CSAT Feedback Collector & Low-Score Alert

The problem this kills

You only find out a customer was unhappy when it's already too late - a one-star review shows up online, or they tell three neighbors, or they simply never call you again. By then the tech is long gone, the job is closed, and there's nothing left to fix. The feedback you needed was sitting in that customer's head the day of the visit, and you had no clean way to ask for it or to act on it fast.

Most shops either don't survey at all, or they bolt on some heavyweight CX platform that nobody checks, or they email a generic "how did we do?" blast that lands in spam and gets ignored. And even when a survey does come back angry, it sits in an inbox until someone notices - so the one moment when a quick, genuine "I'm so sorry, let me make this right" call could have saved the relationship slips right past you.

There's also the coaching problem. When feedback isn't tied to the specific job and the specific tech, you can't see patterns - which crews delight customers, which ones keep generating "they left a mess" comments, which job types consistently disappoint. So the scores, even when you have them, never turn into better service.

This tool turns that around. After each job, the customer gets a short, friendly survey. The score and comment land against the work order and the tech. And the instant a low score comes in, the right manager gets an alert - so they can reach out and recover the relationship while the experience is still fresh, before it ever goes public.

What you'll build

A small internal web app, just for your team, that:

• Imports completed jobs from a CSV - customer, contact, work order number, assigned tech, job type, and completion date.
• Sends each customer a short 1-2 question survey through Resend - a rating plus an optional comment - with a one-click rating link so they don't have to log in or type much.
• Captures the score and comment and ties it to the work order and the tech, so every piece of feedback is traceable and coachable.
• Fires an instant low-score alert to the right manager the moment a score below your threshold comes in - email now, with the customer, job, tech, score, and their words.
• Holds any customer-facing recovery message behind a manager's approval - the manager reviews the flagged feedback, decides the recovery action, and approves the response before a single word goes back to the customer.
• Honors consent and frequency caps - no surveying customers who opted out, and never more than your allowed number of surveys in a window.
• Dedupes on work order + survey so the same job can never be surveyed twice.
• Surfaces a tech / crew scoreboard so you can coach from real data.
• Exports the whole survey + alert log as a CSV.

What's inside the Implementation Plan

The plan is a single markdown file you paste into Claude Code (a free AI coding agent). It walks the agent through building the whole tool, step by step, each step ending with a ready-to-paste prompt.

The most important part: the plan opens by interviewing you about your business. Before it writes a single line, the agent asks how you close out jobs today, where your completed-work data lives (a field-service app, a dispatch board, a spreadsheet), the exact column names in your export, your survey questions and rating scale, what score counts as "low," who should get the alert and how fast, what a recovery looks like in your shop, and your messiest edge cases - commercial jobs with no single contact, multi-visit jobs, customers with no email, repeat visits to the same address. It reads a short tailored spec back to you, you confirm it, and only then does it build - so you get a tool shaped to how your service operation actually runs, not a generic template you have to wrestle into place.

Inside you'll find:

• The discovery interview and how the agent turns your answers into the data model, your survey, and your low-score rule.
• The full build: database, login, CSV import with duplicate guards, the survey send, the rating-capture page, the low-score alert, the manager review-and-approve recovery screen, the tech scoreboard, and the export.
• The hard human approval gate, consent and frequency caps, and the dedupe on work order + survey.
• Verification steps so you can prove it works, and the CSV-export fallback so the tool is fully usable today even before you connect it to your field-service system.

The governance it includes (this is the point)

This isn't a toy. The plan builds in the controls a service business actually needs:

• Login so only your team can see or touch anything.
• Row-level security so people only ever see the jobs, customers, surveys, and scores that belong to your organization.
• A complete audit trail - every survey sent, score captured, alert fired, and recovery approved, logged with who and when.
• A hard human-in-the-loop gate - the tool can draft a recovery response, but a real manager reviews the flagged low score and approves the action and the wording before anything is sent to the customer.
• Consent and frequency caps so customers are never over-surveyed or messaged after opting out.
• Duplicate guards - dedupe on work order + survey so the same job can never be surveyed twice.

Who it's for

Service managers, customer-experience leads, dispatchers, and owners at field-service shops - HVAC, plumbing, electrical, pest control, garage doors, cleaning, appliance repair, landscaping - who want to catch an unhappy customer in the hours after a job, not weeks later in a public review, and who want feedback tied to the tech so it actually improves service. You don't need to write code. You need your completed-jobs CSV, your survey questions, and a weekend.

You've got this - paste the first prompt and let the agent interview you.