What the card fee auditor costs
The model runs once per statement, which is once per merchant account per month. A fourteen-page itemised PDF is the most expensive read in this series and it is still four cents. Nine accounts is a small group of sites or a franchisee. Here is where each cent goes.
Key takeaways
- About $2 a month at 9 accounts. Roughly $3 at 30 accounts.
- One Bedrock read per merchant account is the only line that scales. Everything else is rounding.
- Nothing is always-on, so a quiet month genuinely costs almost nothing.
- Transaction files are parsed without a model at all, so the cost scales with statements rather than with sales volume.
- 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 9 accounts | How it scales |
|---|---|---|
| Bedrock read | $0.38 | Linear. One call per merchant account, roughly 1,800 in and 200 out tokens. |
| SES | $0.01 | Linear. About 4 messages per merchant account. |
| DynamoDB + S3 | $0.31 | 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 call per statement against a capable model, because merchant statements are the worst documents in this entire series: multi-column, abbreviated, inconsistent month to month, and frequently a scan of a print. This is a place where paying for a better model is straightforwardly correct, and it is still under five cents per account per month.
The three ways this bill surprises you
Every one of these has happened to somebody, and all three are cheap to prevent.
- Sending transaction files through a model. They are CSV with a fixed header. Parse them; a model here would cost more than the fees you are auditing.
- Re-reading the statement to answer a new question. Extract once into fees and query the fees. The PDF is the source, not the working set.
- Storing card numbers because the export contains them. Drop the PAN at ingest. You need the card type and the entry mode, and nothing that puts you in scope.
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