Dynamic and Time-Series

The Dynamic and Time-Series extension models systems whose state changes continuously: how data moves between systems, metrics measured as time series, and models that predict one metric from others. It extends the Performance domain and the interoperability extensions with three record types:

• Data flow: a configured movement of data from source systems to target systems, with its trigger, cadence, payload schema, quality measure and governance. A data flow can be a live operational feed, or a monitored source such as a regulator's publication page watched daily for new instruments.
• Timebound metric: a metric measured as a time series, with its unit, update interval, target and alert thresholds.
• Predictive model: a model that predicts a timebound metric from others. Its governance (ownership, validation, monitoring) lives in its Model Risk Management model profile.

Schemas: data-flow.schema.json, timebound-metric.schema.json, predictive-model.schema.json
Specification: JSON Schema Draft-07

Monitored sources

When a data flow feeds governed reasoning, such as an agent that watches regulators for change, sourcePrecedence records how much weight the source carries when sources disagree, and whether a change from it alone may set a record's status or date. freshnessThreshold sets how long records fed by the flow stay fresh without reverification, and governance.humanApprovalRequired records whether changes need a named approver. New instruments the flow finds are recorded as Triggers.

Data Flow

Attribute	Type	Description	Required
dataFlowID	String	Unique identifier for the data flow, e.g. DF-PROC-001	✓
title	String	Name of the data flow	✓
description	String	What the data flow carries and why
sourceSystems	Array[Object]	Systems or sources the data comes from	✓
sourceSystems[].endpointType	Enum	Kind of endpoint: software, sensor, dataStore, webPage, feed, register, api, person, other (required)
sourceSystems[].title	String	Name of the endpoint (required)
sourceSystems[].systemReference	Object	Technology or Information record for the endpoint, when one exists
sourceSystems[].url	String	Address of the endpoint, for web sources and APIs
targetSystems	Array[Object]	Systems that receive the data
targetSystems[].endpointType	Enum	Kind of endpoint: software, sensor, dataStore, webPage, feed, register, api, person, other (required)
targetSystems[].title	String	Name of the endpoint (required)
targetSystems[].systemReference	Object	Technology or Information record for the endpoint, when one exists
targetSystems[].url	String	Address of the endpoint, for web sources and APIs
frequency	Enum	Cadence of the data flow: continuous, realTime, hourly, daily, weekly, monthly, quarterly, semiAnnually, annually, eventDriven, onDemand, adHoc	✓
triggeringEvent	Object	Event that starts the flow, for event-driven flows
triggeringEvent.eventName	String	Name of the event, e.g. PartProduced (required)
triggeringEvent.source	String	Where the event comes from, e.g. Machine101
schemaReference	Object	Schema of the data payload
schemaReference.format	Enum	Payload format: json, xml, csv, html, pdf, rss, atom, binary, other (required)
schemaReference.schemaID	String	Identifier or URI of the payload schema, e.g. SCHEMA-MFG002 or a Triodian schema $id
sourcePrecedence	Object	How much weight the source carries when sources disagree. Used where a flow feeds governed reasoning
sourcePrecedence.rank	Integer	1 (highest precedence) to 5 (lowest) (required)
sourcePrecedence.basis	Enum	Why the source has that rank: officialText, regulatorPublication, standardsBodyPublication, commentary, news, internalSystem, sensor, other (required)
sourcePrecedence.canSetStatusOrDate	Boolean	Whether a change from this source alone may set a record's status or date, or only serve as a lead to a higher-ranked source
externalOrgID	String	External Organization that owns the source, for external sources
jurisdiction	String	Jurisdiction the source covers, for regulatory sources
dataQualityMetricID	String	Performance or timebound metric that measures the flow's data quality, latency or freshness
latency	String	Expected latency of the flow, as an ISO 8601 duration or plain text, e.g. PT0.1S or 100ms
volume	String	Expected volume, e.g. 1000 records/min
freshnessThreshold	String	Maximum age, as an ISO 8601 duration, after which records fed by this flow are treated as stale unless re-verified, e.g. P90D
governance	Object	Governance controls applied to the flow
governance.ownerTitle	String	Person or team that owns the flow
governance.ownerStakeholderID	String	Stakeholder record for the owner
governance.approvedDate	Date	Date the flow was approved
governance.reviewFrequency	Enum	How often the flow is reviewed: continuous, realTime, hourly, daily, weekly, monthly, quarterly, semiAnnually, annually, eventDriven, onDemand, adHoc
governance.humanApprovalRequired	Boolean	Whether changes derived from this flow need a named human approver before they are published
status	Enum	Implementation state of the flow: notStarted, planned, inProgress, implemented, operational, verified, retiring, retired, abandoned, notApplicable

{
  "dataFlowID": "DF-PROC-001",
  "title": "Live Manufacturing Process Data",
  "description": "Real-time ingestion of data from a manufacturing process line.",
  "sourceSystems": [
    {
      "endpointType": "sensor",
      "title": "Sensor Network A"
    }
  ],
  "targetSystems": [
    {
      "endpointType": "software",
      "title": "Operational Dashboard"
    },
    {
      "endpointType": "software",
      "title": "Quality Control Engine"
    }
  ],
  "frequency": "realTime",
  "triggeringEvent": {
    "eventName": "PartProduced",
    "source": "Machine101"
  },
  "schemaReference": {
    "format": "json",
    "schemaID": "SCHEMA-MFG002"
  },
  "dataQualityMetricID": "MET-LATENCY-001",
  "latency": "100ms",
  "governance": {
    "ownerTitle": "DataOps Team",
    "approvedDate": "2025-08-20"
  }
}

Timebound Metric

Attribute	Type	Description	Required
timeboundMetricID	String	Unique identifier for the metric, e.g. TBM-MACHINE-TEMPERATURE	✓
title	String	Name of the metric	✓
description	String	What the metric measures
source	String	Sensor, stream or data flow the values come from
dataFlowID	String	Data flow that supplies the values
unitOfMeasure	String	Unit the metric is measured in, e.g. Celsius	✓
updateFrequency	Enum	How often the metric updates: continuous, realTime, hourly, daily, weekly, monthly, quarterly, semiAnnually, annually, eventDriven, onDemand, adHoc	✓
updateInterval	String	Exact update interval as an ISO 8601 duration, e.g. PT10S
target	Object	Desired or expected value
target.value	Number	Target value (required)
target.comparison	Enum	How a reading is compared with the target: atMost, atLeast, equals, within
target.tolerance	Number	Allowed deviation when comparison is within
alertThresholds	Array[Object]	Values that raise an alert
alertThresholds[].value	Number	Threshold value (required)
alertThresholds[].direction	Enum	Whether the alert fires above or below the value: above, below (required)
alertThresholds[].severity	Enum	Seriousness of the alert: low, medium, high, critical (required)
linkedInitiativeID	String	Strategic initiative the metric supports
performanceID	String	Performance record the metric reports into

{
  "timeboundMetricID": "TBM-MACHINE-TEMPERATURE",
  "title": "Machine 101 Temperature",
  "description": "Temperature of a critical machine component, updated every 10 seconds.",
  "source": "Machine-101-Sensor-B",
  "unitOfMeasure": "Celsius",
  "updateFrequency": "realTime",
  "updateInterval": "PT10S",
  "target": {
    "value": 85.0,
    "comparison": "atMost"
  },
  "linkedInitiativeID": "INIT-ENERGY-EFFICIENCY"
}

Predictive Model

Attribute	Type	Description	Required
predictiveModelID	String	Unique identifier for the model, e.g. PM-INVENTORY-FORECAST	✓
title	String	Name of the model	✓
description	String	What the model predicts and from what
modelAlgorithm	Enum	Kind of algorithm: regression, classification, timeSeriesForecast, clustering, anomalyDetection, neuralNetwork, ensemble, largeLanguageModel, other
inputTimeboundMetricIDs	Array[String]	Timebound metrics that feed the model	✓
predictedTimeboundMetricID	String	Timebound metric the model outputs	✓
accuracy	Object	Measured accuracy of the model
accuracy.metric	String	Accuracy measure, e.g. MAPE, AUC, accuracy (required)
accuracy.value	Number	Value of the measure (required)
accuracy.measuredDate	Date	Date it was measured
modelRiskModelID	String	Model Risk Management model profile that governs the model
useCaseIDs	Array[String]	Standardised AI use cases the model is used in

{
  "predictiveModelID": "PM-INVENTORY-FORECAST",
  "title": "Inventory Demand Forecast",
  "description": "Predicts future inventory demand from sales data, seasonality and promotional events.",
  "modelAlgorithm": "regression",
  "inputTimeboundMetricIDs": [
    "TBM-SALES-VOLUMES",
    "TBM-PROMO-EVENTS"
  ],
  "predictedTimeboundMetricID": "TBM-INVENTORY-LEVELS",
  "accuracy": {
    "metric": "accuracy",
    "value": 0.955
  }
}

Use cases

• Major events. Data flows ingest turnstile and camera feeds; timebound metrics track crowd density at each gate and car park use; a predictive model forecasts peak parking demand from ticket sales, weather and live traffic.
• Manufacturing. Data flows carry sensor readings from assembly machines to a quality engine; timebound metrics track temperature, vibration and energy use; a predictive model forecasts the remaining useful life of a part.
• Healthcare. Data flows move patient vitals from bedside monitors to the health record; timebound metrics track heart rate and oxygen levels with alert thresholds; a predictive model forecasts the likelihood of deterioration.
• Regulatory monitoring. Data flows watch legislation registers and regulator publication pages; each new instrument becomes a Trigger; a timebound metric tracks how many records are past their freshness threshold.