Planning IoT Systems
The Planning domain covers everything that happens before hardware is chosen: turning a business problem into testable requirements, confirming the site and budget can support them, and producing a defensible business case. Items here are scenario-based - you are given a stakeholder, a constraint, or a cost table and asked what should happen next.
Turn goals into testable requirements
- Functional requirements describe behaviour: what is sensed, how often, what triggers an action, what the user sees.
- Non-functional requirements set the limits that drive design later: sample rate, end-to-end latency, availability target, device battery life, data retention, ruggedisation, and data classification.
- Write each requirement so it can be verified - "report tank level to ±1 cm every 5 minutes, 99.5% of the time" beats "accurate monitoring".
Map stakeholders and constraints
Operations owns the process, IT owns the network and identity, OT/engineering owns the physical asset, and security and legal own risk and privacy. A plan that skips the OT change-control board or the privacy review stalls at implementation.
Do the site survey before the design
Confirm power at each mounting point, RF coverage and interference, cable routes, environmental rating (IP class, temperature range), physical access for servicing, and whether the site permits new cabling. Coverage assumptions are the most common planning error: concrete, metal racking, and walk-in freezers all change it.
Build the case
Compare capital cost (devices, gateways, mounting, installation labour), recurring cost (connectivity, cloud, licences, support), and quantified benefit. Payback = capital ÷ (annual benefit − annual recurring cost).
Worked example. 500 sensors at $120 installed = $60,000, plus $15,000 of gateways and platform setup, so $75,000 capital. Recurring cloud and connectivity is $12,000 a year. Avoided spoilage and manual inspection is valued at $45,000 a year. Net annual benefit = $45,000 − $12,000 = $33,000. Payback = $75,000 ÷ $33,000 ≈ 2.3 years. If the sponsor requires payback inside 18 months, the plan must change - fewer sensors, cheaper connectivity, or a larger quantified benefit - not simply be approved.
Traps
- Treating a pilot as production: a pilot needs explicit success and exit criteria plus a scale-up decision.
- Omitting recurring costs, which usually dominate over five years.
- Accepting a stated coverage claim without a survey.
- Deferring data classification and privacy review until the design phase.