Anomaly detection

Meta Ads anomaly detection that
catches drifts in 2 hours,
not 2 days.

crumplz's Anomaly Engine runs a full check on your Meta Ads ROAS, CPA, CTR and MER every 2 hours. Dynamic thresholds. Shopify cross-validation. One actionable alert when something drifts. No false positives. No dashboard to check.

Start free — 3 min setup See the engine

Dynamic thresholds · Shopify MER cross-validation · WhatsApp, Slack, Email

Checks every 2 hours Dynamic per-campaign baselines 4 metrics monitored −19% wasted ad spend
What anomaly detection means

Not alerts on numbers.
Alerts on deviations.

The difference between "alert when CPA > €40" and actually detecting an anomaly is the difference between static rules and real intelligence.

Static threshold alerting

  • Fixed value you set manually (e.g. "alert when CPA > €40")
  • Fires on normal seasonality — Mondays, post-weekend dips
  • Misses slow degradation that never crosses the threshold
  • No cross-validation — alerts even when Shopify revenue is fine
  • Requires manual updates when campaign performance shifts

crumplz Anomaly Engine

  • Dynamic baseline per campaign from historical performance
  • Accounts for day-of-week patterns — no false Monday alerts
  • Detects both velocity (sudden drop) and trend (slow bleed)
  • Cross-validated with Shopify MER before alert fires
  • Self-adjusting — learns as your campaign pattern evolves
What crumplz monitors

Four metrics. One engine.
Full campaign visibility.

ROAS
Return on Ad Spend
Tracks revenue generated per euro spent on Meta Ads
CPA
Cost per Acquisition
Monitors cost per conversion across campaigns and ad sets
CTR
Click-Through Rate
Detects creative fatigue and audience exhaustion signals
MER
Marketing Efficiency Ratio
Cross-validates Meta signals with real Shopify revenue and AOV

Monitoring these four together catches patterns that individual metric alerts miss. A ROAS drop with stable CTR points to landing page or checkout issues. A CTR drop with stable CPA points to creative fatigue before it hits profitability.

How the engine works

Simple algorithms.
Robust signals.

No black box. crumplz uses transparent, explainable anomaly detection — not a model you can't inspect.

Rolling baseline per campaign

For each campaign, crumplz builds a rolling 14-day baseline for each metric, segmented by day of week. Monday's ROAS baseline is different from Friday's. This eliminates the most common false positives from weekly campaign rhythms.

Drift velocity + trend detection

crumplz detects two types of anomaly: sudden drift (large deviation within a single 2-hour window) and trend degradation (consistent directional movement over multiple windows). Both trigger alerts at different severity levels.

Shopify MER confirmation gate

Before an alert fires, crumplz cross-validates the Meta signal against live Shopify MER data. If Meta shows a ROAS drop but Shopify revenue is holding, the alert is held — likely an attribution lag, not a real drift.

Recommended action per pattern

Different anomaly patterns map to different recommended actions. Creative fatigue → rotate creatives. CPM spike → check competitor activity or budget pacing. CVR drop → check landing page. Every alert is contextual, not generic.

Pricing

Anomaly Engine in every plan.

Starter

€79

/ month

  • Full Anomaly Engine (ROAS, CPA, CTR, MER)
  • 1 ad account
  • WhatsApp, Slack, Email
  • Check every 2h

Growth ⭐ Most popular

€149

/ month

  • Everything in Starter
  • 3 ad accounts
  • Custom anomaly thresholds
  • Priority support

Pro

€249

/ month

  • Everything in Growth
  • Multi-store
  • Roles & permissions
  • Budget automation (roadmap)
FAQ

About Meta Ads anomaly detection

How does crumplz avoid false positive alerts?
Three layers of false positive prevention: (1) Dynamic baselines segmented by day of week, so normal Monday patterns don't trigger alerts. (2) Multi-metric confirmation — a ROAS drop alone isn't enough; it needs to be corroborated by CPA or MER movement. (3) Shopify MER confirmation gate — if Shopify revenue is holding, the alert is held. The result is fewer alerts, higher signal quality.
What happens during a campaign's Meta learning phase?
crumplz automatically adjusts sensitivity during the learning phase. When a new campaign has limited historical data, dynamic thresholds are wider — so normal learning-phase volatility doesn't trigger alerts. Sensitivity narrows as the baseline builds over 7–14 days.
Can I customize anomaly sensitivity per campaign?
Yes on Growth and Pro plans. You can set campaign-specific sensitivity levels — for example, making a high-spend campaign more sensitive (alert on smaller deviations) or a test campaign less sensitive (only alert on major anomalies). Starter uses automatic sensitivity calibration.

Static rules miss drifts.
crumplz catches them.

Dynamic anomaly detection. Shopify cross-validation. Action in every alert.

Start free — no credit card

14-day trial · 3 min setup · GDPR · EU hosting