Appearance
What Dalton can do
In short
- Find: the CRO Agent spots what to test.
- Build: the editor agent makes the variants, no developer needed.
- Test: A/B tests on one page or a whole template, winners called on conversion rate or revenue per visitor.
- Target: an audience, a country or a market, with copy translated per market and winners deployed only where they won.
- Connect: one snippet on any site, purchases from Shopify or GA4, results in GA4 and on your Shopify orders, and the MCP server for your own AI.
Find and build
| You want to... | How in Dalton | Page |
|---|---|---|
| Find what to test | The CRO Agent scans your store and builds a test for each problem it finds | CRO Agent |
| Build a variant without a developer | Describe the change to the editor agent, or edit the code by hand | Using the editor |
| Test a new image | Put the image on your store or CMS, then give the agent its URL | Testing images |
| Keep variants on-brand | Dalton learns your look, tone and rules, and the agents follow them | Brand Intelligence |
| Build variants in your Shopify theme (developer) | Your developer builds both versions in Liquid; the variant flips a switch | Below |
Test and target
| You want to... | How in Dalton | Page |
|---|---|---|
| Test one page against a variant | The original against one or more variants, on a split fixed for the test | Your first experiment |
| Test a whole new page against the current one | A URL split test redirects part of the visitors to your new page | URL split tests |
| Test several pages or a whole template at once | One targeting group that covers every page | Testing across many pages |
| Test only on paid traffic, email traffic or mobile | Click the audience in the editor and pick Device, Traffic or Visitor | Personalise by audience |
| Test only for one UTM campaign | Advanced → Query parameter, for example utm_campaign is spring-sale | Personalise by audience |
| Change a headline per URL parameter value, such as one per campaign | Ask the editor agent; it maps each value to its own headline | Personalise by audience |
| Run a test for one or more countries | Ask the editor agent, such as "only for visitors in Belgium and the Netherlands" | Targeting |
| Combine audience rules with AND or OR, such as mobile or paid traffic | Ask the editor agent, or set it through the MCP server | Targeting |
| Translate a variant for every market | The editor agent writes the copy in each market's language, picked from the domain or locale path | One test across markets |
| Run one test across several countries or domains | One targeting group with rules on each domain or locale path | One test across markets |
| Keep a winner only where it won | Deploy → Smart deploy or Custom, by device, channel or country | Deploy to a segment |
| Send each Meta ad to a page version matched to its message (beta) | Landing Pages, with Meta connected and a subdomain pointed at Dalton | Landing pages for ads |
Results and data
| You want to... | How in Dalton | Page |
|---|---|---|
| Know when a test is done | Dalton fixes a 2, 4 or 6 week window before launch, from your traffic | How experiments run |
| Judge on revenue, not only conversion rate | Winners are called on conversion rate or revenue per visitor | Reading your results |
Install and connect
| You want to... | How in Dalton | Page |
|---|---|---|
| Install on your site | One snippet in the <head>. Works on Shopify, WordPress, Wix, React, Next.js or any site | Installation |
| Count purchases and revenue | The Dalton Pixel on Shopify, your GA4 ecommerce events, or the pixel on your order page | Make sure purchases arrive |
| Check results in GA4 or your own tools | Every visitor in a test gets a dalton_experiment event in the data layer | Analytics integration |
| Split Shopify orders by variant | The variants a buyer saw are added to the order, with the Shopify app or a few lines of theme code | Variants on Shopify orders |
| Drive Dalton from Claude, ChatGPT or Cursor | Connect the Dalton MCP server | MCP server |
Two ways to add logic
Targeting decides who enters the test. Device, traffic source, new or returning visitor, country, a query parameter or the referrer, combined with AND or OR. Visitors outside the audience see your page as it is and don't count in the results.
Variant code decides what each visitor sees. A variant can read the page URL, the path, the domain and the page itself. So one variant can show a different headline per campaign, or Finnish copy on /fi-FI/ and German copy on /de/. You don't write this code. Describe the logic to the editor agent and it writes it.
Use targeting when only one group should be in the test. Use logic in the variant when everyone is in the test but the copy should fit where they came from.
After the test
A test that wins for some visitors and loses for others doesn't have to be all or nothing. When you deploy, Dalton can keep the change only for the device, channel or country where it won. See Deploy to a segment.
Variants built in your theme
If your developer prefers to build in the theme, they can put both versions in Liquid and let Dalton decide who sees which. The Dalton variant only flips a switch, for example an attribute your theme's CSS reacts to:
js
document.documentElement.setAttribute('data-ab-header', 'b')Make the original the version that shows when the attribute isn't set, so visitors see it if the script doesn't run. Dalton still handles targeting, the split, keeping visitors in their variant and the results.
Ask us
Targeting on a cookie, a logged-in state or other data from your own site is possible with a custom setup, where your site passes that information to Dalton. Email team@getdalton.com.