Detectors

Detectors analyze loaded series data and produce anomalies. Configure them on each data view Analysis tab (YAML). Validate behavior in Data Preview before enabling production schedules.

How detectors are defined

Detectors live in the data view Analysis tab as a YAML list under detectors:. You can:

Edit the YAML directly in the editor
Insert a starter snippet with Add detector (uses catalog defaults, including shared series filters)
Define the same list inside the full data view YAML (analysis.detectors) — see Data Views

At least one detector is required to deploy a data view. Multiple detectors on one data view run independently and can cover different aspects (quality, period comparisons, structural breaks, and so on).

YAML shape

detectors:
  - type: dod
    params:
      percentualChange: 50.0
      mode: both
      maxGapAmount: 5
      minValue: 1000
      useSpecialDates: false
  - type: gap
    params:
      longerThan: 3
      shorterThan: 0
      maxGapAmount: 5
      minValue: 1000
      useSpecialDates: false

Field	Required	Description
`type`	Yes	Detector type key (for example `dod`, `gap`, `peer`). Must match a catalog type.
`params`	No	Detector-specific settings plus optional shared series filters. Omitted keys use runtime/catalog defaults.

In the full data view document, the same list sits under analysis::

analysis:
  fillGapMethod: zero
  ignoreLastNDays: 1
  scanInLastNDays: 1095
  detectors:
    - type: mom
      params:
        percentualChange: 40.0
        mode: both
        aggregateMode: sum
        useSpecialDates: true

Special-date rules (specialDates, recurringRules, specialPeriods) are configured on the Special Dates tab (or under analysis in the full YAML). They apply only when a detector sets useSpecialDates: true. See Special settings.

Where results go

After load + process jobs, anomalies appear on Home and in Filters. Use Data Preview to dry-run a detector on selected series and see skip reasons before production.

Shared series filters

Most detectors accept these optional parameters in params. They decide whether a series is evaluated at all (or whether matching anomalies are suppressed for special dates):

Parameter	Catalog default	Description
`maxGapAmount`	`5`	Skip the series when its gap amount (%) is above this. Use `null` / omit for no limit.
`minValue`	`1000`	Skip the series when its average value is below this. Use `null` / omit for no limit.
`maxValue`	omit / `null`	Skip the series when its average value is above this. Omit or `null` = no upper limit.
`useSpecialDates`	`false`	When `true`, suppress anomalies that overlap configured special dates (by impact direction).

Add detector inserts maxGapAmount, minValue, and useSpecialDates with the catalog defaults above. Tune or remove them per detector as needed.

When a filter skips a series, Data Preview lists the reason (gap too high, average below minValue, or above maxValue).

Shared concepts

Concept	Where	Description
Direction `mode`	Many detectors	Usually `above`, `below`, or `both` (quantile detectors: `above` / `below` only). Exact meanings are on each detector page.
`aggregateMode`	Period / target-style detectors	How weekly/monthly/quarterly values are built: `sum` or `average`.
Detector-specific params	Each type	Thresholds, windows, methods, and so on — documented on the detector page under Settings and Allowed values.

Each detector page lists Allowed values for enum-like parameters (method, mode, granularity, targetMode, and so on) with meanings aligned to the runtime code.

Baseline detectors

peer, mix, and cardinality maintain shared baseline tables per data view. After changing their settings, use Recalculate baseline tables (ad-hoc) on the data view Management tab (or wait for the next load/process path that rebuilds them).

How it works

Each detector page includes a conceptual illustration plus a short explanation of the detection logic.

Catalog

Data quality & freshness

Detector	Type	Summary
Gap	`gap`	Missing-data intervals between consecutive dates
Bounds	`bounds`	Values outside a hard `[min, max]` range
Stale	`stale`	Last point older than a freshness SLA
Flatline	`flatline`	Near-zero or stuck-constant runs
Zero Rate	`zero_rate`	Elevated (or reduced) share of near-zero days

Periodic comparisons

Detector	Type	Summary
Day over Day	`dod`	Today vs yesterday (% change)
Day Prior Year	`dopy`	Today vs same calendar day last year
Week over Week	`wow`	Current week vs previous week
Week Prior Year	`wpy`	Current week vs same week last year
Month over Month	`mom`	Current month vs previous month
Month Prior Year	`mpy`	Current month vs same month last year
Quarter over Quarter	`qoq`	Current quarter vs previous quarter

Quantile thresholds

Detector	Type	Summary
Quantile (daily)	`qt`	Daily points vs historical percentile
Quantile (weekly)	`qtw`	Weekly aggregates vs weekly percentile
Quantile (monthly)	`qtm`	Monthly aggregates vs monthly percentile

Adaptive & structural

Detector	Type	Summary
Spike	`spike`	Sudden isolated spikes or drops
Cluster Window	`cw`	Level shifts between sliding windows
Dynamic Threshold	`dt`	Adaptive baseline with optional seasonality
Volatility	`volatility`	Regime shifts in variability
Sustained Shift	`sustain`	Multi-day deviations from baseline
Seasonal Residual	`sr`	Trend + seasonality residual outliers
Changepoint	`changepoint`	Structural breaks (IID/SSA + preprocessing)

Business comparisons

Detector	Type	Summary
Target / Budget	`target`	Actuals vs explicit targets
Peer Group	`peer`	Series vs peer cohort
Mix / Share	`mix`	Child share of parent aggregate
Cardinality	`cardinality`	Unusual distinct counts in a group