Course NPS & Improvement Loop

The problem this kills

You ask learners "would you recommend this course?" and "what would make it better?" - and then the answers go to die. A spreadsheet fills up with hundreds of comments nobody reads twice. You can't say whether a course is getting better or worse. Good suggestions never become real changes, and when you do fix something, you have no proof it moved the number. The feedback loop is open at both ends.

This tool closes the loop. It collects the score and the comments, computes a real NPS over time (with honest response counts so a "90" from four people doesn't fool anyone), uses AI to group the free-text suggestions into themes, and - only after the course owner approves - turns those themes into a tracked improvement backlog you can drive to done. Then it shows you the NPS before and after each change.

What you'll build

A small, secure web app your L&D team logs into. It does five things:

• Collects feedback per course: the NPS-style recommend question (0-10) plus an open "what would make it better?" box.
• Computes NPS over time for each course, always shown next to the response count so you know when the number actually means something.
• Themes the suggestions with AI - clustering hundreds of comments into a handful of clear, recurring themes.
• Approves themes into a backlog - the course owner reviews the AI's themes and decides which become tracked improvement items. Nothing becomes an action item without a human saying yes.
• Tracks improvements to done - each item links back to the feedback that prompted it, and closing it needs the owner's sign-off. The course's NPS trend shows whether the change worked.

What's inside the Implementation Plan

The plan is a single file you paste into an AI coding agent (Claude Code), which then builds the app with you step by step - no prior coding needed.

• It starts by interviewing you about your business. Before writing a line of code, the plan has the agent ask about your courses, how you collect feedback today, your real field names and conventions, your typical and peak response volumes, who owns which course, and your approval rules. It reflects a short tailored spec back to you for a thumbs-up - so you get a tool shaped around how you actually run L&D, not a generic template.
• A clear data model for courses, surveys, responses, themes, and improvement items.
• The full NPS calculation (promoters, passives, detractors) with response-count guardrails so small samples are flagged, not trusted.
• AI theming of open feedback, with every theme traceable to the comments behind it.
• The owner approval gate and the improvement backlog with status tracking and sign-off-to-close.
• A "No API yet?" path: import courses and responses from a Google Sheet or CSV and export your scores and backlog as clean CSV - so it's fully buildable today, even with zero integration to your LMS.
• Copy-paste prompts for every step, plus a verification checklist.

The governance it includes (this is the point)

This isn't a toy survey app. The plan builds in the controls L&D and compliance care about:

• Login so only your team can use the tool.
• Row-level security so people only ever see their own organization's courses and feedback.
• A complete audit trail - who themed, who approved, who closed an item, and when.
• A hard human-in-the-loop gate - the AI drafts themes; the course owner reviews and approves which become improvement items, and closing an item requires owner sign-off. The AI never writes an action item on its own.
• Duplicate guards so the same learner response (learner + course + survey) can't be counted twice, and each improvement item has a stable ID.

Who it's for

L&D leads, training managers, and enablement teams who want to continuously improve their courses based on what learners actually say - and who are tired of feedback that never turns into action. If you can use a spreadsheet and describe how your courses work, you can build this.

You've got this - paste the first prompt and let the interview tailor it to your courses.