20 IoT Requirements: Classification and Connectivity
More
Depth: Practitioner
Progress saves as you read. Mark sections complete as you go; your progress appears in the sidebar.
- Podcast audio coming soon
- Character narration audio coming soon
20.1 Turn the Care-Home Idea into a Testable Promise
IoT requirements connect a useful idea to evidence that a system can deliver it. They describe what must be sensed, communicated, decided, and maintained before a team chooses parts. This chapter uses two passes: establish whether a product qualifies as IoT, then establish which qualities make it suitable for its job. Together, those passes lead to a connectivity shortlist and requirements that can be checked across the architecture.
Picture a care-home team proposing a room sensor that warns when heat may harm a resident. The idea sounds useful, but the owner needs to know what is sensed, who gets the warning, how soon it arrives, and what happens when power or the link fails. A reading reaching a distant service is only one part of that promise. The resident needs a useful response, and the team needs evidence that the response remains safe when a component fails.
Write the promise from the resident’s point of view. Name the physical event, useful range, place, update time, allowed error, reader, action, and safe fallback. Add battery life, service life, repair owner, data retention, access rule, and the way the unit receives a safe update. Identify the local, service, repair, and data owners explicitly so that each failure has someone responsible for handling it.
The problem is that words such as fast, secure, reliable, and low cost leave those decisions open. Turn each word into a measured check, record which result passes or fails, and name who accepts the limit. Keep any urgent safe response in the room or building if a remote round trip could miss the deadline. A distant service may store, compare, and notify, but it must not be the only path to immediate protection.
Now work through the approaches: classify the service path, choose the qualities that matter, translate them into measurable requirements, and trace them through five layers. You will use the Three Ingredients Test, prioritize the eleven IoT characteristics, and compare Wi-Fi, LoRaWAN, and cellular against application constraints. The aim is a reviewable decision with explicit inputs, errors, and change rules, rather than a radio chosen by habit.
Basic familiarity with IoT is enough; no technical background is required. IoT Introduction introduces the Five Verbs, while IoT Perspectives and Definitions helps explain why stakeholders ask different questions about the same promise.
KEY CONCEPTS
- Three ingredients
A thing, embedded computation, and internet reach establish the classification.
- Domain priorities
Choose the 3–4 qualities that matter most for the job.
- Measurable evidence
Attach a limit, a test, and an owner to each promise.
- Five-layer responsibility
Carry identity, data meaning, and recovery rules through the whole system.
20.2 Part Objectives
- Define related chapters and resources with explicit inputs, errors, and change rules.
- Trace requirements cross the five layers across its components and failure boundaries.
20.3 Classify the Service Path, Then Judge Its Quality
The Three Ingredients Test asks whether a physical thing, embedded computation, and internet reach work together in a useful data or control loop. Embedded computation means logic running within the device. Internet reach may be direct or supplied through a gateway, an intermediary that carries the device’s data onward. The test forces you to name the asset, the local logic, and the path carrying readings, commands, alerts, configuration, or updates beyond the local environment.
A timer microwave has a thing and computation but no connectivity, so it is an embedded device. A Bluetooth-only lock adds local wireless communication but lacks internet reach, so it is a connected product. A soil sensor with a LoRaWAN gateway has all three ingredients and qualifies as IoT. The sensor does not need to open its own TCP connection: the gateway can complete the path. Removing an ingredient changes what the product can observe, decide, or control remotely.
The worked Three Ingredients figure in Definition and Value Flow makes the test explicit. Device Evolution develops the distinction between embedded, connected, and IoT products. Use that distinction as a requirement filter, rather than treating the category name as evidence of usefulness.
A Bluetooth-only lock gains a different requirement set when it reaches a cloud service through a working Wi-Fi bridge or integrated Wi-Fi radio. A factory vibration sensor can reach a historian through an edge gateway and still count as IoT when the complete path supports remote visibility or action. The requirement is the service path, not a specific radio. For the care-home sensor, trace that path all the way to the person or service that must respond to the warning.
Make the classification record concrete. For the thing, name the physical asset, installation point, user, operating environment, enclosure limits, and expected lifetime. For computation, name the logic that samples, filters, stores, controls, protects, compresses, or signs data. For internet reach, name the direct or gateway-mediated path carrying data, commands, updates, or alerts beyond the local link. Then name usefulness: the action, decision, maintenance task, safety response, or service workflow that changes because the device is connected.
Qualification is only the first pass. The second asks whether the device is good IoT for its job. The eleven quality benchmarks are Ubiquitous, Smart, Agile, On Demand, Blend into Background, Secure, Low Maintenance, Fast, Upgradable, Growing, and Adaptable. No device excels at all eleven. Choose the 3–4 that matter most in the domain, assign priority weights, and record the tradeoff you accept.
A smart thermostat, cold-chain tracker, irrigation valve, medication dispenser, and fleet sensor can all satisfy the three ingredients while needing different strengths. A cold-chain tracker may prioritize battery life, location coverage, tamper evidence, and data retention. A thermostat may prioritize local control, comfort, update safety, and household privacy. A fleet sensor may prioritize mobility, cellular roaming, ruggedness, and remote diagnostics. A farm sensor optimized for millisecond response while ignoring battery life is solving the wrong problem.
Use Application Domains to compare domain priorities and IoT Use Cases to examine implementation examples. IoT History and Paradigm Shifts adds lessons from technology evolution when reviewing a claim about a newer technology. These comparisons help separate qualifying as IoT from fitting a particular job.
20.4 Turn Priorities into Requirements and Connectivity Choices
Once the care-home team has named its priorities, each quality needs a measurable design constraint. An acceptance criterion is the testable condition used to decide whether a requirement is met. “Low maintenance” becomes battery access, enclosure sealing, calibration interval, replacement process, spare-device handling, and an over-the-air (OTA) update policy. “Fast” becomes a latency budget, sampling rate, queue behavior, command timeout, and local fallback. Latency is the delay before the required response. “Secure” becomes provisioning, credential storage, access control, audit logging, firmware signing, and revocation.
Write each requirement as a budget, threshold, or operational rule. A cold-chain sensor might require temperature samples every five minutes, upload within fifteen minutes when coverage exists, buffer seventy-two hours offline, survive washdown, rotate credentials on ownership transfer, and preserve audit records for a defined retention period. A smart-home lock might require local unlock when the cloud is down, revocation within minutes, low-battery warning before failure, and a clear household recovery path. These statements are design inputs, not documentation added after choosing the hardware.
For the room sensor, keep the response time separate from data age: how soon a warning arrives and how old its underlying reading is are both part of the promise. Count each power draw, test weak and dead cells, and connect the battery budget to the maintenance process. A service-life claim also needs a field-check plan and a next-check plan, with the repair owner able to act on the result.
Connectivity choices become easier to defend once range, power budget, update frequency, and environment are explicit. Wi-Fi suits high-bandwidth local devices, particularly mains-powered equipment with installation support. Thread or Zigbee may fit home and building meshes where a border router is acceptable. Bluetooth Low Energy (BLE) can support phone-assisted setup, short-range sensing, or local wearables. LoRaWAN suits small, infrequent telemetry over long range for battery-powered sensors when downlink limits are acceptable. LTE-M or NB-IoT may fit wide-area assets when gateway ownership is impractical and the business model can carry subscription cost; cellular also suits mobile assets.
Those are candidates to test against requirements. Do not choose Wi-Fi, LoRaWAN, cellular, MQTT, or a cloud platform because it sounds modern. Choose the least demanding combination that still satisfies range, power, latency, security, maintenance, data quality, and user experience. Streaming everything raw to the cloud can raise latency, cost, and fragility, so the architecture must also decide where sensing, filtering, networking, and decisions belong.
Use a repeatable sequence to defend the shortlist:
- Prioritize characteristics. Choose the few qualities that matter most and state the tradeoff being accepted.
- Translate each quality. Write measurable requirements for power, latency, freshness, update rate, range, identity, maintenance, recovery, privacy, and support.
- Check interactions. Lower latency may worsen battery life; a higher sample rate may increase data cost; stronger authentication may add setup friction.
- Attach verification. Choose a bench test, field trial, protocol trace, security review, battery calculation, accessibility check, or support drill as appropriate.
- Keep requirements alive. Revisit assumptions from prototype to pilot, pilot to fleet, and one domain to another. Reopen the review when parts or use change.
For the separate warehouse total cost of ownership (TCO) comparison and further prioritization of the eleven characteristics, continue to IoT Requirements: TCO and Design Trade-offs. Its cost comparison extends the connectivity decision beyond choosing a radio. Price both the normal case and the hard case before accepting the design’s known limits.
The Simulation Playground provides separate interactive tools for exploring choices, and the Video Gallery offers visual learning resources alongside this requirements review.
20.5 Carry Each Requirement Across the Five Layers
A connectivity shortlist is only defensible if the whole system can keep the promise. The five layers are physical, edge, connectivity, cloud, and application: they carry the flow from sensing through local processing and transport to remote processing and the user’s view or action. A requirement usually touches several of them. Freshness, for example, may need a sensor timestamp, device clock behavior, gateway buffering rule, MQTT or HTTPS message policy, cloud ingestion timestamp, application stale-data label, and operator response rule. A dashboard label alone leaves that requirement ambiguous.
Keep source, event time, stated unit, quality mark, and code version through every hand-off. Data semantics means the agreed meaning of the data: units, timestamp source, calibration state, location precision, firmware version, quality flags, device state, and command meaning. Define these before analytics or automation use the readings. A Celsius value sampled inside an enclosure, averaged over one minute, and marked stale means something different from an instantaneous probe reading in Fahrenheit. In the care home, the team must be able to interpret the reading before treating it as evidence for a warning.
Recovery also crosses layers. The device may need safe local behavior when its radio is down. A gateway may need store-and-forward queues, which hold data until forwarding is possible. The cloud may need idempotent command handling, meaning repeated handling does not repeat the command’s effect. The application may need to show whether a command is pending, applied, rejected, expired, or unsafe to retry. Otherwise, a retry storm, duplicated command, stale reading, or silently offline device can appear normal until a user notices the failure.
Specify buffering, retry limits, duplicate handling, command expiry, and local safe behavior for each failure mode. Test lost and slow links, an old or bad clock, a full store, and a stale view. Also test a moved probe, loose wire, failed rule, safe stop, hard restart, failed update, and safe rollback. These tests connect the resident’s promise to what the system actually does under strain, rather than assuming that a successful demonstration covers recovery.
Identity has its own chain of responsibility. Manufacturing may assign a device identifier and certificate. Provisioning, the setup that binds a device into service, may associate it with an owner, site, room, vehicle, patient, or asset. The gateway may authorize which devices join; the cloud may map device identity to tenant identity and permissions. The application may expose ownership transfer and decommissioning. Define authentication, authorization, credential rotation, and retirement before field rollout so support teams can safely replace, resell, retire, or investigate devices. Include a wrong-user test in the acceptance evidence.
Finally, make failures visible. Observability means the logs, counters, health checks, trace IDs, fleet reports, and alert thresholds used to inspect operation. Define them for every layer that can fail. Keep the raw proof and final view, record passed and failed cases, and link the evidence to the claim. State each known limit and who accepts it; a record without that decision cannot explain whether an observed failure is tolerable.
The Concept Navigator offers another place to test understanding. The chapter check brings classification, the service path, and domain priorities together before you carry this evidence into the formal definition.
Summary
The care-home sensor’s promise becomes reviewable when it names the sensed event, useful range, allowed error, timing, action, safe fallback, and responsible owners. The Three Ingredients Test establishes whether a physical thing, embedded computation, and direct or gateway-mediated internet reach work together. Quality then depends on selecting the right 3–4 characteristics for the domain and turning them into measurable requirements.
Connectivity follows those requirements. Identity, semantics, reliability, and observability must remain testable across the physical, edge, connectivity, cloud, and application layers. Verification connects a design claim to evidence, including failure behavior, and keeps the decision open when parts, scale, or use change.
Name the thing, computation, and internet path; then prove that the complete system can keep its promise to the user, including when power, links, or updates fail.
Carry this evidence into IoT Requirements: Definition and Value Flow, beginning with the formal definition of the Internet of Things.

