Testing Your Campaign

Every ZipTier campaign has a Sandbox alongside its Live version. Use it to try the AI Assistant yourself, catch mistakes, and fix them before real prospects see the campaign. This page covers how to open the sandbox, what to test, how sandbox usage is metered, and when to run another pass.

What Is Sandbox Testing?

Creating a campaign gives you two independent versions of the AI Assistant:

• Sandbox. A private testing version for you and your team.
• Production (Live). The version your prospects see once the campaign is live.

Both share the same campaign configuration and the same knowledge base. Only the conversations are separate. Sandbox usage does not draw from your Live AI Conversation credits, and sandbox chats stay out of your production analytics so test data never skews your metrics.

Opening the Sandbox

The sandbox lives on the View Campaign page. Open a campaign from My Campaigns, then switch to the Sandbox Testing tab. You'll see the same set of distribution options as the Production tab: a direct link, a JavaScript embed tag, and a QR code.

Sandbox Testing Tools

Sandbox Direct Link

A full-page URL that opens the AI Assistant as a standalone chat. Share it with teammates, subject-matter experts, or reviewers. The URL carries a mode=sandbox parameter so every message routes to the sandbox environment.

JavaScript Embed Tag

A snippet you can paste into a staging or beta copy of your website. Use it to check that the widget sits correctly on the page, doesn't fight your site styling, and behaves the way you expect across screen sizes.

QR Code

Scan the sandbox QR code with your phone to check the mobile experience on a real device. Simulators tend to miss keyboard behavior and touch-target sizing.

Widget Builder

The Sandbox Testing tab includes a shortcut into the Widget Builder. Tune colors, position, launcher style, and open/close behavior there, then lock in your choices before you push the Production embed to your live site.

What to Test

Try the assistant the way a real prospect would use it. Type past the greeting, ask the questions you're worried about, and take the flows to the end. The checklists below are a starting point.

Content and knowledge base

• Ask the questions your top buyer personas are most likely to ask, and confirm each answer matches your source documents.
• Try ambiguous, incomplete, or off-topic questions. The assistant should acknowledge the gap gracefully instead of fabricating an answer.
• Focus on numbers, dates, and specific product details, where accuracy matters most.
• Confirm the assistant can pull from both uploaded documents and crawled web pages.
• Open the Resources drop-down in the chat header and make sure only the documents and web pages you set to visible are listed.

Assistant behavior

• Read the greeting message as a first-time visitor and check whether the tone matches your brand.
• Click each kick-off question and make sure the responses are still relevant.
• Push a conversation past 10–15 messages to see how context holds up.

Lead capture and meeting booking

• Ask an intent-heavy question (for example, “can I get a demo?”) and confirm the lead capture form triggers.
• Verify that required and optional fields match what your sales team needs downstream.
• If in-chat meeting booking is enabled, book a meeting end to end to confirm your Calendly (or equivalent) scheduler is wired correctly.
• Click your CTAs and confirm they land on the right destination pages.

Deployment and UX

• Embed the sandbox JS tag on a staging page and confirm the widget renders correctly in context.
• Scan the sandbox QR code and walk through the mobile experience.
• Test in Chrome, Safari, and Firefox, on both desktop and mobile.
• Open the widget in an incognito window to rule out browser cache issues.

Sandbox vs. Live

The sandbox and Live campaigns are functionally identical. A few things are handled separately so testing doesn't pollute your production numbers.

Sandbox Limits by Plan

Each plan includes a monthly allowance of sandbox conversations and sandbox chat messages. They're independent of your Live allowances and reset with each billing cycle.

For a full breakdown of what counts as a conversation and how messages are metered, see Subscription Plans.

When to Retest

Testing isn't only a pre-launch step. Retest in the sandbox whenever any of the following change:

• Documents or web pages in the knowledge base.
• The greeting message, kick-off questions, tone, CTAs, or contact information.
• The lead capture form: added, removed, or newly required fields.
• Scheduler or CRM integration settings, such as a new Calendly link or a different HubSpot pipeline.
• Questions surfaced by the Knowledge Base Gaps analytics panel. Re-ask them in sandbox after you add content, and confirm the assistant now answers correctly.

✅ Best Practices

• Test before every promotion. Content that was accurate last month may be stale after a knowledge base update.
• Check mobile on a real device via the QR code. Simulators miss keyboard and touch-target quirks.
• Embed on a staging page, not just the direct link. The widget behaves slightly differently as a standalone page versus embedded.
• Ask the boring questions too. Pricing, timelines, refund policies, and contact details are where accuracy matters most.
• Validate in sandbox first, then publish to Live. Edits flow to both once published, so it's cheap to check first.