Reports
Anomaly Detection (Smart Alerts)
Last updated 19 May 2026
Anomaly Detection (Smart Alerts)
Anomaly Detection is a rule-based system that surfaces unusual patterns in your venue's data — cancellation spikes, revenue pace warnings, staff gaps, traffic surges. Up to 2 alerts render as Smart Alerts at the top of your Intelligence dashboard. Available on every plan; not Spark-charged.
Where to find them — Web: Intelligence → top of the page (Headlines zone) → Smart Alerts strip. Mobile: Intelligence tab → top of the screen.
TL;DR
- Four detector types: cancellation spike, revenue pace warning, staff gap, traffic surge.
- Surfaces up to 2 active alerts at a time at the top of Intelligence.
- Rule-based (no AI, no Sparks). Thresholds are fixed.
- Dismissible per-alert per-day; reappears next day if still firing.
- All plans. Mobile parity.
The 4 detector types
Cancellation spike (warning)
Fires when:
- Cancellations in the last 24 hours > 2× the 28-day daily average, AND
- Cancellations in the last 24 hours ≥ 2 absolute
Why: catches the situation where a particular day or two is unusually bad. The double threshold (relative ratio + absolute count) means a quiet venue with 0.1 cancellations/day average doesn't fire just because 1 cancellation happened.
Needs ≥2 weeks of history to compute meaningfully.
Revenue pace warning (warning)
Fires when:
- Projected next-7-day revenue (from confirmed + pending bookings) < 65% of the 4-week weekly average
Why: catches the situation where your forward book is unusually empty for the upcoming week. Gives you time to react — fill the gap with Smart Waitlist, a Concierge campaign, or a one-off Bulk Compose.
Needs ≥4 weeks of history.
Staff gap (warning)
Fires when:
- Any active staff member who historically books ≥4 times per 28 days has zero confirmed or pending bookings in the next 48 hours
Why: catches the case where a normally-busy stylist has empty time coming up — useful for proactive outreach or a personal client text.
Needs ≥2 weeks of history. Surfaces per staff member.
Traffic surge (opportunity)
Fires when:
- Today's storefront views > 2× the 7-day rolling average, AND
- Today's storefront views ≥ 5 absolute
Why: catches viral moments — a social post hits, a press mention, a referral wave. Useful for spinning up a flash deal or campaign while attention is hot.
Needs ≥7 days of storefront analytics data.
Confidence scaling
Anomaly detection scales confidence by how much history is available:
| Weeks of history | Confidence multiplier |
|---|---|
| 0 | 0.0 (no alerts) |
| 1 | 0.4 |
| 2 | 0.7 |
| 3+ | 1.0 |
Alerts with confidence below 0.4 are suppressed. This means brand-new venues see fewer alerts in the first couple of weeks — the noise floor is too high to detect real anomalies.
Sort order and limit
Active anomalies are sorted warnings first, then by descending confidence, then sliced to top 2. If you have 5 active anomalies, only the 2 highest-priority show. Dismissing one promotes the next-highest.
Dismissing alerts
Each alert has a small X to dismiss. Dismissal persists in your browser's localStorage keyed {venue}:{anomaly-type}:{date} — so the alert dismisses for the current day only.
If the underlying condition still fires tomorrow, the alert reappears. You're not muting it permanently; you're just acknowledging it for today.
When all anomalies are dismissed, the strip briefly shows "All caught up" for 2 seconds, then collapses.
How anomalies relate to the AI narrative
Anomaly Detection is purely rule-based — no LLM, no Sparks, no async generation. It runs on every Intelligence dashboard load against a 15-minute server cache.
The AI Narrative and Business Health narrative (both PRO) are AI-generated week-level interpretations. They might cover the same data ("Your bookings are down this week") but in different forms — Anomaly Detection is a tight numeric trigger; AI narrative is a paragraph of context.
Use them together: anomalies for immediate "look at this" prompts; narratives for the why and the suggested response.
Caching
- Server cache: 15 minutes per venue
- Mobile client cache: 5 minutes stale-time
- Dismissals: per-browser, per-day
Tier and cost
- All plans — anomaly detection is an intentional safety net, not a premium feature
- Not Spark-charged — rule-based, no AI calls
- All four detectors run for all plans
Mobile parity
Full parity. Mobile Intelligence tab shows the same Smart Alerts strip at the top with the same dismissal pattern. Tapping an alert routes to the relevant detail surface via the mobile-equivalent deep link (e.g. cancellation spike → calendar; staff gap → that staff member's schedule).
What's NOT detected
- Specific client churn — handled by Retention Intelligence card with health scores
- Single low-rating review — handled by Reviews & Feedback low-satisfaction queue
- SMS delivery issues — handled by the Activity Feed needs-action filter
- Wallet balance — surfaced as its own notification, not via anomalies
- Membership expiring — surfaced as its own notification
Anomalies cover trend changes specifically. Single-event situations have their own dedicated alerts elsewhere.
Common mistakes
| Problem | What to check |
|---|---|
| No alerts ever appear | Confirm you have at least 2 weeks of history. Newer venues won't see alerts until the noise floor stabilises. |
| Alert won't go away | Dismissals are per-day. If the condition is still firing tomorrow, the alert reappears — that's by design. Address the root cause or wait for the trend to normalise. |
| Want to mute a specific staff's gap alert permanently | Not supported. The system will surface it as long as the staff is active and historically busy. |
| Cancellation spike alert seems wrong | Check the calendar for that 24-hour window. The trigger needs both a high ratio AND ≥2 absolute cancellations. |
| Revenue pace warning fires every Monday | Some venues have weekly patterns the 4-week average doesn't fully smooth. If yours is one, treat the alert as informational. |
FAQ
What's the difference between Anomaly Detection and Smart Alerts?
Same thing — Anomaly Detection is the underlying capability; Smart Alerts is the surface where the anomalies appear. Up to 2 active alerts render at the top of the Intelligence dashboard on web and on the mobile Intelligence tab.
What kinds of anomalies are detected?
Four types: cancellation spikes (recent cancellations >2× the 28-day daily average), revenue pace warnings (next-7-day projection <65% of 4-week weekly average), staff gaps (a normally-busy staff member with zero bookings in the next 48h), and traffic surges (today's storefront views >2× the 7-day rolling average).
Can I tune the sensitivity?
Not today — thresholds are fixed. Cancellation spike triggers at 2× normal; revenue pace at 65% of weekly average; traffic surge at 2× rolling average. Confidence scaling does mean newer venues see fewer alerts until they have 2-3 weeks of history.
Do anomalies cost Sparks?
No. Anomaly detection is rule-based — no AI calls, no Spark consumption. It's intentionally a free safety net on every plan, not a premium feature.
Are anomalies pushed to me as notifications?
No. Anomalies surface only on the Intelligence dashboard, not as push notifications or in the Activity Feed. Specific high-priority events (booking cancelled, no-show, SMS delivery failed, low wallet) have their own notifications.
What's the difference between this and a 4-star drop in revenue I noticed manually?
Anomaly Detection sets a programmatic threshold (65% of weekly average) so you don't have to spot the drop yourself. Manual review of the Revenue card still works; anomalies just surface them automatically when the numeric trigger fires.