Chapters

19 IoT Perspectives: Shared Definitions and Decisions

applications
iot
perspectives
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.

19.1 Make the Thermostat Decision Together

IoT design brings technical feasibility, user value, business viability, operational support, and risk into the same decision. This chapter shows how to recognize a single-perspective failure, define the system under discussion, and resolve cross-functional trade-offs with testable evidence. It is part 2 of 2; IoT Perspectives: Role Lenses introduces the viewpoints used here.

A thermostat review has five confident answers: security wants stronger protection, hardware wants longer battery life, data wants more samples, UX wants fewer prompts, and operations wants simple recovery. Each answer protects something real. The problem is deciding together without letting one answer hide the others. We will first map the system and choose a useful definition, then compare alternatives and record the evidence behind the compromise.

A perspective is a role’s view of the system, including its primary question and success metric. Six recurring perspectives are security, maker or DIY, hardware, architecture, data analytics, and user experience (UX). Disagreement often means that the team has not aligned its evidence. Assign a perspective advocate for each viewpoint in a review, and require every architectural decision to address at least three perspectives, with those advocates signing off.

A security-dominated project can become unusable through too many passwords or slow interactions. A hardware-dominated project can be cheap but vulnerable; a data-dominated one can drain batteries and bandwidth. UX alone can sacrifice safety for convenience, while architecture alone can produce an over-engineered system. These are failure patterns to look for, rather than reasons to dismiss a specialist’s concern.

The Three Ingredients Test and Five Verbs establish the shared starting point. Systems Evolution to IoT explains the computing economics and chip trends that made such systems practical. The next approaches turn that shared background into a review the thermostat team can use.

KEY CONCEPTS

  • Role perspectives

    Make competing questions and success metrics explicit.

  • Five system layers

    Check hardware, software, network, data, and security.

  • A definition for the decision

    Separate identity and context from product classification.

  • Trade-off evidence

    Measure the compromise and record when to revisit it.

19.2 Learning Objectives

By the end of this chapter, you will be able to:

  • recognize the failure caused by a single role perspective
  • build a practical layered definition for an IoT system
  • resolve cross-functional trade-offs with testable evidence

19.3 Map the Whole System

Before comparing designs, map what the team is responsible for. A taxonomy groups the system into layers so that a missing responsibility is visible.

Figure 1 gives the five layers a whole IoT system has to cover, and it is worth checking a project against.

Five-layer IoT taxonomy showing hardware, software, network, data, and security layers with representative examples for each
The five layers cover hardware, software, network, data, and security. Production systems need all five.

Read the five columns as a checklist rather than a ranking. The first three are the ones teams naturally staff: hardware to sense and act, software from firmware up to the user’s app, and a network layer that already offers four different ranges to choose between. The fourth column is where value appears, because data only matters once collection turns into an alert or an action. The fifth is the one that gets postponed. Security is drawn as a full layer rather than a feature inside another, since secure boot, transport encryption, and access control each belong to a different part of the stack. The line at the foot of Figure 1 is the test: production systems need all five.

Use the following inventory to check each layer:

  • Hardware: sensors such as temperature, motion, pressure, light, GPS, and humidity devices; actuators such as motors, relays, valves, LEDs, speakers, and displays; controllers and computing platforms such as Arduino, ESP32, Raspberry Pi, and ARM Cortex; and power systems such as battery, solar, mains, and energy harvesting.
  • Software: firmware such as RTOS, Embedded C, and MicroPython; edge applications such as gateway logic, local analytics, and protocol bridges; cloud services such as data storage, ML/AI, APIs, and dashboards; and user applications such as mobile apps, web portals, and voice control.
  • Network: PAN technologies such as Bluetooth, Zigbee, Z-Wave, and Thread; LAN technologies such as Wi-Fi and Ethernet; WAN technologies such as 4G/5G, NB-IoT, and LTE-M; and low-power wide-area network (LPWAN) technologies such as LoRaWAN, Sigfox, and RPMA.
  • Data: collection through sampling, streaming, and event-driven messages; storage through time-series databases, NoSQL, or data lakes; analytics through real-time, batch, ML/AI, and predictive workflows; and action through alerts, automation, visualization, and reporting.
  • Security: device protections such as secure boot, attestation, and TPMs; network protections such as TLS (and legacy SSL terminology), VPNs, and firewalls; data protections such as encryption, access control, and privacy controls; and application protections such as authentication, authorization, and API security.

The thermostat review needs the complete map even when its immediate dispute concerns battery life. Reference Architectures expands these five layers into system design patterns; the map here keeps the review focused on which responsibilities each option covers.

19.4 Choose the Definition That Fits the Question

Now that the roles are clear, define the object they are arguing about. IoT definitions are tools: choose the one that fits the decision you need to make.

The Internet of Things (IoT) gives physical objects digital identities and interfaces for communicating in connected environments. Those interactions take place in social, environmental, and user contexts. Two foundational definitions explain different parts of that idea: identity and context, and network addressability.

19.5 Academic Definitions

Figure 2 sets the three definitions beside each other so you can pick by the decision you face.

Three complementary IoT definitions compared across academic, semantic, and practical perspectives, showing what each emphasizes and when to use it
Academic definitions emphasize identity and context. Semantic definitions emphasize unique addressability. Practical definitions test the product boundary.

Each horizontal card gives an emphasis, a use, and a question; the question is the test to take away. The academic card asks what makes a system meaningfully smart, which is the right question when framing research or arguing about scope. The semantic card asks how a thing exists and works with others on a network, which is what a standards or architecture decision turns on. The practical card asks whether a product qualifies at all, which is what a product manager needs before writing a spec. Figure 2 makes the point without ranking anyone. The definitions do not compete, they answer different questions, and choosing the wrong one produces a long argument about the wrong thing.

The academic identity-and-context definition describes things with identities and virtual personalities in smart spaces, using intelligent interfaces to communicate within social, environmental, and user contexts. Its strength is explaining the intelligence and context awareness that distinguish IoT from simple networking. It fits research papers and vision documents, although it can be operationally vague.

The semantic definition starts with the two words. The Internet is the worldwide network of interconnected computer networks using the Internet protocol suite, TCP/IP. A thing is a physical object; together, the definition describes a worldwide network of interconnected, uniquely addressable objects using standard communication protocols. Its strength is grounding IoT in established networking and unique identity. This fits standards and protocol design, although addressability alone says little about intelligence.

IoT History traces how formal definitions evolved as connected things moved from vision to deployment. For a thermostat specification, the practical classification test below will be easier to verify; the broader definitions still help the team explain what its system is meant to do.

19.6 Connected Things Diversity

Figure 3 is worth reading for its final band rather than its inventory of objects.

Five connected-thing families surround an Internet of Things hub. Below, a smart parking sensor detects car presence and reports an occupied or vacant state to a city dashboard over LoRaWAN or NB-IoT. The sensor's unique ID maps to a specific parking space; the dashboard can query individual spaces. The final band links IoT value to sharing data and coordinating actions.
Five families of connected things share an IoT hub. The parking example maps a sensor identity to a space and reports its occupancy. Shared data supports coordinated action.

The upper panel groups everyday objects into five families around an Internet of Things hub, making the breadth visible. The middle panel shows what sharing data means in the chapter’s parking example: a smart parking sensor detects car presence and reports an occupied or vacant state to a city dashboard over a standards-based network such as LoRaWAN or NB-IoT. The sensor’s unique ID maps to a specific parking space, so the dashboard can query individual spaces. The band along the foot carries the broader lesson: value comes from devices sharing data and coordinating actions, not simply from the number of connected objects. Take that away from Figure 3 when the list below starts to look like a catalogue. A thousand connected objects that never exchange anything are just a thousand separate products.

This academic visualization captures a key insight: virtually any physical object can become an IoT device. The diversity shown here includes:

  • Transportation: Aircraft, buses, ships, cars
  • Buildings: Smart homes, offices, factories
  • Appliances: Refrigerators, washing machines, beds
  • Personal devices: Phones, cameras, headphones, speakers
  • Infrastructure: Trash bins, street lights, utility meters

The connecting lines illustrate that IoT value emerges not from individual smart devices, but from the network effects when thousands of diverse devices share data and coordinate actions.

Source: University of Edinburgh - Principles and Design of IoT Systems

19.7 Test the Product Boundary

For practitioners, the most operationally useful definition provides clear, verifiable criteria:

Minimum Requirements Definition: Thing + Computation + Internet

This framework enables:

  • Product classification (is this smart toaster IoT?)
  • Requirements validation (missing internet connectivity = not IoT)
  • Evolution tracking (embedded -> connected -> IoT progression)

The source material offers two broader lenses alongside this operational test. One frames IoT as people and things being connected anytime, anyplace, with anything and anyone, over any suitable path or service. Another describes interrelated devices, machines, objects, animals, or people that carry unique identifiers and transfer data across a network without requiring a person to move every message. These definitions describe reach, identity, and machine-to-machine exchange; they do not replace the engineering boundary below.

The same source also presents a four-part minimum: an everyday thing, computation, internet connectivity, and computational intelligence. In this course, the first three remain the quick classification test because they are directly observable. Intelligence is then reviewed as a separate capability question: does the system merely report or relay state, or does it use computation to interpret context and support a decision? Keeping the two questions separate prevents a connected object from being called intelligent by default while preserving the source’s reminder that useful IoT systems usually do more than move raw readings.

19.8 IoT Classification Test

For the thermostat, turn that boundary into three verifiable questions:

  1. Is it a physical object, or “thing”? If not, it is not IoT.
  2. Does it have computation capability, such as a processor or firmware? If not, it is only a sensor or passive object.
  3. Can it communicate over the internet, directly or through a gateway? If yes, it meets the practical IoT definition; if not, it is an embedded or local system.

A Bluetooth-only fitness band that syncs to your phone, which then uploads to the cloud, qualifies. Under the gateway model used in this course, the phone provides the device’s internet path, so the three criteria are satisfied. The trade-off is that remote reach disappears when the gateway phone is absent.

  • Academic, identity, and context definitions capture intelligence and context well, but can be operationally vague. They fit research papers and vision documents.
  • Semantic, network, and unique-ID definitions emphasize networking foundations, but can miss the intelligence aspect. They fit standards bodies and protocol design.
  • Practical Thing plus Compute plus Internet definitions are clear and verifiable, but can oversimplify. They fit product development and project scoping.

Best Practice: Use the practical definition for product development and the academic definitions for conceptual understanding and research.

The Embedded → Connected → IoT progression helps track changes in this boundary. A product can pass the classification test while still lacking a viable battery, usable recovery, or a safe authentication flow. Classification settles scope; the next review must settle whether the design can operate successfully.

Check knowledge

Definitions, layers, and perspective conflicts · Review the first four sections on a separate check page.

Open check

19.9 Compare Security and Battery With Explicit Assumptions

Definitions classify the system; multidisciplinary review decides whether the system is worth shipping. The same device can satisfy the IoT definition and still fail if one perspective quietly dominates.

Understanding diverse IoT perspectives is not merely academic — it is essential for project success:

Cross-Functional Challenges:

  • Security engineer vs Hardware engineer: Security demands encryption overhead; hardware wants minimal power consumption
  • Data scientist vs Network engineer: Analytics wants high-frequency data; network wants minimal bandwidth
  • UX designer vs System architect: Users want seamless experience; architecture requires complexity

Resolution Strategy:

  1. Acknowledge that each perspective is valid and necessary
  2. Identify conflicts early in design phase
  3. Make explicit trade-offs with stakeholder buy-in
  4. Document decisions for future reference

Example: In this simplified medical-wearable comparison, assume ChaCha20-Poly1305 uses half the encryption energy overhead of AES-256-GCM. Measure power and verify that the chosen authenticated-encryption implementation meets the deployment’s security requirements.

19.10 Security-Battery Tradeoff

Given: Medical wearable, AES-256-GCM vs ChaCha20-Poly1305 authenticated encryption

  • AES-256-GCM is assumed to add about 15% battery drain and about $3 to the bill of materials.
  • ChaCha20-Poly1305 is assumed to add about 7.5% battery drain and about $1.50 to the bill of materials, half the encryption energy overhead in this simplified comparison.

Impact on a nominal 7-day battery (180 mAh): this simplified model applies the stated penalties as reductions in effective capacity, rather than deriving runtime from measured current traces.

  • With AES-256-GCM, a 180 mAh battery at 85% effective capacity gives about 153 mAh, or 5.95 days in this example.
  • With ChaCha20-Poly1305, the same battery at 92.5% effective capacity gives about 166.5 mAh, or 6.48 days.

Key insight: Under the example’s assumed costs and loads, ChaCha20-Poly1305 extends battery life by about 8.8% and saves $1.50 BOM per unit; assess security suitability independently. At 100,000 units, that is $150,000 in assumed hardware savings.

Neither option’s security can be expressed as a percentage of the other’s. Compare complete authenticated-encryption implementations on the target hardware and check applicable FIPS requirements before selection. A more capable chip, different sampling policy, or radio duty cycle may change the result. Security Threats supplies the threat-model, encryption, and authentication detail behind this role’s requirements.

For the thermostat team, the lesson is to expose the workload behind a claimed improvement. Do not remove a required control simply because its modeled cost is visible; look for compensating changes elsewhere and measure their effect. The calculation audit and explorer let you inspect these assumptions separately.

The maths

What Encryption Costs the Battery · Follow Ada’s calculation audit.

Open explorer

Try it

Battery vs Security Explorer · Compare modeled runtime under assumed encryption penalties.

Open explorer

19.11 Give the Compromise an Owner

The battery example needs several reviewers. A smart building energy management system extends the same approach to a complete project team.

Use Figure 4 to see how six specialist views meet at product choices that no single role can settle alone.

Cross-functional IoT project team map for a smart building energy system, showing security, hardware, architecture, data, UX, and maker/prototyping roles coordinated around shared product decisions
Six specialist perspectives meet at shared product decisions. Sampling, access, retention, and processing each require an explicit compromise.

Start at the shared goal: reduce waste while preserving comfort and safety. Follow the six perspectives: security, maker, hardware, architecture, data analytics, and user experience. Then read the shared decisions. Sampling balances one-second readings against network capacity through a 15-second average and event-driven bursts. Authentication balances MFA against one-tap access. It uses biometric confirmation for critical controls and a PIN for dashboards. Retention keeps one year of detailed data and five years of aggregates, with documented deletion behavior. Processing combines edge anomaly detection with cloud predictive modeling. Finally, at least three perspectives must sign off on each architectural decision.

Key design decisions and their perspective trade-offs:

  • Sensor sampling rate: Data wants one-second readings while architecture worries about network capacity. A balanced resolution is a 15-second average with event-driven bursts.
  • User authentication: Security wants MFA while UX wants one-tap access. A balanced resolution is biometric confirmation for critical controls and a PIN for dashboards.
  • Data retention: Data teams want five years while security and privacy want minimal retention. A balanced resolution is one year of detailed records, five years of aggregated records, and documented deletion behavior.
  • Edge versus cloud processing: Hardware has limited edge compute while data teams want complex models. A balanced resolution is edge anomaly detection plus cloud predictive modeling.

Return to the thermostat with the same questions: who owns sampling, authentication, retention, and recovery, and what evidence will the other roles accept? The system is one product even when its specialists use different success metrics. Privacy and Compliance helps connect data-use ambitions to privacy constraints; consent and deletion belong in the design discussion alongside analytics.

Common Pitfalls

Three recurring pitfalls show why these reviews must begin early. Adding too many features before validating core needs wastes weeks when user testing reveals that people wanted simpler interactions. Define and test a minimum viable version, then add complexity in response to validated requirements. This gives the maker perspective a useful role without letting prototype speed decide production policy.

Treating security as phase 2 leaves hardcoded credentials, unencrypted channels, and unsigned firmware that are expensive to fix after deployment. Include security in the initial review, because prototype patterns become production patterns. Designing only for the happy path is equally costly: test sensor failures, connectivity outages, and cloud unavailability, and make devices fall back to a safe, locally functional state. For the thermostat, simple recovery is an operating requirement that the other roles must review together.

19.12 Set Constraints Before Scoring Preferences

The review now needs a repeatable procedure. Use it when a role conflict is real enough to affect cost, risk, service life, compliance, or usability.

When different perspectives clash during IoT product development, use this structured resolution framework:

  • Security versus usability: Security engineers ask for stronger authentication while UX designers protect low-friction access.
  • Performance versus battery: Hardware engineers protect service life while data scientists ask for real-time inference.
  • Cost versus features: Product teams push bill-of-materials limits while engineers price radios, sensors, memory, and certification realistically.
  • Privacy versus monetization: Legal and privacy teams constrain data use while business teams may see value in location, behavior, or occupancy data.
  • Speed versus reliability: Software teams want fast rollout while architecture teams protect device variants, rollback paths, and field reliability.

Score the conflict on two dimensions:

  • Regulatory risk, 1 to 5: Does this violate GDPR, safety standards, FCC regulations, or another binding rule?
  • User impact, 1 to 5: Does this make the product unusable, unsafe, misleading, or worthless for customers?

Use the following resolution rule as a review heuristic. If regulatory risk ≥ 4, prioritize the binding security, privacy, safety, or other compliance requirement. If user impact ≥ 4, prioritize UX and product needs unless the regulatory threshold also applies. If both are < 4, seek a compromise that protects both perspectives. A score does not waive a binding requirement, even below 4; the team must establish which rules actually apply.

The six perspectives prioritize different characteristics from the eleven ideals in IoT Requirements. Weights help compare options that satisfy the constraints. They cannot make a noncompliant option acceptable. In one illustrative medical-device weighting, Secure 20% + Fast 15% + Low Maintenance 20% totals 55% of the criteria. Another example gives security and reliability a combined 40% weight. These percentages are decision-model choices, not weights mandated by healthcare regulations.

A medical wearable can involve six stakeholders across engineering, legal, and product. In the worked comparison, Device A scores Secure 7/10 and Low Maintenance 4/10; Device B scores Secure 9/10 and Low Maintenance 10/10, with seven-day rather than two-day battery life. These are example ratings that need supporting evidence before they guide a product decision. Consider both usability and the complete power budget rather than assuming a cipher name determines either outcome.

Try it

IoT Perspective Tradeoff Tool · Adjust stakeholder weights and compare the supplied design options.

Open explorer

19.13 Find a Third Option and Record the Evidence

With the constraints clear, look for an alternative that addresses both needs. The following examples illustrate synthesis rather than measured product outcomes. Each names what to test before claiming that the compromise works.

For a smart door lock, security asks for fingerprint or facial recognition while UX asks for one-tap daily access. The proposed third option is tiered authentication: one-tap at home on Wi-Fi with authenticated phone presence, biometrics away from home, and escalation to two-factor authentication (2FA) after 3 failed attempts. Treat the proposed 95% one-tap and 5% biometric split as a hypothetical target requiring a measured workload. Home-network presence alone does not establish an authorized user; security must validate the complete authentication design.

For an activity-tracking wearable, the data scientist wants an ML model every 10 seconds, while hardware reports 8-hour runtime against a 48+ hour requirement. The proposed adaptive-sampling and edge/cloud hybrid uses a low-power accelerometer at 50 Hz and an activity-versus-rest classifier on the microcontroller (MCU), assumed below 1 mW. It invokes the 150 mW ML model only during activity. The proposed 90% rest time, 52-hour runtime, 97% accuracy, and 6.5× battery life are hypothetical targets. Measured accuracy and power traces must establish whether the workload supports them; the rest fraction alone does not prove runtime.

For an agricultural sensor, product wants a $12 GPS module, but business requires a bill of materials (BOM) of at most $50; adding GPS produces $62. With the $50 base BOM, alternatives include a $2 Bluetooth beacon with a setup-mapping phone app (total $52), a $0.05 QR code with the installer’s phone GPS during deployment (total $50.05), or LoRaWAN geolocation (total $50). Only the last meets the BOM target. The last proposal assumes time-of-arrival measurements from 3+ gateways and ±50 m accuracy, with no added sensor BOM cost so that the $50 target remains intact. Treat accuracy and gateway availability as assumptions to validate. Zero added sensor BOM does not establish that infrastructure or service operation is free.

These examples give the thermostat team a method: change the design, rather than merely deciding which advocate wins. Context-dependent authentication, adaptive sampling, and deployment-time mapping each preserve a useful part of both requirements. Testing and Validation helps make each perspective’s definition of success testable.

Document who advocated for what, regulatory-risk and user-impact scores, the selected approach, why it balances the needs, the supporting evidence, and the condition that would reopen it. A review trigger is a change or observation that requires the team to reconsider its decision. Six months later, the record prevents the same rejected approach from returning without reference to the original evidence. Six-month reviews can also reveal that an assumption has changed.

Use a short record that a later reviewer can understand without attending the meeting:

  • Decision and date: a short description, dated YYYY-MM-DD.
  • Perspectives: Security (John), UX (Sarah), Product (Maria).
  • Conflict: Security requests mandatory 2FA; UX requests one-tap access; Product prioritizes adoption.
  • Scores: illustrative Regulatory Risk = 3 and User Impact = 4, with the applicable obligations recorded separately.
  • Resolution: propose context-based authentication, with one-tap at home and biometrics away, subject to authentication review.
  • Rationale and evidence: explain the access friction being reduced, the required protections, and the measured workload supporting the choice.
  • Review trigger: a breach, failed validation, or an audit finding. Review HIPAA obligations only where they actually apply to the deployment.

For the wearable calculation, record: “Assumed ChaCha20-Poly1305 has half the encryption energy overhead of AES-256-GCM; measure both on the target hardware and check FIPS requirements before selection.” Keep the assumption, measured evidence, and selection decision distinguishable.

The meeting procedure is identify, score, find a third option, and document. It acknowledges that each perspective is valid and necessary, identifies conflicts early, secures stakeholder agreement, and leaves a design log with evidence and review triggers. The thermostat review succeeds when every advocate can explain both the chosen compromise and the evidence that could overturn it.

Check knowledge

Evidence and shared decisions · Apply the trade-off method and review the remaining activities.

Open check

Summary

Six perspectives—security, maker, hardware, architecture, data analytics, and UX—bring distinct priorities, questions, and metrics. The five-layer taxonomy keeps hardware, software, network, data, and security visible. Academic definitions explain identity and context; semantic definitions explain network addressability; the practical Thing + Compute + Internet test supports product classification, including a gateway path.

A thermostat, medical wearable, or pet tracker can qualify as IoT and still fail its users. Strong reviews make requirements explicit, compare complete designs, and distinguish assumed benefits from measurements. Document the advocates, scores, decision, rationale, evidence, and review trigger so that later teams can understand the compromise. The Quiz Navigator provides practice across chapters.

Every IoT project negotiates technical feasibility, user value, business viability, operational support, and risk. Make those perspectives explicit early, and support the compromise with testable evidence.

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.