Alerts
Site settings → Alerts is where you decide what interrupts you. It has alert rules, the places alerts go, a log of recent alerts, and ownership rules that assign issues to people.
Viewers can read it. Owners, admins and members can change it. The demo is read-only.
Alert rules
A rule says when to alert and where to send it. New alert rule opens the form: choose When, set the rule's numbers, and tick the destinations under Send to. Each rule can be turned on and off, and the list shows when it last fired.
| When | Alerts when | You set |
|---|---|---|
| A new issue appears | An issue SnowAnalytics has never seen shows up. | Optionally, pages such as /checkout, /cart to watch only those. |
| An issue comes back after a release | A resolved issue gets a new error. | Nothing. |
| Errors spike for the hour | The last hour's errors reach a multiple of the usual for that hour over the week before. | Errors reach (3 times by default, at least 1.5) and At least (20 errors by default). |
| Revenue at risk passes an amount | Today's revenue at risk, counted since midnight, passes an amount. | Passes, in your currency (500 by default). |
| A page loads slowly | A page's LCP at the 75th percentile over the last hour is slower than a threshold. | Slower than (4 seconds by default, 0.5 to 60) and At least (20 visits by default). |
| The site goes down | No page views for a while on a site that normally has traffic at that hour, and the home page fails two checks a minute apart. | For (2 minutes by default, 1 to 120). |
| A cron job misses its check-in | A cron monitor is missed or failing. | Nothing. |
| Someone sends feedback | A visitor sends a note through the feedback widget. Spam and empty notes stay quiet, and a note that says the same as an open one counts there. | Nothing. |
| AI agents hit broken pages (weekly) | A digest every Monday at 9:00, site time, of AI agents that hit pages answering 404 or 5xx. | Nothing. |
The Errors spike for the hour rule is also the Spike alerts switch under Site settings → Emails: turning one on or off turns the other. See Email reports.
How often it alerts
Alerts for the same issue are sent at most once an hour, and a spike rule alerts at most once an hour. This keeps one bad deploy from sending a hundred messages.
An alert carries the title, the people hit and the revenue at risk, the page, the release and a link to the issue. GitHub and Linear alerts also carry the fix prompt.
Quiet hours
Quiet hours hold alerts at night. Choose From and Until in half-hour steps, in your site's time zone. While they last, only The site goes down sends. Everything else waits and goes out as one message when they end. They are off until you turn them on.
Destinations
A destination is a place an alert goes. Destinations lists the six kinds with where each points, and not connected until you add one. Connecting sends a test message first, and a destination that fails stays disconnected and shows why.
| Destination | What you give SnowAnalytics |
|---|---|
| Slack | A channel name and an incoming webhook URL for it. |
| The members of your workspace who should get the mail. | |
| GitHub issues | A repository from your connected GitHub. Alerts open an issue with the fix prompt. |
| Linear | An API key and a team key such as ENG. Alerts create an issue. |
| Discord | A channel name and a webhook URL. |
| Webhook | A name and an endpoint URL. |
A webhook receives JSON by POST. SnowAnalytics signs each body: the x-snow-signature header is an HMAC-SHA256 of the body with your signing secret, which SnowAnalytics shows when you connect the webhook, so copy it then. Check the signature before trusting a request.
Secrets such as webhook URLs and tokens are stored encrypted. A GitHub issues destination added with a token, before the GitHub App, keeps working.
Recent alerts
Recent alerts logs the last seven days: what fired, when, and whether it was delivered. Hover a failed one for the reason.
Cron check-ins
The page has no cron card. A monitor is created by your job's first check-in that includes its schedule, sent with a server key; see Cron check-ins. After that, a check-in that doesn't come in time, or reports an error, fires A cron job misses its check-in like any other rule, and it shows in Recent alerts.
Who owns what
Who owns what assigns issues without anyone doing it by hand. An issue someone assigned stays theirs. Otherwise the first rule from the top that matches decides, and Everything else at the bottom is unassigned.
Each rule matches one of:
| Match | Pattern example |
|---|---|
| A file in the stack trace | src/checkout/** |
| The page the error happened on | /stores/* |
| The host of the failing request | api.kilnrose.com |
The owner is a person or Whoever is on call. Each row counts its open issues. Owners show on Issues and on each issue.