Thresholds vs. learned normal
A threshold alert fires when a number crosses a line you picked. That works when you know the right line, but accounts are noisy and what's 'normal' drifts over time, so static thresholds either miss real anomalies or cry wolf constantly.
Anomaly detection learns your baseline from your own history and judges new activity against it. A spike that's normal for you stays quiet; a genuinely unusual move gets flagged, even one you'd never have written a rule for.
What it catches that you'd miss
The valuable anomalies are the unexpected ones: engagement falling off a format that used to work, a quiet surge from a post you underestimated, a break in your posting-to-reach relationship. Because the system isn't limited to rules you thought of, it surfaces things you didn't know to look for.
frequently asked
- How is anomaly detection different from a normal alert?
- A normal alert needs you to set a threshold. Anomaly detection learns what's normal for your account and flags deviations automatically, including ones you'd never have set a rule for.
- Won't it flag every little fluctuation?
- No, it judges against your own baseline, so routine ups and downs stay quiet and only genuine breaks from your pattern surface.
- Do I need to configure anything?
- Much less than with manual alerts. x-signal learns your account's normal rhythm from its history rather than asking you to define every rule.
- Does it work on a new account?
- It gets sharper as it learns your pattern, so the more history it has, the better it distinguishes normal from unusual.
Last updated June 5, 2026