What the feedback follow-up router costs
The model is used once per response with a comment, for theme classification and the serious-content check. Two hundred and fifty responses a month is a business surveying a few thousand orders a year. Here is where each cent goes.
Key takeaways
- About $2 a month at 250 responses. Roughly $4 at 1,000 responses.
- One Bedrock read per response is the only line that scales. Everything else is rounding.
- Nothing is always-on, so a quiet month genuinely costs almost nothing.
- Responses with no comment skip the model entirely, and they are a large fraction of the total.
- The duplicate test runs before the read, so resends are free.
- The three real risks: a retry loop, storage nobody expires, and a bigger model than the job needs.
The bill at three volumes
These are US East prices at the time of writing, at three volumes that bracket most small businesses. Find the bar closest to your own and read across.
Line by line
| Line | At 250 responses | How it scales |
|---|---|---|
| Bedrock read | $0.23 | Linear. One call per response, roughly 1,800 in and 200 out tokens. |
| SES | $0.18 | Linear. About 2.4 messages per response. |
| DynamoDB + S3 | $0.32 | Storage grows with what you retain, not with throughput. |
| Lambda + SQS | $0.12 | Linear, and effectively free at this scale. |
| CloudWatch | $0.16 | Flat, if you set retention. Unbounded if you do not. |
| Secrets Manager | $0.40 | Flat. One secret, $0.40 a month. |
| AWS Budgets | $0.46 | Flat. Two actions, so you find out before the bill does. |
One model call per response that has a comment, doing two things at once: theme classification and the serious-content check. Scores with no comment cost nothing.
The three ways this bill surprises you
Every one of these has happened to somebody, and all three are cheap to prevent.
- Surveying every order with no frequency cap. Not a cost problem so much as a list-destruction problem: a weekly customer surveyed weekly stops opening your email entirely.
- Re-classifying on every report run. Themes are assigned once per response and stored. Re-running classification at report time multiplies the cost and makes trends unstable.
- Log retention left at never. Survey comments are customer content, and unbounded logs of them are both a cost and a data question.
What it costs when nothing happens
This matters more than the headline number for a seasonal business. In a month with nothing to process the bill is the fixed band: Secrets Manager at forty cents, AWS Budgets at forty-six, and a few cents of storage. Call it a dollar. There is no instance to stop and nothing to remember to turn off.
- Management
- Analytics
- Front-end & mobile
Set these on day one
- A dead-letter queue on every SQS queue, with a maximum receive count of three.
- Thirty-day retention on every CloudWatch log group. There is no default that is safe.
- An S3 lifecycle rule on the object prefix, tiering at 90 days and expiring at your actual record-keeping horizon.
- Two AWS Budgets actions — one that emails at half your expected spend, one at double it. The second is how you find out about a loop in an hour instead of a month.
- Provisioned concurrency: none. Nothing here is latency-sensitive enough to justify paying for a warm function.
Next: the same system drawn for engineers — service names, resource identifiers, IAM scopes, table schemas and the model id.
All posts