As your Appcues account grows, it gets harder to know what's working and what isn't. You may have dozens of Flows, Pins, Checklists, Banners, and NPS surveys live at the same time — but which ones are actually driving engagement? Which have high issue rates silently hurting their reach? Which have been live for months with barely any completions?

Instead of clicking into each experience one at a time, you can ask an MCP-compatible AI assistant to pull performance data across your entire account, rank experiences by effectiveness, and flag the ones that need attention.

Where to run these prompts

These prompts work in two places:

• Captain AI — Appcues' built-in AI assistant, available directly in Studio. No setup required. Best for quick performance checks while you're reviewing an experience in Studio. Captain AI also supports step interaction analysis for both legacy and Flows 2.0.
• External MCP clients (Claude, ChatGPT, Cursor, or another MCP-compatible tool) connected to the Appcues MCP. See Connect the Appcues MCP for setup. Better for building visual dashboards, charts, and side-by-side comparison tables — external clients like Claude can generate graphs, format data into custom report layouts, and combine Appcues performance data with data from other tools.

Both options access the same underlying data. Captain AI is faster for quick lookups; external clients are stronger for visualization and cross-tool reporting.

Prerequisites

• Captain AI access in Studio, or an external MCP client connected to the Appcues MCP.
• At least one published experience with analytics data.

Get a performance overview across all experience types

Start with the big picture: what's live, how much is it being seen, and how is it performing?

Example prompts:

• "List all my published experiences — Flows, Pins, Embeds, Checklists, Banners, Launchpads, and NPS surveys — with their completion rates and user counts over the last 30 days."
• "Which of my live experiences have the highest completion rates? Which have the lowest? Rank them across all types."
• "Give me a performance summary of everything that's currently published. Flag anything with a completion rate below 10% or an issue rate above 20%."

What you can learn:

• High completion + high volume — These experiences are doing their job. Study what makes them effective (targeting, timing, content length) and replicate the pattern.
• High completion + low volume — The experience works, but barely anyone sees it. The targeting segment may be too narrow, or the page it targets gets low traffic.
• Low completion + high volume — Many users see it but don't finish. Investigate step drop-off for Flows, or check whether Checklists have items that users get stuck on.
• Experiences with zero recent activity — If a published experience hasn't been seen in weeks, it may be targeting a segment that no longer has qualifying users. These are candidates for cleanup or retargeting.

Find your top-performing experiences

Narrow in on what's working best so you can double down.

Example prompts:

• "Show me my top 10 best-performing Embeds ranked by completion rate over the last 30 days."
• "Which Checklists have the highest item completion rates? Are any Checklists fully completed by most users?"
• "Compare completion rates across my NPS surveys. Which one gets the most responses?"

What to look for:

• Patterns across top performers — Do your best Embeds share something in common? Design styles, specific copy, targeting to engaged segments? Use this to inform new content.
• Checklist item completion patterns — If Checklist items 1-3 have high completion but items 4-5 don't, the Checklist may be too long or the later items may be unclear.
• NPS response rates — Low NPS response rates may indicate bad timing (showing the survey before users have enough context to answer) rather than bad content.

Investigate step and item drop-off

When an experience has low completion, you need to know where users are leaving.

Example prompts:

• "Show me the step-by-step breakdown for the Flow 'Onboarding Tour'. Where's the biggest drop-off?"
• "For my Checklist 'Getting Started', how many users completed each item? Which item has the lowest completion rate?"
• "Which step in 'Feature Walkthrough' loses the most users? What percentage who start the Flow never reach Step 3?"

What to look for:

• Sharp drop after Step 1 — The experience may be interrupting users at the wrong time, or the opening step isn't compelling enough to continue.
• Drop-off at a Tooltip or Hotspot step — If the step is anchored to a UI element users can't find (or that loads asynchronously), the step has a high issue rate and users abandon the Flow.
• Checklist items that nobody completes — The item's completion condition may be misconfigured (wrong event name, unreachable page), or the task itself may be unclear to users.

Analyze element interactions within Flow steps

Beyond step-level completion, the MCP can break down what users are clicking on within each step — which buttons, links, and interactive elements get engagement. This works with both legacy Flows and Flows 2.0.

Example prompts:

• "For the Flow 'Onboarding Tour', show me a step-by-step breakdown of what users are clicking on in each step."
• "In Step 2 of 'Feature Walkthrough', how many users clicked the primary CTA button versus the dismiss link?"
• "Which interactive elements across all steps in 'Upgrade Prompt' get the most clicks? Show me the breakdown in a dashboard."

What you can learn:

• CTA effectiveness — If your primary button gets 80% of clicks and the secondary link gets 2%, the step design is working. If neither gets clicks and users dismiss, the content isn't compelling.
• Unintended click patterns — Users clicking a "Learn more" link instead of the "Get started" button may indicate they aren't ready for the action you're pushing. Adjust the step content or add an intermediate step.
• Step design optimization — Comparing element interactions across steps reveals which layouts and content patterns drive the most engagement, so you can apply those patterns to underperforming steps.

Spot experiences with issues

Experiences can have flagged issues — broken CSS selectors, missing elements, unreachable pages — that silently reduce performance. The MCP can surface these across your entire account at once.

Example prompts:

• "Are there any issues flagged across any of my published experiences? List them with the issue rate for each."
• "Which of my Flows have steps with issue rates above 50%? Those are probably broken."
• "Check my Pins and Tooltips for selector issues. Are any anchored to elements that aren't being found?"

What to look for:

• 100% issue rate on a step — The anchored element is never found on the page. Users either skip the step or abandon the experience. Update the CSS selector or switch to a non-anchored step type (Modal, Slideout).
• Issues on experiences you didn't know were broken — This is the real value of a cross-account audit. A Pin you set up months ago may be broken because the UI element it was anchored to was renamed in a product update.
• Patterns by experience type — If most of your issues are on Pins and Tooltips, your product's CSS selectors may be unstable. Consider using data-appcues attributes for more durable anchoring.

Tips for better results

• Run a weekly health check — Ask for a full account performance summary every week: "How did all my published experiences perform this week compared to last week? Flag anything that degraded."
• Audit by experience type — Don't just check Flows. Pins, Checklists, and Launchpad items can quietly break or underperform without anyone noticing. A monthly prompt like "List all published Pins with their issue rates" catches problems early.
• Combine with user data — If a Flow has poor completion, ask who's dropping off: "List users who started 'Onboarding Tour' but didn't complete it." You can upload that list as a segment for re-targeting.
• Use the data to deprecate — Performance data isn't just for optimization. Experiences with consistently zero engagement should be unpublished. Cleaning up stale content improves your account's signal-to-noise ratio.

Limitations

• Real-time data — Analytics data may have a brief delay before appearing in API results. If an experience was just published, wait a few minutes before pulling performance data.
• Read-only — The MCP shows performance data but can't edit experience content or targeting. Make changes in Appcues Studio.