your widget automatically. The results page offers:

• Make B my widget — B’s settings become your live widget for every shopper, and the test ends. Daima lists the changes in a confirmation first: “Make B your widget?” This is the main button when B is winning.
• Keep A — your widget stays as it is and the test ends with A as the winner. This is the main button when A is winning.
• Stop test — ends the test without a decision. Your widget stays as it is and B isn’t applied. You can also stop from the A/B tests page.

While the verdict is “too early” or “no clear difference,” both Make B my widget and Keep A are offered. You can decide early, but the result may not be reliable. Whatever you choose, the results are kept and show up under Past tests.

Designing a useful test

• Change one idea at a time. If B changes the layout, colors and text at once, you won’t know which change made the difference.
• You need traffic. A test needs at least 100 visitors on each version and 20 orders before Daima will compare them. With low traffic, test bigger changes, like a different layout, not a button color.
• Give it time. Weekday and weekend shoppers behave differently. Let a test run at least a couple of weeks before you trust an early lead.
• Start with offers and layout. Pre-selecting Subscribe or switching layouts usually moves orders more than fonts and colors.