# Costs

What an install costs to run, and the limits that cap it.

GOAT runs in your Google Cloud project, so what it uses is charged by Google Cloud on your bill. This page covers what drives those charges, and the limits that cap them.

## What you pay for

| Charge | What drives it |
|---|---|
| Cloud Run | GOAT's service. One instance always runs, with 1 CPU and 512 MiB of memory, so the install costs about one instance's always-on rate even when idle. Busy times add instances, up to 10 by default. |
| BigQuery queries | Every chart that is not answered from a cache runs one query. You pay for the data each query reads. |
| BigQuery storage | GOAT's managed copies of your data, and its record of each query. Live connect datasets add no storage, since GOAT reads your tables where they are. |
| Cloud Storage | The files operators upload. They stay for as long as their datasets use them. |
| Firestore, Identity Platform, Secret Manager, and logging | Small, for what GOAT keeps and logs. |

The cost of a chart depends on how much of the table it reads. A table partitioned or clustered on the columns charts filter by costs less to query. See [How queries run](/under-the-hood/how-queries-run).

## The cost cap per query

Every query GOAT runs has a cap on the data it may read. A query that would read more fails instead of billing, and its chart says so. Previewing a dataset's rows and reading the rows behind a chart are capped the same way; there, GOAT checks the query before running it.

| Level | Default | Where it is set |
|---|---|---|
| The install | 10 GiB per query | When GOAT is installed. |
| One dataset | The install's default | The dataset's Settings tab, under Cost cap per query: Default, Custom (Cap at), or No cap. |

With No cap, a query on that dataset may read the whole table, and bills whatever it reads.

When a chart stops at the cap, narrow it: add a filter, especially on a date or partitioned column, or choose fewer fields. Or raise that dataset's cap.

## Reusing answers

GOAT keeps a chart's answer for a while and reuses it for the next viewer, instead of querying again. For Live connect tables, views, and files read in place, a longer reuse time reads the table less often and costs less; a shorter one shows new data sooner. Set it per dataset, under Cache answers for when you create it, or Reuse answers for on its Settings tab. Managed copies reuse answers until their next load, whatever the setting.

## Capping the whole bill

The per-query cap limits one query, not a day's total. To cap or watch total spending, use Google Cloud's own controls on the install's project:

- A budget with alerts, to be told when spending passes an amount.
- A BigQuery custom quota on query usage per day, to stop queries past a daily total.

If query costs spike, lower the cost cap on the datasets behind them, on each dataset's Settings tab.

## Common problems

| What you see | What to check |
|---|---|
| A chart says it would read more data than allowed | It hit its dataset's cost cap. Add a filter or choose fewer fields, or raise the cap. |
| "Previewing this dataset would scan more than its cost cap allows." | The same cap applies to the Data tab's preview. |
| The bill grows with no one using GOAT | The always-on instance, storage, and logging continue while idle. That part is steady. |
