These cookies are required for the website to function properly. They ensure security, enable basic features like page navigation, and store user session data. You cannot disable these cookies.
- IoT Data Analytics
- Features
- Anomaly review
Anomaly review
Examine measured values against bounds and keep the review tied to its source revision.
OverviewStreams ready for anomaly review×
IoT Data Analytics · Anomaly review
New
Streams ready for anomaly review
Analyzed waits for the operations analyst
| Number | Name | Status | Reading value | Lower bound | Upper bound | |
|---|---|---|---|---|---|---|
| SS-43757 | Stream 692 | Validated | 440 | 200 | 500 | |
| SS-72941 | Stream 206 | Validated | 560 | 200 | 680 | |
| SS-54286 | Stream 421 | Validated | 950 | 190 | 590 | |
| SS-27330 | Stream 151 | Validated | 400 | 580 | 780 | |
| SS-42260 | Stream 881 | Validated | 930 | 480 | 670 |
6 of 6 records
Stream 692Sensor streams · SS-43757
ValidatedReading value440
Lower bound200
Upper bound500
Quality flagGood
What you can do
Review an anomaly
The Operations Analyst opens a validated stream and compares its reading value with the effective lower and upper bounds. The quality flag and anomaly score help explain the recorded outcome. The anomaly workflow persists a run and numbered steps, then reads the stream back before calling the run successful.
How the work is organized
- Value. The Operations Analyst compares the stream reading with its lower and upper bounds.
- Effect. The run and numbered steps show what changed and whether read-back confirmed it.
- Decision. The Independent Reviewer records a decision against the reviewed revision when required.
A returned decision gives the Stream Owner a reason to inspect or correct the stream. The reviewer can see the source revision and evidence reference. The owner can then create a successor revision, preserving the original run and judgement for later comparison.
What to agree during setup
Agree who sets bounds and how an uncertain or bad quality flag is handled. Define when a review decision is required and who may make it. A decision must identify the exact stream revision so a later correction cannot silently change the subject of an earlier approval.
Use a normal reading and one beyond the upper bound. The Operations Analyst checks the stream value, quality flag and anomaly score, then compares the saved fields with the run result. The Independent Reviewer uses the exact revision and evidence reference when a decision is needed.
Check the handoff on an actual case
Suppose a reading lies above the upper bound. The analyst checks the parent stream, observation date and collection basis, reviews the bound in force, and records the outcome. The read-back confirms the stream fields actually changed. The Independent Reviewer checks supporting evidence and records an approved, returned or rejected decision for that revision when approval is required.
If read-back cannot confirm a changed stream, the Operations Analyst checks the run steps before trying again. The Independent Reviewer should see the actual stream fields and the revision under review; a successful write attempt alone is not the final outcome.
Before the decision, compare the recorded stream status with the run status. If they disagree, the analyst investigates the step trail and the current stream fields. The reviewer should see the corrected explanation before deciding on that revision. A visible anomaly score alone does not settle the source decision.
What your team sees
- Streams ready for anomaly review — Validated streams with observed reading values and bounds.
- Pending review decisions — Source revisions awaiting a reviewer decision.
Follow the work through
See the related process for the handoff and the data model for the connected records.
What it keeps track of
Data model- Sensor streams
- Measured web properties with lifecycle status and observed outcome fields.
- Sensor readings
- Consented observation-level session and conversion event facts.
- Operation runs
- Workflow execution runs with frozen revision, input hash and result identity.
- Operation steps
- Numbered steps of an operation run with input/output and receipt state.
- Review decisions
- Independent reviewer decisions recorded against an exact source revision.
Who uses it
Permissions- Operations Analyst
- , Independent Reviewer, Stream Owner.
Set up IoT Data Analytics your way.
Start free, then use Proto to change fields, steps and approvals before your team goes live.