Flatline

YAML type: flatline | Category: Data quality

Flags periods where values are near zero or effectively constant for multiple days.

Description

Flatline detects stuck metrics -- broken feeds that keep sending zeros or the same number. Mode selects near-zero, constant, or both.

How it works

Flatline illustration

Flatline scans for runs of days that are near zero, nearly constant (low standard deviation), or both, according to mode. When the run lasts at least minDurationDays, it emits a Flatline anomaly.

Jitter tolerance allows tiny noise while still treating the series as stuck. This catches broken feeds that keep publishing zeros or the same placeholder value.

Setup

Choose mode (zero, constant, or zero_or_constant). Raise minDurationDays to ignore short quiet periods.

Configure detectors on the data view Analysis tab or in YAML. Validate with Data Preview before enabling production schedules.

Settings

Parameter	Default	Description
`minDurationDays`	`3`	Minimum run length to flag
`nearZeroThreshold`	`0.001`	Absolute near-zero cutoff
`maxStdDev`	`0.0001`	Max std-dev for "constant"
`mode`	`zero_or_constant`	Flatline kind -- see Allowed values
`jitterTolerancePct`	`0.5`	Allowed relative jitter for constant mode

Allowed values

`mode`

Value	Meaning
`zero`	Point counts when near zero (`\|v\| <= nearZeroThreshold`)
`constant`	Point counts when it matches the previous value within `jitterTolerancePct`; segment also needs `stdDev <= maxStdDev`
`zero_or_constant`	Near-zero or constant (default)

Shared series filters

Most detectors also accept optional series filters in params:

Parameter	Purpose
`maxGapAmount`	Skip series when gap amount is too high (`null` = no limit). Catalog create-default is often `5`.
`minValue`	Skip series whose average value is below this (`null` = no limit). Catalog create-default is often `1000`.
`maxValue`	Skip series whose average value is above this (optional; omit/`null` = no limit)
`useSpecialDates`	When `true`, suppress anomalies that overlap special dates

When a filter skips a series, Data Preview shows the reason.

Configuration example

analysis:
  detectors:
    - type: flatline
      params:
        minDurationDays: 3
        nearZeroThreshold: 0.001
        maxStdDev: 0.0001
        mode: zero_or_constant
        jitterTolerancePct: 0.5

Using the detector

Investigate whether the source system stalled, a filter emptied the series, or a process stopped publishing.

Use cases

Broken IoT or app telemetry
Zero-sales after a store system outage
Constant placeholder values left in ETL

Specifics

Related: zero_rate for elevated zero share over a lookback window.