---
title: "Monitor your usage"
description: "Watch request volume, error rates, and latency for each of your applications."
url: "https://nov18-prod-sg.movespot.fyi/guides/monitoring"
image: "https://nov18-prod-sg.movespot.fyi/_og/d/c_Ocean.takumi,title_Monitor+your+usage,description_~V2F0Y2ggcmVxdWVzdCB2b2x1bWUsIGVycm9yIHJhdGVzLCBhbmQgbGF0ZW5jeSBmb3IgZWFjaCBvZiB5b3VyIGFwcGxpY2F0aW9ucy4,props_eyJ0aGVtZSI6eyJtb2RlIjoibGlnaHQiLCJjb2xvcnMiOnsicHJpbWFyeSI6IiNDQjUzMjQifX19,p_Ii9ndWlkZXMvbW9uaXRvcmluZyI,s_MF6VX3wV31-LnUX1.png"
---

# Monitor your usage

Watch request volume, error rates, and latency for each of your applications.

## [Watch three signals](#watch-three-signals)

Most integration problems show up in one of three numbers. Track all three per application, per API:

| Signal         | What it tells you                                                   | Where to find it                         |
| :------------- | :------------------------------------------------------------------ | :--------------------------------------- |
| Request volume | Sudden spikes or drops usually mean a deploy went sideways          | Your own metrics or gateway analytics    |
| Error rate     | A rising 4xx or 5xx share is the first sign of a broken integration | Your own metrics, grouped by status code |
| Latency        | Tail (p95, p99) latency matters more than the average               | Your own metrics, measured end to end    |

## [Log what you need to debug](#log-what-you-need-to-debug)

For every request, log the outcome and the identifier that ties it back to us:

```json
{
  "method": "POST",
  "path": "/v1/orders",
  "status": 422,
  "request_id": "req_8k3j2l",
  "duration_ms": 143
}
```

Include the `request_id` whenever you contact support. It lets the team find the exact request in seconds instead of guessing from timestamps.

## [Set alerts before you need them](#set-alerts-before-you-need-them)

Alert on the signals that predict trouble, not just on outages:

-   **Error rate** above a threshold you'd consider an incident, such as more than `1%` of `5xx` responses over five minutes.
-   **`429` responses** at all. Any rate limiting means your traffic pattern needs attention. See [Rate limits](https://nov18-prod-sg.movespot.fyi/guides/rate-limits).
-   **`401` spikes** right after a deploy. They usually mean a credential didn't make it into the new environment.
-   **Latency** on your `p95`, not your average. Averages hide the tail that users feel.

Retry blindly and you'll alert on yourself

Retries multiply load exactly when the API is struggling. Use exponential backoff with jitter, and only retry the statuses that are safe to retry. See [Errors](https://nov18-prod-sg.movespot.fyi/guides/errors).

## [Read rate limit headers continuously](#read-rate-limit-headers-continuously)

Every response includes `X-RateLimit-Remaining` and `X-RateLimit-Reset`. Emit them as metrics so you can see quota burn before you hit a `429`, and so a new deploy's traffic pattern is visible immediately.

## [Rehearse the failure path](#rehearse-the-failure-path)

A monitor nobody has responded to is a fire drill waiting to happen. Once per quarter:

1.  Force a `429` in a staging application by exceeding the rate limit.
2.  Confirm your backoff and retry behavior behaves as designed.
3.  Confirm the alert fired and the runbook pointed someone in the right direction.

## [Next steps](#next-steps)

### Rate limits

Read your remaining quota and back off before you hit `429`.

<https://nov18-prod-sg.movespot.fyi/guides/rate-limits>

### Errors

Understand error responses and decide which requests to retry.

<https://nov18-prod-sg.movespot.fyi/guides/errors>

### Security best practices

Rotate credentials and protect data in transit.

<https://nov18-prod-sg.movespot.fyi/guides/security>