Chapters

6 IoT History: Forecast Boundaries

applications
iot
history
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.

6.1 Ask What the New Capability Changes

An IoT forecast can get the arithmetic right and still describe the wrong market. A forecast is an estimate of a future outcome. Its product boundary specifies what is being sold: a device, a fleet capability, or a managed service. When that boundary changes, current buyers and current willingness to pay may no longer describe the opportunity. This chapter follows those changes and shows how to compare a forecast with evidence without confusing phones, subscriptions, and people.

6.2 Sensor Squad Time Travel

This is part 2 of 2, following IoT History: Early Lessons. Keep the Sensor Squad’s time-travel scenario in mind. Temperature Terry finds a magical time machine in the lab, and Sammy, Light Lucy, Motion Marley, and the battery-management controller join him. Their first stop is London in 1878, where Sir William of the Post Office dismisses the telephone because messenger boys already deliver messages. Lila points out to Sammy that he is judging the new capability through a familiar job.

Their next stop is an American phone company in 1983. Max shows a mobile phone the size of a brick; the engineers ask why anyone would carry a phone when there is one on the desk. Bella compares a reported forecast of 900,000 units with ITU’s 738 million worldwide mobile subscriptions in 2000. The quotient is about 820, but those labels matter: it is not a validated forecast error without matching geography and units. Back home, Sammy hears the same question about a fridge with Wi-Fi or a connected light bulb. The story asks what connectivity could make possible beyond the existing appliance’s job.

A smart light looks unnecessary if the only question is whether it turns on and off. The original examples widen that question to saving energy, supporting health, and helping elderly people live safely. They suggest possibilities to investigate; a previous technology’s success does not establish demand for this one. The problem is to recognize a changed job while still requiring evidence. We will first define the forecast boundary, then examine the old comparison habits, and finally calculate what the numbers can support.

KEY CONCEPTS

  • Forecast assumption

    State the adoption population and time horizon before comparing outcomes.

  • Installed base

    A device shipment is not the same as an active service.

  • Revision trigger

    Update the forecast when cost or deployment evidence changes.

6.3 Learning Objectives

  • Test forecasts fail when boundaries move with a concrete scenario and pass criteria.
  • Calculate forecast error calculator from stated measurements and limits.

6.4 Read a Forecast as a Dated Hypothesis

The Sensor Squad’s phone comparison raises a question that applies equally to connected fleets: what exactly did the forecast count? IoT expands a device into a fleet, a protocol ecosystem, a cloud or edge pipeline, and a service workflow. Installed base, interoperability, regulation, cybersecurity, and data quality can matter as much as the device bill of materials. Counting today’s buyers at today’s price misses a market that appears after the product changes the job, the data, or the service model.

Low-power wide-area network (LPWAN) forecasts published during the mid-2010s are a useful historical case. Analysts expected rapid connection growth and divided the opportunity among infrastructure, agriculture, consumer, utility, transport, and industrial sectors. Preserve those charts as dated expectations, not as current market measurements. Their value is the hypothesis they expose: long range, low device energy, and inexpensive small-message connectivity could unlock fleets that conventional cellular and short-range radios served poorly.

A connection forecast compresses several uncertain steps into one rising curve: coverage rollout, standards and operator support, module availability, procurement, installation, recurring fees, battery servicing, and conversion of a connected asset into a useful workflow. A rising connection count does not establish that customers find enough value to keep paying for the service.

A retrospective review therefore records the forecast date, forecast horizon, unit of measure, technology boundary, geography, and sector definitions. It then compares like with like: forecast connections against observed connections, forecast revenue against observed revenue, and forecast sector share against a consistently classified installed base, the population already deployed. If definitions or sources differ, the comparison is directional evidence rather than a score of whether the prediction was right.

For a current LPWAN design, replace the old market total with present coverage, device, certification, service-life, and operations evidence. Historical charts can explain why investment accelerated. Selecting a network requires evidence about the deployment being proposed. The applications module index supplies the wider application context for deciding which workflow the connection must support.

6.5 Make the Operating Model Testable

Once the forecast names its boundary, the pilot needs measurements that can test it. Use log-scale thinking for cost and adoption, but do not assume cost decline solves everything. A LoRaWAN sensor, LTE-M tracker, Matter device, or OPC UA gateway still needs identity, provisioning, permissions, monitoring, update policy, and lifecycle ownership. Hardware cost can fall while total operating cost remains high: truck rolls, false alarms, certificate expiry, firmware updates, privacy review, and support tooling do not shrink at the same rate.

The technical boundary changes the measurement plan. Collect sensor readings alongside missing data, duplicate messages, latency, battery drain, false positives, user overrides, maintenance outcomes, and support cases. For predictive maintenance, the evidence record needs fault labels, lead time, avoided downtime, and enough negative examples to show the model is not just noisy. For workflow improvement, it needs handoff time, alert fatigue, escalation accuracy, and evidence that staff actually changed behavior.

Three boundaries help organize that record. The adoption boundary separates early users from mainstream buyers because they may value different outcomes. The system boundary includes network effects, standards, integrations, and support channels that create value a standalone device cannot show. The trust boundary covers privacy, safety, security, and maintenance obligations that can slow adoption even when sensor cost falls.

Return to Sammy’s connected light. Asking whether it creates a useful new action widens the review; asking who owns power, identity, integration, security, support, and trust makes the answer testable. A historical analogy should lead back to a specific signal, workflow, owner, and pilot threshold. State the old metric, new behavior, technical contract, adoption assumption, evidence threshold, and trigger for stopping or expanding the pilot. Without that operating model, history remains a slogan rather than a review tool.

6.6 Recognize the Old Frame Before Extrapolating

Time: ~15 min | Level: Intermediate | ID: P03.C01.HIST

The operating record helps expose paradigm blindness: judging a new capability only through an established way of thinking. A paradigm is that familiar frame, such as “phones stay on desks.” Innovation creates something new and useful; disruption changes an established way of doing things; adoption is people starting to use the new thing. These distinctions explain why an invention’s initial comparison may miss its later market.

The chapter’s four-stage pattern is a way to inspect that mistake, rather than proof that every dismissed invention will succeed:

  1. Anchoring to current behavior. The evaluation trap asks which current users would want a worse product instead of what new users could do with new capabilities. The mobile example contrasts “Who needs a $4,000 mobile phone?” with “What happens when it costs $200 in 15 years?” The latter makes the cost assumption visible.
  2. Measuring by old metrics. The comparison trap chooses metrics that favor the existing paradigm. Sir William Preece’s messenger-boy comparison emphasizes message delivery while missing instant voice communication. An old-frame objection and a real engineering burden need separate examination.
  3. Extrapolating linearly. The math trap assumes adoption follows a simple trend, although it may follow an S-curve. Even a compelling curve must be compared with observations in the same geography and unit.
  4. Ignoring second-order effects. The imagination trap considers only the direct use case and misses later applications: SMS, mobile internet, app stores, ride-sharing, and mobile banking. Text messaging, ride-sharing apps, and watch-based health monitoring illustrate uses beyond the initial phone or watch comparison.

For the connected bulb, the same review considers switch-flipping behavior, lumens per watt, assumptions that cost will not fall, and possible later uses such as health monitoring through light-usage patterns, Li-Fi communication, and emergency evacuation guidance. These examples broaden the investigation; they do not establish that a particular bulb will become essential infrastructure. Established expertise can miss a new market, while practical limits still determine whether a service works.

The two tracks in the linked paradigm-blindness figure make that distinction inspectable. Compare incumbent metrics on the expert-evaluation track with new behaviors and markets on the actual-adoption track before dismissing an early prototype. Then return to the evidence threshold rather than treating imagination as a substitute for measurement.

Check knowledge

History as Evidence Filter: distinguish changed jobs, dated forecasts, and testable operating evidence.

Check knowledge

6.7 Separate the Phone Forecast from the Observation

With the comparison traps visible, return to the number Bella used. In the early 1980s, AT&T commissioned McKinsey & Company to forecast the mobile phone market. A widely reported forecast put the year-2000 market at 900,000 mobile phones. The chapter’s source does not establish a matching geography and unit for comparison with the later worldwide subscription count.

The ITU recorded 738 million mobile-cellular subscriptions worldwide in 2000. A subscription is an active service record, not necessarily one unique phone or person. That distinction prevents the story’s striking quotient from becoming a misleading claim about measured forecast accuracy.

The historical question remains useful: “Why would anyone want to walk around with a phone?” Analysts can anchor to high initial prices, extrapolate early adoption, or miss later uses such as SMS and apps. Understanding the technology and its current costs does not automatically reveal how human behavior could change once it becomes available. Conversely, a large numerical gap alone cannot establish which assumption caused it.

The same caution applies to forecasts of 50B IoT devices in 2025. This is a dated forecast example, not a current installed-base measurement. Assuming today’s $50 sensor price rather than examining a $2 cost trajectory can change the adoption model, as can applications beyond monitoring. Neither a lower assumed price nor a historical surprise proves the new forecast. Record what would have to change and what evidence would show it happened.

6.8 Calculate the Ratio, Then Qualify the Claim

The calculation is straightforward once its limits are explicit. Using the reported mobile numbers from the forecast and the ITU observation gives an observed-to-forecast quotient of 738M / 0.9M = 820. Subtracting the forecast first gives (738M - 0.9M) / 0.9M = 819, or 81,900% above the forecast. The factor is 820; 819 is the relative excess before conversion to a percentage. Neither is a validated mobile-market forecast error when the underlying geography and units have not been matched.

For matching counts, the same method does support a forecast comparison. A forecast of 1,000 connections and an observed count of 1,500 connections gives 1.5 times forecast, or 50% above it. A forecast of 1,200 connections with 900 observed gives 0.75 times forecast, or 25% below it. The differences of 500 and 300 connections are gaps in counts, not multiplication factors. These calculations still do not decide which network to buy today.

6.9 Use the Calculation to Review Assumptions

A calculator makes a numerical gap visible, but it cannot repair a mismatch in the input definitions. Before interpreting the display, state whether the values represent connections, revenue, devices, or subscriptions. Keep the forecast date separate from its target year, and keep sector classifications consistent. An old connection projection does not become a fresh count of willing customers simply because it can be entered into a calculator.

The review should also distinguish a falling device price from the cost of keeping a useful service running. Return to coverage, recurring fees, battery servicing, installation, and support evidence before accepting an adoption curve. Ask which assumption changed: the buyer population, the job, the service boundary, or the deployment conditions. A large ratio is a reason to inspect those assumptions; it is not a replacement for that inspection.

The separate calculation audit shows the full working and rounds only at the end. Use the calculator to try the reported mobile inputs or your own IoT forecast after checking scope. Its original exercise also offers the attributed IBM “5 computers” example from 1943, with inputs 5 and more than 5,000,000,000. Treat those as illustrative exercise inputs, not a verified historical forecast comparison. A large multiplier by itself does not diagnose paradigm blindness.

The maths

Ada separates the 820× quotient from relative percentage error, with full working.

Read calculation audit

Try it

Forecast Error Calculator: explore ratios and percentage gaps, then check whether the inputs match.

Open calculator

Check knowledge

Forecast Scale: report ratios and percentage gaps without confusing units or procurement evidence.

Check knowledge

Summary

A forecast needs a stated product boundary, adoption population, horizon, geography, and unit. A device can become part of a service whose value depends on integrations, operating costs, maintenance, and trust. Historical analogies help reveal new behaviors, but a useful review returns to a signal, workflow, owner, and evidence threshold.

The reported 900,000-phone forecast and ITU’s 738 million worldwide subscriptions yield an arithmetic quotient of 820. Matching scope and units is required before calling it a measured forecast error. Revise an IoT forecast when cost or deployment evidence changes, and test the service someone would use and pay to keep.

Ask what new behavior connectivity enables, then test the operating model. Compare a forecast with observations only after their boundaries and units match.

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.