---
title: "Choose a Cloud Monitoring aggregation and missing-data rule | Proofline Help"
description: "In the deploy verification signal editor, match the statistic to the metric’s kind. A rate, a count of returned points and a latency distribution describe different things."
url: https://proofline.sh/help/gcp-monitoring-aggregation/
---

[Help center](/help/)

# Choose a Cloud Monitoring aggregation and missing-data rule

In the deploy verification signal editor, match the statistic to the metric’s kind. A rate, a count of returned points and a latency distribution describe different things.

## Check the statistic and how empty reads behave

Each check reduces every series the labels select, such as one for each revision, to one value. **Mean**, **Maximum** and **Minimum** are compared as read. **Sum per second** and **Data points per second** add the series together, and because a sum or count grows with the window, each check divides it by the window's seconds: enter **Minimum change** per second for these statistics. **Data points per second** counts the metric's data points, not the requests or errors they record. `run.googleapis.com/request_count` writes one point a minute for each series it selects, so a minute with 1,000 errors reads like a minute with one. For a request or error total, choose **Sum per second**. Saving does not check that the statistic suits the metric. One that cannot apply, such as the minimum of a latency distribution like `run.googleapis.com/request_latencies`, leaves every check incomplete, and the Inbox offers **Edit signals**.

Select **No data means zero** only for a counter that reports nothing while nothing happens, such as 5xx responses. Otherwise a read that returns no data leaves the check incomplete, and the Inbox offers **Edit signals**.

## Related guides

-   [Connect Google Cloud](/help/connect-google-cloud/)
-   [Prepare Terraform changes](/help/terraform-changes/)
