How it works

Alerts watch your accounts for unusual swings in spend, conversions and ROAS, and flag the days that fall outside each metric’s normal range — across every connected Source. You get them in-app on the Alerts page and as an email digest.

Like the rest of TheAdsHub, alerts are advisory: they point you at what looks off so you can investigate — nothing is changed on the platforms.

How it decides

For each metric, TheAdsHub compares every day against that metric’s own recent behaviour using a z-score:

1. It builds a rolling 30-day baseline for the metric (its mean and standard deviation over the window).
2. For each day it computes z = (observed − mean) / standard deviation — how many standard deviations the day sits from normal.
3. A day is flagged when |z| ≥ 2.5.

So an alert means “this day is far enough outside the recent norm to be worth a look”, not that a fixed limit was crossed. Detection runs automatically every few hours over data up to yesterday (the latest settled day).

Severity

The bigger the deviation, the higher the severity:

Severity	Deviation
High
Medium
Low

What’s monitored

Alerts are evaluated at two levels, for each Source:

• Portfolio — the Source’s daily totals (e.g. all of Google, all of Meta).
• Campaign — each individual campaign.

Both cover spend, conversions and ROAS. To stay signal-rich, campaign-level checks ignore campaigns that spent very little over the window, and the strongest anomalies per organisation are kept so the list never floods.

Where alerts show up

• In-app — the Alerts page always shows the current snapshot of flagged days, filterable by Source.
• Email digest — each newly detected anomaly is emailed once (de-duplicated), grouped per organisation, with campaign names resolved. Notification settings on the Alerts page control email delivery.

Things to keep in mind

Note

A metric needs enough history (at least a few days in the window) and some day-to-day variation for a baseline to exist. Brand-new accounts or perfectly flat metrics won’t produce alerts until there’s something to compare against.

• It’s relative, not a target — a stable account has a tight baseline, so even a modest change can flag; a volatile account has a wide baseline and needs a bigger swing. That’s by design.
• Direction — both spikes and drops are flagged (a conversions drop or a ROAS collapse is as important as a spend spike). Open the alert to see observed vs. expected.
• Recent days settle — detection uses data up to yesterday; very fresh conversions can still be attributed, so revisit anything borderline once the day has settled.