Key takeaways
- ROI starts from avoided downtime cost, not the number of sensors installed.
- Build two scenarios: a conservative case and an expected case.
- Only the avoidable share of failures counts — not every breakdown is predictable.
- Add secondary savings: spare parts, scrap, overtime, safety and warranty.
- Target high-value, high-failure machines first; credible payback often falls within 12–18 months.
Start from the cost of stopping, not the cost of sensors
Most predictive-maintenance business cases fail because they begin with the bill of materials. The right starting point is the cost of an unplanned stop. A line that stops loses throughput, may scrap in-process material, pulls technicians off planned work, and sometimes triggers overtime or missed shipments. Quantify that first. The sensor and platform spend is the smaller number you subtract at the end.
The avoided-cost formula
The core model is straightforward. Annual benefit equals the downtime you avoid, valued at what an hour of downtime actually costs, plus the secondary savings predictive maintenance unlocks:
Annual benefit = (downtime hours × cost per hour × avoidable share) + secondary savings
The word that does the most work is avoidable share. Predictive maintenance catches developing mechanical and electrical faults — bearing wear, misalignment, overheating, lubrication loss. It does not catch a forklift hitting a panel. A conservative model assumes you address only part of the failure population.
The input table
| Input | What it means | Conservative | Expected |
|---|---|---|---|
| Unplanned downtime hours / yr | Hours the asset is stopped unexpectedly | Use your CMMS history | Use your CMMS history |
| Cost per downtime hour | Lost margin + labour + scrap + penalties | Lower-bound estimate | Fully-loaded estimate |
| Avoidable share | % of stops a model can predict in time to act | 25–35% | 40–55% |
| Secondary savings / yr | Spare parts, scrap, overtime, warranty, safety | Counted modestly | Counted in full |
| Sensors + integration (one-time) | Hardware, wiring, edge, engineering | Full quote | Full quote |
| Platform + ops (annual) | Connectivity, platform, monitoring effort | Full quote | Full quote |
Worked logic (illustrative)
Suppose a critical asset records 120 hours of unplanned downtime a year at a fully-loaded $4,000 per hour. In the conservative case you assume a 30% avoidable share: 120 × $4,000 × 0.30 = $144,000 in avoided downtime. Add modest secondary savings and subtract one-time and annual costs to get net benefit and payback. In the expected case, a 50% avoidable share lifts the same line to $240,000. The two scenarios bracket your real outcome — commit on the conservative one.
These figures are illustrative placeholders for the method, not an Amotus client result. Replace every number with your own CMMS and finance data before you decide.
Payback and prioritisation
Divide net first-year benefit by one-time cost to get a simple payback. The pattern across plants is consistent: the best returns come from a short list of high-value, high-failure machines, not a blanket roll-out. Instrument those first. For the right machines, a credible payback often falls within 12–18 months, after which the avoided-cost benefit continues while incremental cost is mostly connectivity and platform.
- Rank assets by downtime cost × failure frequency.
- Pilot the top few; measure avoidable share against reality.
- Re-run the model with measured numbers, then scale to the next tier.
The mistakes that inflate the model
Two errors flatter the spreadsheet and then disappoint in production. The first is using a generic cost-per-hour instead of your own fully-loaded number — lost margin, labour, scrap and any contractual penalties for the specific line. The second is assuming a high avoidable share before you have measured one. Pad neither. A model you can defend to finance uses your CMMS history for downtime, your accounting for the hourly cost, and a deliberately cautious avoidable share until the pilot gives you a real figure to replace it.
It also helps to keep secondary savings honest. Spare-part optimisation, reduced scrap, less overtime and fewer safety incidents are real, but they are easy to double-count or overstate. Count them modestly in the conservative case and only in full in the expected case.
What makes the savings real
A model on a slide is not a saving. The benefit only materialises when an alert reliably changes a maintenance decision before the failure — which means clean data from the floor (often OPC-UA and MQTT), a model tuned to real failure modes, and a maintenance team that trusts and acts on the signal. The economics are sound; the discipline of acting on the prediction is what turns the spreadsheet into cash. For the full pilot-to-scale picture, see the IIoT pilot-to-scale guide.
Where Fundamentum fits
Fundamentum, our Canadian IoT platform, is the control plane that turns a predictive-maintenance pilot into a fleet-wide program: device identity, governed OTA, role-based access and a SOC 2 Type II audit trail across every instrumented machine, so adding the next asset is configuration, not re-engineering — and it bridges OT data to your existing cloud (AWS, Azure) only when required. See the platform →
Frequently asked questions
What is the basic formula for predictive-maintenance ROI?
Annual benefit equals downtime hours × cost per hour × avoidable share, plus secondary savings (spare parts, scrap, overtime, safety, warranty). Subtract one-time sensor and integration cost and annual platform cost to get net benefit, then divide first-year net benefit by one-time cost for a simple payback.
What is the “avoidable share” and why does it matter?
It is the percentage of unplanned stops a model can predict early enough to act on. It matters because not every failure is predictable — a forklift impact is not. A conservative model uses 25–35%; an expected case 40–55%. Measure your real share during the pilot before scaling.
Why model a conservative case and an expected case?
Because a single optimistic number is easy to attack and hard to defend. Two scenarios bracket the real outcome: commit the budget on the conservative case and treat the expected case as upside. It makes the business case credible to finance and survivable if reality lands low.
What payback period is realistic?
For the right machines — high downtime cost and frequent, predictable failures — a credible payback often falls within 12–18 months. A blanket roll-out across all assets typically pays back far slower, which is why prioritising a short list of high-value machines matters.
Which machines should I instrument first?
Rank assets by downtime cost multiplied by failure frequency, and instrument the top few first. The best returns come from a short list of high-value, high-failure machines, not a plant-wide roll-out. Use the pilot to measure the real avoidable share, then scale to the next tier.
Related reading
Talk to an IoT engineer — free
Book a FREE 30-minute consultation with our team. No slides, no obligation — a working session on your connectivity, platform or compliance questions.
