Get an alert when something suddenly gets worse

What Varly watches

Alert	What it means	On at the start
JavaScript errors	A larger share of visits than usual gets an error in the page code. Varly looks at all errors together and at each of the most common error messages.	Yes
Rage clicks	Many more visits than usual click the same element over and over, for example a button that stopped working.	Yes
Conversion rate and orders	Fewer visits than usual place an order, or orders drop. Needs your orders connected.	Yes
Funnel steps	Fewer visits than usual get from one step of a saved funnel to the next. Checked over the last 24 hours.	Yes
Page speed	A page takes clearly longer than usual to show its main content (Largest Contentful Paint, LCP).	No

Turn each one on or off under Settings → Alerts.

What an alert tells you

Each alert is about one problem, for example one error message or one button. It says:

• What changed and how much, such as JavaScript error “Cannot read properties of undefined”: 0.3% → 33% of visits, mostly on product pages.
• Since when, rounded to the hour, such as Since 14:00.
• What is at stake, when Varly can work it out from your own orders: About €1,940 less revenue than usual so far for a drop in orders, or If this keeps up: roughly €800–€1,400 a month at stake for errors and rage clicks, calculated the same way as on Connections. With too few orders for a fair estimate, Varly leaves it out.
• 3 recordings that show it, such as visits that got this error, with a link to all matching recordings. See Watch recordings.
• A link to look closer: the heatmap for rage clicks, the funnel, or speed and errors in Traffic & pages.

How Varly decides

The baseline is what is usual for your site: the same hours on the same weekday in the previous 4 weeks. Comparing Tuesday afternoon with earlier Tuesday afternoons means a busy weekend or a quiet night never looks like a problem. The hours follow your site's clock (its time zone under Settings), so in the week after the clocks change, 10:00 is still compared with 10:00.

Varly opens an alert only when all of these are true:

1. Enough visits. The site has at least 200 visits a day, and there are enough visits in the hours that changed (for example 100 visits for errors, 60 visits to the page for rage clicks).
2. A real number of visits affected. For example at least 10 visits with the error, or 8 with rage clicks on the element. For a drop: at least 12 orders would have been usual in those hours.
3. A clear change, both relative and absolute. For errors: at least twice as many as usual and at least 1 percentage point more visits. For orders: at least 30% fewer and at least 5 orders fewer.
4. Not chance. A statistical check that compares the counts now with the baseline says the difference is very unlikely to be chance (z-score of 4 or more; roughly 1 in 30,000).
5. Worse than every one of the earlier weeks, not just worse than their average.
6. It lasts. At least 3 hours, and the last hours don't look normal again.

An alert closes with Back to normal when the last 3 hours look like a usual day again, or when the change is no longer clear. If the same problem comes back within 24 hours, the same alert opens again, so you don't get a pile of alerts for one problem. Varly opens at most 5 new alerts per site per hour.

Small sites

A site with fewer than 200 visits a day doesn't get alerts. With so few visits, two or three visitors with an error can look like a big jump, and alerts would mostly be false alarms. Settings → Alerts tells you when this is the case. A new site also needs 2 weeks of data before Varly knows what a usual day looks like.

Where alerts show up

• Research → Connections, under Alerts: open alerts first, then the ones from the last 14 days that are back to normal.
• Overview: a short list when something is open.
• Email to the owners and editors of the site, when an alert opens and when it's back to normal. Turn it off under Settings → Alerts.
• Slack or a webhook: tick Something suddenly got worse for a destination under Settings → Notifications. See Get notifications in Slack, Zapier or Make.
• The weekly email: one line per site, such as 2 spike alerts this week, all back to normal.

Quiet hours

Under Settings → Alerts, tick Quiet hours and pick when they start and end, in the time zone of the site. In those hours Varly sends no emails or Slack messages. Alerts still show in Varly, and anything still open is sent when the quiet hours end.

Mute an alert

Is an alert about something you expect, such as a carousel where repeated clicks are normal? Click Mute this alert on the card (owners and editors). The alert closes and Varly won't alert about this problem again. To undo it, go to Settings → Alerts → Muted alerts and click Alert me again.

Common questions

Why did I get no alert while the error count went up? The jump was too small, too short or within what earlier weeks also showed. Varly would rather miss a small blip than send false alarms. You can always see all errors under Traffic & pages.

The alert says “Since 14:00” but it started a bit earlier. Varly checks per hour and rounds to the hour.

What happens when the clocks change? Nothing you need to do. Varly compares by the hour on your site's clock, not by exactly 7 × 24 hours, so daylight saving time doesn't shift the usual pattern by an hour. On the night of the change itself, one hour is skipped or happens twice, which is too small to cause an alert.

Does an alert use AI? No. Alerts are Varly's own counts and a fixed statistical rule, the same for every site.