Alerts

Alerts are rules evaluated after every data update. Each alert carries the facts that triggered it — values, baseline, threshold and window.

Rule types

Rule	Fires when	Plans
Monthly budget	Month-to-date spend reaches your warning share, or the month-end pace is ≥ 110% of budget; critical at 100%	All
Spend anomaly	A complete day is above its 28-day median by your threshold, by at least a minimum amount, and outside normal variation	Starter, Pro
Margin floor	Trailing gross margin falls below your floor (with a minimum revenue)	All
Sync failure	A connection fails 3 syncs in a row (critical after 10 or 48 hours)	All, built in
Re-authentication	A provider rejects stored credentials	All, built in

How anomalies are detected

The baseline is the median of up to 28 previous complete days, with the median absolute deviation (MAD) as the measure of normal variation. A day is anomalous when it exceeds the baseline by your percentage threshold and your minimum amount and a robust z-score of 4, 3 or 2.5 for low, normal or high sensitivity. At least 14 complete days are needed before anomaly alerts start.

Dedupe and auto-resolve

There is at most one open alert per rule and scope. Repeated detections update the existing alert; you are only notified again if severity escalates. Alerts resolve themselves when the condition clears — for example when a provider revises a day's figure, after two normal days, or when a budget month ends.

Delivery

New and escalated alerts appear in the inbox and the notification bell. Email goes to owners and admins by default (configurable per rule), combined into one message when several fire together, and capped at 10 per workspace per day. Every email has a one-click link to change your preferences.