Chapters

7 Domain Requirements Contracts

applications
domain-requirements
architecture
governance
Listen
Uses your device's voice. Tap ▶ beside any paragraph to hear just that one.
  • Podcast audio coming soon
  • Character narration audio coming soon
Lecture
Video lecture coming soon. The slides are the lecture for now.
Ask AI
Pick a prompt. We copy it for you and open an AI service; paste it there. Nothing is sent to IoTClass.

7.1 Turn a Domain Claim Into a Testable Promise

An IoT requirement is useful when a team can test whether its system meets it. “Low latency,” “high reliability,” “long battery life,” “medical compliance,” and “industrial integration” leave too much unstated. Latency is the wait between an event and a useful result. A testable latency rule names who must act, how long they may wait, and what happens when the result is late.

Think of two temperature sensors. One warns a grower about frost; the other guards vaccines in a clinic. They may use the same thermistor, but the harm from a late or wrong reading differs. Their power, records, alarms, maintenance, and backup plans must match their jobs. Imagine their review meeting: each claim must become a contract before the team can build.

  1. Physics Phoebe leads a review where vague domain claim cards hover above an empty contract sheet, while the team waits for measurable limits.

    How can each claim become a promise the team can test?

How can each claim become a promise the team can test?

A requirement envelope is the full set of limits a design must meet together. It names the time limit, power source, number of devices, data to keep, applicable rules, and safe response to a fault. Comparing Domain Requirements introduces latency, reliability, scale, power, data volume, regulation, and cost as selection dimensions. Here, those comparisons become budgets, interfaces, ownership, validation tests, and failure behavior.

Try it

Rank latency, reliability, power, and other limits to build a domain requirement envelope.

Rank the requirements →

Lab

Compare domain profiles by latency, power, coverage, governance, and consequences of failure.

Compare domain requirements →

Ask who needs the result, how soon they need it, what harm a bad result can cause, how long proof must remain available, and who looks after the device. Then turn the answers into tests: time the full alert path, cut the network, lower the battery, send a wrong value, and replace a device. Check whether the record still explains what happened and what the user should do next. Safety, privacy, and legal work may require further tests and named standards.

The problem is choosing a technology before these obligations are clear. The approaches below start with the operating decision, compare candidates against the complete envelope, and carry each promise into engineering budgets and interfaces. Across healthcare, agriculture, industry, transport, retail, and smart homes, the aim is traceability from a requirement to a technology decision, validation test, and fallback plan.

KEY CONCEPTS

  • Operating limits

    A requirement names the decision, the allowed delay, and the consequence of failure.

  • Linked budgets

    Latency, reliability, power, security, data, lifecycle, and integration must fit together.

  • Interface evidence

    Telemetry, commands, and alerts need fields and states that explain what happened.

  • Validation and ownership

    A decision record names the test, fallback, and operating responsibility.

7.2 Learning Objectives

After this page, you should be able to:

  • Explain why domain requirements are measurable operating constraints, not technology preferences.
  • Translate healthcare, agriculture, industrial, transport, retail, and smart-home needs into requirement envelopes.
  • Compare candidate radios, protocols, platforms, and governance controls against pass/fail boundaries.
  • Define latency, reliability, power, security, data, lifecycle, and integration budgets for an IoT system.
  • Preserve traceability from requirement to technology decision, validation test, and fallback plan.

7.3 Write the Operating Contract

A domain requirement is not a preference for a fashionable sensor, radio, cloud, or dashboard. It is a measurable constraint created by the decision the IoT system has to support. A patient-alert system, irrigation controller, factory line, delivery vehicle, and smart home lock can all measure temperature or motion, but they do not have the same consequence of delay, missed data, power loss, privacy exposure, or maintenance failure.

Start by writing the job in operational terms: who acts, how quickly, what happens if the signal is wrong, how many devices are involved, how the device is powered, and which rules apply. Those answers narrow the design space before the team argues about LoRaWAN, NB-IoT, Wi-Fi, BLE, Zigbee, Ethernet, MQTT, HTTP, edge computing, or cloud dashboards.

Figure 1 separates a product’s position in a device matrix from the limits needed to accept its design.

A power-consumption and market-focus matrix places LoRa sensors and smart meters in low-power infrastructure, factory robots and CNC machines in higher-power industry, and wearables and consumer devices in the consumer half. A separate checklist names six limits needed for a full requirement envelope.
Device profiles differ by power consumption and market focus. LoRa sensors and smart meters prioritize low-power infrastructure. Factory robots and CNC machines accept higher power use for industrial control. A complete requirement envelope also names time, power source, device count, retained data, applicable rules, and safe failure behavior.

A device matrix can help locate a product, but it cannot define its contract. A device-quadrant comparison uses two axes: power consumption and industrial or consumer use. LoRa sensors and smart meters occupy the low-power industrial corner, where battery life shapes the design. Factory robots, CNC machines, and other machine tools accept higher power use to meet industrial control demands. Wearables and consumer devices occupy the consumer half. Those positions explain broad operating profiles; they do not specify acceptance tests.

Alongside those axes, write six limits: time limit, power source, device count, data to keep, applicable rules, and safe failure behavior. A grid can place the grower’s sensor and the clinic’s sensor near similar products while leaving their different duties unresolved. The written limits must establish how each reading becomes a useful action and what evidence remains afterward.

The important shift is from feature thinking to constraint thinking. “We need temperature monitoring” is too broad to design. “We need a calibrated temperature record that proves vaccine storage stayed within policy, alerts a responsible person before excursion, and survives an audit” is a requirement. “We need a field temperature reading that warns a grower before frost damage and can run through a growing season” is a different requirement. Both may use a thermistor, but their acceptable latency, connectivity, calibration evidence, battery replacement, and data retention are different.

Three questions keep that shift concrete. The time constraint asks how much delay is allowed from physical event to useful action. The failure constraint asks what harm, cost, or lost trust follows a missed, late, duplicated, or false reading. The operating constraint assigns who installs, powers, patches, calibrates, replaces, and retires the device fleet. Each answer belongs in the same contract as the sensor’s measurement range.

A requirements comparison should also include the people and systems around the device. A hospital monitor may need a nurse workflow, biomedical engineering ownership, EHR context, clinical alarm policy, privacy controls, and cleaning procedures. A city air-quality node may need public-space permissions, calibration visits, publication rules, and a way to explain uncertainty. A factory sensor may need plant-network segmentation, scheduled downtime for installation, CMMS integration, and clear authority to stop or inspect equipment. These are requirements because they decide whether the telemetry can safely become action.

7.4 Compare Candidates Against the Whole Envelope

With the operating job written down, compare candidates against a pass/fail boundary: the limit that determines whether a candidate meets a requirement. For a remote soil-moisture node, the envelope may include a 15-minute sampling interval, five-year battery target, LoRaWAN Class A uplink, IP67 enclosure, local rainfall context, and seasonal service visit. For a hospital asset tracker, it may include BLE tags, Wi-Fi access-point location or Ultra-Wideband anchors, patient-area privacy rules, HL7 FHIR integration, and a support path that does not interrupt care.

Write the envelope as numbers, interfaces, and operational commitments. Latency should be an end-to-end budget, not only a network figure: sampling interval, edge processing, radio access, broker routing, rule evaluation, notification, and human response all count. Reliability should include how the system behaves when a node is silent, a gateway is offline, a certificate expires, a firmware update fails, or an integration API is unavailable. Scale should name not just device count but provisioning rate, event rate, data retention, support tickets, and replacement logistics.

  • Connectivity: compare BLE, Zigbee/Thread/Matter, Wi-Fi 6, Ethernet/PoE, LoRaWAN, NB-IoT, LTE-M, 5G, and satellite links against range, duty cycle, handoff, gateway ownership, and subscription cost.
  • Message path: specify MQTT QoS, Sparkplug B state, OPC UA tags, DDS topics, CoAP resources, REST APIs, broker retention, duplicate handling, and offline queue behavior.
  • Platform fit: map the domain to Amazon Web Services (AWS) IoT Core, Azure IoT Hub, ThingsBoard, Home Assistant, Ignition, AVEVA PI System, or a custom service only after the latency, lifecycle, security, and data-retention envelope is written.
  • Governance: name the actual rules where they apply: Health Insurance Portability and Accountability Act (HIPAA) and Food and Drug Administration (FDA) expectations for some healthcare workflows, International Electrotechnical Commission (IEC) 62443 for industrial cybersecurity, International Organization for Standardization (ISO) 26262 for automotive safety work, General Data Protection Regulation (GDPR) for personal data, and GS1 EPCIS or Digital Link for supply-chain traceability.

Use the envelope to eliminate attractive but incompatible choices. A battery node that must sleep most of the day should not be forced into high-frequency telemetry just because a dashboard looks better with live charts. A safety-critical machine-control loop should not depend on a distant cloud rule engine when deterministic local control is needed. A consumer smart-home device may accept cloud-assisted features, but a door lock or smoke alarm still needs local behavior when the internet is unavailable. A retail shelf sensor may prioritize low cost and easy installation, while a pharmaceutical cold-chain sensor may prioritize auditability and chain-of-custody evidence.

Return to the grower and clinic before approving a choice. The review should end with a traceable decision record. For every selected technology, record the requirement it satisfies, the requirement it weakens, the test that will prove it, and the fallback if it fails. That record prevents later scope drift, such as adding video analytics to a low-power sensor network, changing a daily reporting system into a safety alert path, or expanding personal-data collection without updating retention and consent controls.

7.5 Carry the Budgets Into Interfaces

The envelope now needs an implementation that preserves its promises. An engineering budget allocates an allowed limit across the parts that consume it. Latency is divided across sensor sampling, filtering, edge inference, radio access, gateway processing, broker routing, database write, alert rule, and human or actuator response. Reliability is divided across device hardware, firmware watchdogs, network coverage, broker availability, cloud region design, alert escalation, and field replacement. Power is divided across sleep current, sensor warm-up, radio join, transmit time, retries, firmware updates, and battery chemistry.

Those budgets should be represented in interfaces, not only in architecture diagrams. A telemetry event should carry units, timestamp source, sequence or message id, firmware version, calibration state, quality flags, and enough asset or user context to support the domain decision. A command interface should separate requested, accepted, dispatched, acknowledged, applied, failed, and timed-out states. An alert interface should distinguish raw threshold crossing from validated event, suppressed event, escalated event, and operator acknowledgement. Without these states, the system cannot explain whether it acted late, acted twice, or failed to act.

  • Security budget: define device identity, X.509 certificate or secure-element storage, TLS policy, topic or API authorization, key rotation, over-the-air (OTA) signing, and decommissioning.
  • Data budget: define payload schema, units, timestamp source, clock drift tolerance, compression, edge aggregation, retention period, lineage, and deletion behavior.
  • Lifecycle budget: define provisioning, calibration, firmware versioning, configuration state, replacement mapping, support logs, rollback, and end-of-life handling.
  • Integration budget: define which systems consume the event: EHR, CMMS, MES, ERP, supervisory control and data acquisition (SCADA), GIS, mobile app, analytics warehouse, or customer notification service.

A technology is a candidate only when it fits all of the budgets at the same time. A radio that meets the range target but cannot meet the latency, power, provisioning, or regulatory budget is not a domain fit.

Implementation details vary by stack. MQTT topics may need tenant, site, device, and capability boundaries so authorization can be enforced without brittle string parsing. OPC UA nodes may need units, engineering ranges, quality status, and references to the physical asset. HL7 FHIR integrations may need device, observation, encounter, and patient identity handled with clinical safety checks. A LoRaWAN payload decoder may need to preserve battery voltage, frame counter, spreading factor, received signal strength indicator (RSSI), signal-to-noise ratio (SNR), and decoder version so field failures can be diagnosed later.

The hidden cost of poor requirements is usually paid during operations. If the team does not budget for certificate rotation, calibration expiry, alert fatigue, firmware rollback, gateway replacement, schema versioning, or decommissioning, the first successful pilot becomes hard to scale. Good domain requirements make those future operating costs visible while the architecture can still change. For the grower, the review must still explain who services the field node. For the clinic, it must still explain who receives an alert and how the calibrated storage record survives an audit.

Check knowledge

Review feature wishes, complete requirement envelopes, and operating ownership.

Check knowledge

7.6 Summary

Domain requirements become useful when they describe the operating decision, measurable budget, validation test, and fallback behavior. The grower’s frost warning and the clinic’s vaccine record may share a sensing component while requiring different latency, connectivity, calibration evidence, battery replacement, and retention commitments.

Compare radios, protocols, platforms, and governance controls against the whole envelope. Record which requirement each choice satisfies or weakens, how it will be tested, and who owns the response to failure. Carry those decisions into telemetry fields, command states, alert states, security controls, and lifecycle plans so the contract remains testable during operation.

Choose a technology only when it meets the complete operating contract, with evidence and ownership for what happens when it fails.

Hi, I’m Ada. What do you need?

Find help

Let’s find what you need

Find a lesson or return to your module guide.

Give feedback

Help improve this chapter

Tell us what helped or what needs fixing.

Tips

Try a small next step

A few ways to get more from this chapter.

Ask AI

Continue the conversation

Take the conversation further with your preferred assistant.

AI help is available in supported modules when enabled in Settings.

No live person answers on independent study. Feedback does not promise a personal reply.

For welfare or safety concerns, follow your school’s procedures outside IoTClass.