The "10:1 to 30:1 ROI within 12 to 18 months" figure for predictive maintenance is widely attributed to McKinsey, but we could not trace it to a primary McKinsey publication, so treat it as unverified. The closest primary benchmark is the US Department of Energy's FEMP guidance: about a tenfold return, and 8 to 12 per cent savings over preventive maintenance. Build your building's case from your own maintenance data.
Key takeaways
- Quote the US DOE FEMP benchmarks with their source. Do not put the "McKinsey 10:1 to 30:1" figure in a board paper without a primary citation.
- AI predictive maintenance pays best on HVAC, chillers, lifts, fire systems and switchgear. Plumbing and cosmetic work suit scheduled maintenance.
- Most commercial buildings already collect useful data in their BMS. Configure that before buying sensors.
- The operational gain often comes from automated work orders and contractor dispatch, as much as from prediction.
Is the McKinsey predictive maintenance ROI of 10:1 to 30:1 real?
The claim that "leading organisations achieve 10:1 to 30:1 ROI from predictive maintenance within 12 to 18 months" circulates widely online, usually credited to McKinsey, sometimes alongside Deloitte. When we searched for the original, we did not find a McKinsey report, article or dataset stating those ratios and that payback window. It appears to be a secondary statistic repeated from site to site.
That matters if you are writing a business case. A CFO or auditor who asks for the source will find the same loop of citations.
What the primary sources do support:
| Source | What it says | How to use it |
|---|---|---|
| US Department of Energy, FEMP Operations and Maintenance Best Practices Guide | Predictive maintenance can save roughly 8 to 12 per cent over preventive maintenance and up to 40 per cent over reactive maintenance, and cites an industry average return of about ten times investment | The most defensible public benchmark, though dated and US-based |
| McKinsey publications on Industry 4.0 and maintenance | Discuss reduced downtime and maintenance cost from advanced analytics, with figures that vary by study and sector | Cite the specific article and figure you rely on, never a paraphrase |
| Siemens, True Cost of Downtime (2024) | Unplanned downtime costs range from around US$36,000 an hour in FMCG plants to US$2.3 million an hour in automotive | Shows downtime cost varies hugely by sector; commercial buildings differ again |
The practical answer: ratios in the 10:1 range are plausible for well-targeted programs on critical equipment, but your number depends on your equipment, failure history and what downtime costs your tenants.
How does AI predictive maintenance work in buildings?
Predictive maintenance for buildings combines three elements.
- CollectBMS points and sensors: temperature, vibration, pressure, energy, airflow
- DetectModels learn normal behaviour per asset and flag drift
- Raise work orderFault categorised, priority set, job created
- DispatchRight contractor scheduled before failure
- Sensors and data collection. Real-time temperature, vibration, pressure, energy draw, airflow and humidity. For HVAC: motor vibration, refrigerant pressures, current draw and temperature differentials.
- Machine learning analysis. Models learn what normal looks like for each asset and flag deviations early, typically weeks ahead for gradual failure modes.
- Automated alerting and work orders. The system raises a work order, matches a contractor and schedules the repair off-peak, with a facilities coordinator approving.
Not every system justifies monitoring. The strongest returns come from HVAC and chillers, lifts, fire safety equipment and electrical switchgear.
Why is HVAC where predictive maintenance pays most?
HVAC is typically the largest energy load in a commercial building and a frequent source of reactive callouts. A chiller failing in a February heatwave becomes a tenant and business continuity problem fast.
Monitoring typically covers:
- Compressor health: vibration and current draw.
- Refrigerant charge: pressure and temperature differentials.
- Air handling efficiency: airflow against energy use.
- Heat exchanger fouling: degrading temperature differentials.
- Belt and bearing wear: changing vibration signatures.
Vendor case studies report large energy and maintenance savings, but they usually come from buildings with modern BMS and existing sensors. Retrofitting an older building costs more and pays back more slowly.
What is automated facilities support, and how does contractor dispatch work?
Automated facilities support links fault detection to work orders and contractor dispatch, so faults move from alert to scheduled job without phone tag. Platforms such as QFM, FMI Works and Taskflo offer versions of this:
- Skills matching: electrical to electricians, HVAC to refrigeration mechanics.
- Geography: jobs grouped by site and postcode.
- Availability: contractor calendars checked before assignment.
- Mobile completion: contractors receive, action and close jobs on a phone.
- Compliance checks: tools such as WHS Monitor verify contractor credentials and insurance before a job is issued.
| Manual | Automated | Impact | |
|---|---|---|---|
| Fault notice | BMS email reviewed when someone has time | Anomaly categorised and prioritised on arrival | Faster triage |
| Contractor selection | Phone calls to check availability | Matched by skill, location and workload | No phone tag |
| Job details | Relayed verbally | Sent to mobile with history and photos | Better first-time fix |
| Completion | Coordinator chases paperwork | Closed on mobile with documentation | Audit trail |
| Invoicing | Matched by hand | Matched to the work order | Less admin |
For multi-site operators, this is the same pattern we apply across field service and facilities work.
How should you build a predictive maintenance business case?
Skip borrowed ratios. Build it from your data.
- Current maintenance spend by systemFrom 12 months of work orders
- Reactive vs planned ratioEmergency callouts cost more
- Downtime costTenant impact, lost trading, penalties
- Benchmark savingsUS DOE FEMP: 8-12% over preventive
- Implementation and licence costDepends on BMS age and sensors needed
Implementation cost depends most on infrastructure. Buildings with a modern BMS mainly need integration and configuration. Buildings needing sensor retrofits cost far more. Payback is usually measured in years for a single building and faster across a portfolio where the platform cost is shared.
How do you implement predictive maintenance without disrupting operations?
- Weeks 1-4Audit current stateBMS inventory, maintenance spend by system, reactive vs planned ratio, contractor performance
- Months 2-3Use existing BMS dataEnable alerting, set up trend logging, connect BMS to the facilities platform
- Months 3-6Targeted sensorsAdd sensors only where the audit shows clear value
- Months 6-12Prediction and automationConfigure models per asset, automate work orders and dispatch
Audit. Find which systems drive reactive callouts and spend. In most portfolios, a minority of systems account for most of the cost. Target those.
Use what you have. Many buildings have BMS capability they do not use. Enable alerts, log history and connect the BMS to your facilities system before buying sensors.
Add sensors selectively. High-value targets: chiller vibration and temperature, AHU airflow and pressure, cooling tower water quality, lift motor condition. Often not worth it: individual VAV boxes without chronic problems, general plumbing.
Automate. Platforms include Siemens Senseye, IBM Maximo and Australian facilities platforms. Expect months before the models are useful, because they need history to learn your building.
What does not work yet?
- Small single buildings rarely justify sensor and software cost. Well-run preventive maintenance is often enough.
- Buildings without a BMS need significant capital to retrofit.
- Generic "any building" platforms underperform without tuning to your equipment.
- Instant prediction. Prediction works on gradual failure modes over weeks. Sudden faults need fault detection, which is different technology.
- Very old BMS may lack integration options, which means an upgrade or limited connectivity.
What skills does a facilities team need?
Facilities staff need training on new platforms, contractors need onboarding to mobile job management, and building managers need dashboards they trust. The teams that succeed treat AI as a first opinion. It flags the issue; an experienced technician decides whether a vibration signature is normal wear or imminent failure.
How Solve8 can help
We connect BMS, facilities and finance systems and build predictive maintenance and work order automation that runs on your infrastructure in Australia, with a coordinator approving each dispatch. Our consultants have worked on asset and operations data across major Australian industrial sites. See our field service and facilities work and system integration service, or talk to us.
Sources: US Department of Energy FEMP Operations and Maintenance Best Practices Guide; Siemens, The True Cost of Downtime 2024; vendor documentation for platforms named.
