Shopify Sidekick data extension

ABConvert in Shopify Sidekick lets you ask questions about your tests and results in plain English without leaving Shopify Admin.

“Which ABConvert tests are active right now?”

​

When to use it

Use Sidekick when you want to:

• Check which tests are active, paused, scheduled, or closed.
• Summarize one test’s results and data-quality warnings.
• Explore results by date, measurement view, product group, or audience dimension.
• Find active tests with a winner ready for review.
• Review recent changes to a test before you make a decision.

​

When NOT to use it

Use the ABConvert app instead when you need to:

• Create, edit, preview, start, pause, end, or archive a test. Sidekick access is read-only.
• Refresh results immediately or export order data. Open the test’s Analytics dashboard for those actions.

To manage tests from an external AI assistant, use the ABConvert Model Context Protocol integration.

​

How it works

Sidekick reads ABConvert data for the Shopify store you are currently viewing. It can combine several reads in one conversation, such as finding a test by name, checking its result, and then opening its Analytics page.

​

Setup

You do not need an API token or an MCP connection. ABConvert and Shopify Sidekick must be available for the same store. Try this prompt in a new Sidekick conversation:

“List my active ABConvert tests.”

​

What you can ask

The following prompts cover the main read-only workflows: Sidekick can identify a test by name or numeric test ID. Use the numeric ID when two tests have similar names.

​

Ask for a specific result scope

State the scope in the same prompt as your question. Sidekick supports the following result scopes: For example:

“For September 1 through September 10, show the Free shipping over $50 test results for visitors who experienced the test. Only include the Best sellers product group.”

If you do not name a measurement view, Sidekick follows the measurement basis configured for the test in Analytics. If you explicitly request visitors who experienced the test but that view is not supported, Sidekick uses assigned visitors and tells you about the fallback. See Measure results by for the difference between these views.

​

Break results down by audience

Ask Sidekick to break a result down by one or two dimensions. Available dimensions include:

• Country, market, device, and visitor type.
• Traffic channel, traffic platform, referral domain, and landing page.
• UTM source, medium, campaign, term, and content.
• Page URL, template, and click destination for engagement results.

For example:

“Break down last week’s results for the Fall landing page test by country and device.”

Sidekick returns the highest-traffic rows first and may shorten a large result. Ask a narrower question when you need a smaller audience or date range.

​

Read result answers safely

Use these checks before you act on a Sidekick answer:

1. Check the measurement view. By default, Sidekick follows the measurement basis configured for the test in Analytics. If you requested visitors who experienced the test, check whether Sidekick reported a fallback to assigned visitors. These views use different populations.
2. Read the warning list. A sample ratio mismatch, low sample size, or stale filter can make a result unsafe to use.

If you ask for profit metrics without configured product costs, Sidekick tells you that profit is unavailable. It does not use revenue as a substitute. See Revenue and profit metrics.

​

Use custom events as metrics or filters

You can ask Sidekick to show a custom event as a metric, such as its event rate for each test group. You can also filter results to visitors who did or did not trigger an eligible custom event. Custom events must be enabled for the store. The filter event must belong to the current store, use a supported event type, have been created before the test started, and have remained active throughout the test.

“Show the Sign up event rate for visitors who did not click Size guide.”

When you combine both, state which event you want to measure and which event you want to use as the filter.

​

Common mistakes

• Expecting Sidekick to change a test. Sidekick can read ABConvert data only. Open ABConvert or use MCP for supported test actions.
• Overlooking filters kept from earlier questions. In a longer conversation, Sidekick may apply the measurement view, product group, date range, or custom event filter from an earlier question. Before you decide to roll out a test group, confirm which filters and measurement view the answer uses.

​

FAQ