Testing messaging in a sandbox

A sandbox lets you try messaging changes without customers seeing them, which matters because the widget is public.

A sandbox lets you try messaging changes without customers seeing them, which matters because the widget is public.

Why messaging in particular

Most Zendesk configuration is internal. The messaging widget is on your website, so a mistake is visible to every visitor immediately, and there is no draft state.

Set it up

1. Create or open a sandbox from Admin Center.
2. Configure messaging in the sandbox.
3. Put its widget on a test page, not your live site.
4. Work through the conversation as a customer would.
5. Apply what you learned to production.

What is worth testing there

• A new greeting or flow, where the wording is being reworked.
• An AI agent, before it talks to anyone real.
• Trigger changes, where getting them wrong means every visitor sees it.

What a sandbox will not tell you

Volume, and real customer language. A sandbox has the questions you thought of. The value is in checking that the mechanics work, not in predicting what people will say.

Configuration does not copy itself

What you built in the sandbox has to be recreated in production, unless you are on a plan with configuration deployment. Note what you changed as you go, because rebuilding from memory is where mistakes come back.

See also

• Creating a sandbox environment
• Deploying a configuration from one environment to another
• Testing an AI agent before you publish it