Chapters

22 IoT Case Studies: Transferable Evidence

applications
iot
use
cases
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.

22.1 Read a Success Story as a Design Decision

A case study is a detailed account of how a real organization solved a problem, including the surprises and mistakes. For an IoT designer, its value is the evidence it supplies for a new decision. The practical question is: what would you copy, what would you change, and what would you need to prove?

Picture a city report claiming that smart parking cut search time. Before copying the design, ask what was measured, where, for whom, and against which earlier state. Change the street shape, network reach, repair team, or service goal, and the result may no longer hold. A clear success story can reveal a useful pattern while hiding upkeep, weak areas, or a special local deal. A shared platform may lower repeated work while adding one large point of control.

Keep that parking claim in view as we compare Barcelona’s public-service platform with Volkswagen’s industrial reliability work. Barcelona uses sensors in parking spaces, waste containers, street lighting, and water services. Volkswagen uses production data and sensors for predictive maintenance, including welding-gun fault prediction. Both cases make ownership, integration, and staff response as important to assess as the sensing hardware.

The problem is deciding how much of a reported result survives a move to another setting. We will approach it through a claim card, four comparison lenses, a separation of reusable decisions from local conditions, and a trace from the physical event to the recorded outcome. This prepares you to analyze deployments, compare city and industrial strategies, evaluate predictive-maintenance return on investment (ROI), and account for total cost of ownership (TCO), including hidden costs.

This is part 1 of 2, with an estimated reading and review time of about 25 minutes at Advanced level. For the companion industrial case, use Lessons from Real Deployments: Volkswagen Predictive Maintenance and Cross-Case Lessons. Here, the parking claim gives us one decision to carry through the evidence framework.

KEY CONCEPTS

  • Case evidence

    Separate observed results from a vendor promise.

  • Transfer boundary

    Check site, load, and support conditions before reuse.

  • Decision record

    Keep the source and uncertainty attached to the claim.

22.2 Part Objectives

  • Test a case study is a transfer test with a concrete scenario and pass criteria.
  • Validate lessons from real iot projects with a concrete scenario and pass criteria.

22.3 Attach the Claim to Its Evidence

Start the parking review by making a claim card. Put the exact claim at the top, then add the source and date, place and group, earlier state, time span, measure, and owner. Record the site, users, sample, missing data, costs, and local limits. Leave a blank wherever the report supplies no fact. A blank makes uncertainty visible; a guessed sample size hides it.

Next, draw the path behind the claim. Start at the field event and follow the reading, link, store, rule, application, and work action. Ask where data could be lost or changed, which parts were new, and which were already in place. Ask who paid to fit, run, fix, and replace them. A result without this path may be real, but it is hard to transfer. Mark each statement as a reported fact, a team choice, or a lesson to test elsewhere.

Technical labels alone cannot fill those gaps. Bandwidth means how much data a link can carry in a given time. Latency means the time from an event to the useful result. A protocol is an agreed set of rules for passing data. Naming the link or protocol does not establish that its capacity and delay suit the new street layout or service goal. The architecture needs to explain how the measured problem reaches an accountable action.

Barcelona and Volkswagen are deliberately different, so read their evidence through the same four lenses:

Transfer lensBarcelona smart cityVolkswagen predictive maintenance
ProblemParking, waste, lighting, water, and city-service coordinationUnplanned downtime on high-value production assets
ArchitectureSentilo, open APIs, fiber, LoRaWAN, MQTT, REST, city applicationsIndustrial sensors, edge gateways, TensorFlow Lite, OPC UA, MES integration
AdoptionMunicipal departments, citizens, developers, service operatorsTechnicians, maintenance planners, line managers, plant leadership
EvidenceService savings, utilization, route efficiency, ecosystem participationDowntime reduction, false-positive reduction, payback, work-order use

For the problem lens, ask which cost, risk, delay, or service failure justified funding. For architecture, follow the devices, networks, gateways, middleware, analytics, and user interfaces carrying the operating loop. For adoption, identify who had to trust the output and change their work. For evidence, distinguish a measured result from an asserted one. A case that cannot answer all four questions offers weaker support for a new design.

The table identifies what to examine; it does not verify every result. Barcelona connected a city with over 19,000 sensors to manage parking, waste collection, street lighting, and water. The city was reported to save $58 million annually through smart water technology and gain $50 million annually in parking-fee revenue, according to the Cisco city case profile. Keep those separately reported water and parking results separate from the opening claim about parking search time. Coordinating city services remains a further implementation challenge.

For Volkswagen, the production-data and welding-gun examples establish the application being discussed. Treat sensor-count, payback, and alert-adoption figures as unverified unless a primary case source is supplied. Likewise, a hypothetical factory model can help examine ownership, integration, and staff response, but its arithmetic is not a measured deployment result.

This discipline protects the review from advertising language. If a vendor calls a city platform “open,” ask how device data, APIs, departments, and developers connect. If a maintenance vendor reports high model accuracy, ask for alert acceptance, work-order routing, false-positive control, and downtime reduction. Neither a platform label nor an accuracy headline supplies the missing operating evidence.

Integration and change management can demand substantial effort alongside the technology work in both city-scale and factory-scale projects. A reported service saving does not remove the need to establish who owns the system, how it connects to existing processes, and how it will be maintained over time.

The reusable lesson is the connection between problem, architecture, adoption, and evidence. With that connection recorded, we can ask which decisions would survive a change of site.

22.4 Separate the Pattern from Local Conditions

A transfer boundary states where the original result’s supporting conditions stop applying. For the parking claim, a new street shape or repair team may cross that boundary even if the same platform is available. Begin with the decision logic behind the deployment before choosing its visible hardware or interface.

Barcelona’s open Sentilo platform matters because the city has many departments, many vendors, and a long-lived public-service mandate. A private campus may still need open APIs without needing the same civic data portal. Volkswagen’s edge analytics matter because production downtime is expensive and high-rate machine data is too large to ship raw. A small workshop may still need predictive maintenance without the same gateway density or model-training pipeline. The cases support examining these choices in context; they do not establish a universal sensor catalogue or cloud stack.

Write a translation note that names the decision boundary. For Barcelona, it might ask which city services can share sensor events without exposing personal data or locking the city into one vendor. For Volkswagen, it might ask which signals should be processed locally so technicians receive actionable alerts instead of noisy dashboards. These questions turn a story into requirements and reveal assumptions to recheck before promising a similar result.

Carry out that translation in four steps:

  1. Copy the decision logic. Reuse the sequence of problem definition, pilot selection, measurement, trust-building, and scale-out.
  2. Recheck the economics. Replace the case’s downtime, labor, maintenance, and public-service numbers with numbers from the new organization. Include fitting, operation, repairs, and replacement when assessing TCO.
  3. Recheck the operating owner. Identify who receives alerts, approves action, maintains devices, and explains failures.
  4. Recheck the data contract. Define timestamps, identifiers, event meanings, privacy boundaries, retention, and API ownership before scaling.

For the new parking site, establish baseline cost and response time before a measurable pilot. Scale only after independently verified savings and integration work. The point of the pilot is to test the changed conditions, rather than assume that another city’s reported improvement applies unchanged. Local network reach, repair capacity, and the service goal belong beside the financial comparison.

Adoption needs the same attention. City projects often fail when departments cannot agree on data ownership or procurement rules. Factory projects often fail when technicians do not trust alert quality. False positives can erode that trust, so predictive-maintenance alerts need validation with technicians and a process for reviewing false alarms. Hardware and software must fit the people and existing processes that turn a signal into action, and someone must maintain that fit over time.

End the translation note with three lines. Keep names the part worth testing. Change names the local part that must differ. Prove names the test needed at the new site. If the case cannot support all three lines, use it to generate questions. It does not yet support a design answer.

Those lines describe what might transfer. To assess whether the underlying result is trustworthy, we now need to follow its data back to the physical event.

22.5 Follow the Event to the Outcome

Data lineage is the trace from a physical event through the system to a decision record. In a Barcelona-style service, a parking bay, waste container, or irrigation point produces an event. LoRaWAN, a fiber-connected gateway, or cellular backhaul carries it. Middleware such as Sentilo normalizes the message; APIs expose it to applications; city operators use it to dispatch, price, maintain, or report service quality.

Follow that same path for the parking claim. If timestamps, device identifiers, calibration status, or service ownership disappear along the way, the dashboard may remain convincing while the evidence weakens. Being able to see a reading is different from being able to explain which event it represents and which service action followed it.

The Volkswagen-style path is faster and more local. Vibration, current, acoustic, or thermal signals are sampled near the asset. An edge gateway extracts features, a model scores anomalies, and the maintenance system creates or updates a work order. A technician then confirms whether the alert matched a real condition. That confirmation connects the model output to maintenance work and an eventual outcome.

Ask whether every component preserves enough context for a responsible decision. The appearance of TensorFlow Lite, OPC UA, SAP Manufacturing Integration, or a digital twin in an architecture does not answer that question. The audit fields below make the context explicit without losing the main case-study thread.

A smart-city event should record device identity, measurement time, location granularity, calibration state, consent or privacy class where relevant, API consumer, and action taken. These fields let reviewers connect a field reading to its use without silently losing its privacy or measurement context.

An industrial alert should record asset identity, signal window, model version, confidence or severity band, recommended action, spare-part status, operator acknowledgement, and eventual maintenance outcome. The record must carry enough context to connect the alert to the technician’s response and later result.

Use three checks when reviewing either record. Freshness: can you tell when the physical event happened and whether the value is stale? Ownership: can you identify who acts on the event and who maintains the measurement path? Outcome link: can you connect the event to a cost, service, safety, reliability, or adoption measure?

These audit fields distinguish a traceable engineering pattern from a one-off success narrative. They make the route from physical system through IoT components to a human or automated decision available for review.

Return to the parking report with the claim card, transfer note, and lineage together. You should now be able to explain the earlier state, the measured change, the operating owner, and the local conditions that need another test. A missing field remains a visible question, rather than becoming an unsupported assumption in the new city’s design.

Check knowledge

Evidence frame: review claim cards, the four transfer lenses, and traceable outcomes.

Check knowledge

Summary

A useful case connects a measured problem to its architecture, operating workflow, adoption, and accountable result. The parking claim needs a source, baseline, measure, time span, owner, and visible gaps before it can guide another city’s design.

Barcelona and Volkswagen expose different local constraints while sharing a need for trusted data, integration, maintenance, and staff response. Preserve the distinction between reported results and unverified claims. Carry the decision logic into a measurable pilot, replace local economics, and trace the evidence from field event to outcome.

Reuse a case’s decision logic only with its evidence and limits attached: state what to keep, what to change, and what the new site must prove.

Carry this evidence frame into IoT Case Studies: Smart Parking Design, beginning with Design Contract: Smart Parking Sensors.

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.