Feedback follow-up router
A small system that takes a satisfaction score and its comment, routes the unhappy ones to a person within the hour with the context already gathered, groups the rest into themes, and measures whether following up actually changed anything. It never replies on anybody’s behalf. Seven posts on the same system — one diagram at a time — with a cost breakdown and an engineering reference at the end.
- 01
A feedback follow-up router on AWS for a few dollars a month
The whole system on one page — a receiver, a router and a theme grouper — built on the observation that response time matters more than response wording.
- 02
How a response arrives and matches
Why survey links carry an order token rather than a customer id, how anonymous responses are handled, and the timing of the ask itself.
- 03
How urgency gets decided
Why the score alone is insufficient, the comment patterns that escalate regardless of score, who each band routes to, and the response that needs somebody senior.
- 04
How the context gets attached
The five pieces of context that make a fast reply a good one, why the delivery outcome matters most, and the history that changes the tone of what somebody writes.
- 05
How the signal stays honest
Grouping responses into fixed themes, why response rate governs whether the score means anything, the measurement that shows whether following up works, and the pressure to game it.
- 06
What the feedback follow-up router costs
About $2 a month. One Bedrock read per response is the only line that grows; the queue, the table, the mail and the storage are rounding errors. Plus the three ways the bill could surprise you.
- 07
Engineering reference: the feedback follow-up router architecture
Same system, drawn purely for engineers. Service names, region, Lambda inventory, IAM scopes, the schemas and the exact model id.
Frequently asked questions
- What is a feedback follow-up router?
- A small serverless system that routes each survey response by what it says: unhappy scores reach a person within the hour with the order history attached, and the rest are grouped into themes. It routes and gathers context; a person writes every reply.
- Why does speed matter so much?
- Because a complaint answered within an hour reads as attentiveness and the same complaint answered in four days reads as indifference. The content of the reply matters less than most people expect and the delay matters more.
- Does it write the reply?
- No. A generated apology to somebody who is already annoyed is worse than a short human one, and it is usually obvious. It gathers the context so a person can write three honest sentences in ninety seconds.
- What about people who leave a low score and no comment?
- They get the same follow-up, and it is the more valuable one: a score with no comment is a person who could not be bothered to explain, and asking one specific question converts a surprising number of them.
- What does it cost to run?
- A couple of dollars a month at a few hundred responses. See part six.