Skip to content

Troubleshooting ​

Find your symptom, then follow the fix. Still stuck? Email team@getdalton.com with the page URL, your browser and a screenshot.

SymptomGo to
Verify doesn't find the scriptInstall
"Identifier has already been declared" in the consoleInstall
Purchase says not detectedInstall
A running test doesn't show on your siteA test doesn't show
The agent can't change an elementEditor
A variant looks wrong on mobileEditor
Results show a dashResults
The test ended with Not enough dataResults
Numbers differ from Shopify or GA4Results
Preview looks different from the live sitePreview

Install ​

Verify doesn't find the script ​

  • Check the snippet is in the <head> of the published theme or template, as high as possible.
  • Clear your site's cache (and any caching plugin), then reload in a private window.
  • Look for dalton-installation-script in the page source. If it's missing, the snippet isn't on the page.
  • In your browser's developer tools, open Network and search for dalton. The script loads from cdn.getdalton.com. If it's blocked, see CSP and firewalls.

The snippet is on the page twice ​

The browser console shows Identifier 'DALTON_ID' has already been declared. The snippet was added twice, for example in your theme and in Google Tag Manager. The first copy keeps working and tests still run, but remove one to clear the error.

Purchase says not detected ​

Dalton hasn't received a purchase recently, so tests can't resolve. Set up purchase tracking for your platform in Installation step 3, place a test order, and check Settings → Conversions again.

A test doesn't show ​

Work down the list:

  1. The test is running. On the Experiments page it shows Collecting, not Draft or paused.
  2. The page is in the test. Open the test's targeting group and type the page's URL in the search box to confirm it matches. Paths are case-sensitive.
  3. You're in the audience. A test for mobile or paid traffic won't show on a desktop visit from a bookmark.
  4. You may be in the original. Visitors are split, and you keep your variant for the visit (24 hours). Open a private window to be assigned again, or use Preview in the editor to force a variant.
  5. Something blocks the script. Some ad blockers and privacy extensions block testing tools. Try another browser.

Check what a visitor sees ​

Add one of these to any page URL on your site:

ParameterWhat it does
?dalton_disable=1Turns Dalton off for that page view, so you see your original page
?dalton_debug=1Logs what Dalton does to the browser console

Editor ​

The agent can't change an element ​

Content inside iframes, web components and videos is hard to reach, and parts that your site fills in after loading (prices, stock, size selectors) may need Edit code. Try the element around it, or describe the change more precisely: "make this one sentence" works better than "make it shorter". See Harder to change.

A variant looks wrong on mobile ​

Check every variant with Preview on mobile before launch. If the change only makes sense on one device, ask the agent for a mobile-specific version, or limit the test to one device in the audience.

Results ​

Results show a dash ​

Figures appear once every variant, the original included, has 350 visitors and 50 conversions. Before that, the numbers are too thin to read. If conversions stay at zero, check that Purchase isn't not detected.

Not enough data ​

The test ran its full window without reaching the data bars. The page is too quiet for that question, and running longer won't fix it. For the next test:

  • Widen the targeting group, such as every product page instead of one.
  • Test fewer variants, so each gets more traffic.
  • Check time to results in the targeting picker before you launch.

See How experiments run.

Numbers differ from Shopify or GA4 ​

That's expected: each tool counts visits, orders and revenue differently. Every variant is measured the same way, so the comparison between them holds. See Why numbers differ.

Preview ​

Preview looks different from the live site ​

Preview forces the variant on in your own browser, where a cached page, a logged-in session or a browser extension can change what you see. Open the preview link in a private window. On the live site, a tested element can wait up to 1 second for its variant (see anti-flicker).

Preview shows the same variant ​

Preview always shows the variant selected in the editor when you opened it. Select the other variant, then open Preview again. The original has no Preview button; select A to see it on the canvas.

Other tools on your site ​

  • Another A/B testing tool (VWO, Optimizely, your own tests): don't test the same page with two tools. Both change the page, and neither can tell which change caused a result.
  • Heatmaps and session recordings (Hotjar, Microsoft Clarity): work alongside Dalton.
  • Speed and caching plugins: if a plugin delays or bundles scripts, exclude the Dalton snippet from it so it still loads early.