Operate APIs

Anomaly Scoring

Name: Anomaly Scoring
Brand: AttenTech

Score every incoming record for anomalies and flag what needs a human look.

Get API AccessPrivate beta only

// OVERVIEW

How Anomaly Scoring works

Anomaly Scoring watches your operational data — transactions, logs, sensor readings, whatever streams through — and scores each record against learned normal behavior in real time. Records that deviate get flagged with a score and a plain-language reason, so review queues fill with signal instead of noise.

Real-time scoring

Every record is scored the moment it arrives, against a continuously updated baseline of normal behavior — no nightly batch jobs.

Explainable flags

Each flagged record ships with a human-readable reason, not just a number, so reviewers know exactly what tripped the score.

Tunable sensitivity

Thresholds adjust per data source and time window, so a noisy feed and a strict one don't share the same trigger point.

anomaly

attentechth.com/api/operate/anomaly-scoring

01// score a record for anomalies

02import { Anomaly } from '@attentech/sdk'

03

04const anomaly = new Anomaly({

05 source: 'payments-stream',

06 sensitivity: 'balanced',

07})

08

09const result = await anomaly.score({

10 record: transaction,

11})

12

13if (result.flag) {

14 const { score, flag, reason } = result

15 queueForReview({ score, flag, reason })

16}

// PROCESS

How Anomaly Scoring flags what actually needs review

Every record is scored the instant it arrives against a continuously updated baseline for its source, and anything that deviates comes back with a plain-language reason — so review queues fill with genuine outliers, not noise from a fixed rule.

Ingest

Score

Evaluate

Structured Output

Ready to integrate Anomaly Scoring?

Get API access in days, or bring us in to run the full workflow for you — sovereignty by design at every step.

Get API AccessPrivate beta only