Predictive Maintenance ROI: How Japanese Plants Justify It
You run maintenance or reliability at a Japanese-owned plant in Ohio, Chennai, or Rayong. A vendor has quoted sensors and software for predictive maintenance, and your plant manager likes the idea. The budget still has to be approved by the Japanese parent company through ringi, the written approval route, and the first question that comes back is simple: "How much will this save, and how many failures will it actually prevent?"
Predictive maintenance ROI starts with one number: how much you lose each time a machine stops without warning. Your past failure records give you that number. Multiply it by the stops the program can prevent, set it against the purchase and running costs, and you have a payback period. The method is simple, but the result depends heavily on which equipment you choose.
This guide checks the scale of downtime and maintenance costs against published surveys, sorts the benefits into four items you can price, and lists what belongs on the cost side. It then works through two fictional examples, shows how to pick equipment that pays back, and ends with a one-page summary you can attach to a ringi-sho (approval request) and a checklist of the records to pull first.
Contents
- Predictive Maintenance ROI Comes From Downtime You Avoid
- How Big Maintenance Budgets Are: Survey Data From Japan
- Four Benefits to Put a Number On
- What Goes on the Cost Side: More Than the Initial Quote
- Worked Example: A Line-Critical Drive vs. a Pump With a Standby
- Which Equipment Pays Back First
- Presenting the Case for Ringi Approval
- One-Page Summary to Attach to a Ringi-sho
- Where This Method Has Limits
- FAQ
Predictive Maintenance ROI Comes From Downtime You Avoid
When people think of the benefits of predictive maintenance, they usually think of lower inspection and parts costs. Those savings are real, but the larger amount is usually the production you keep by avoiding unplanned stops.
Siemens estimated these losses in its 2024 report "The True Cost of Downtime 2024." It puts the cost of unplanned downtime for the world's 500 largest companies (the Fortune Global 500) at about USD 1.4 trillion a year, or 11% of their revenue. The cost of one lost hour varies widely by sector: USD 2.3 million at a large automotive plant, and USD 36,000 in fast-moving consumer goods (FMCG).
The report also notes that for small and mid-sized manufacturers, an hour of downtime can reach USD 150,000 at the top end.
The same hour of downtime can cost very different amounts. That means the return on predictive maintenance also differs from one machine to the next.
Source: Siemens, "The True Cost of Downtime 2024." Estimates based mainly on surveys of large manufacturers.
This spread is where any ROI calculation should start. A machine that loses a large sum per hour and a machine that can switch to a standby unit will return completely different amounts from the same monitoring system.
The report lists the "hidden costs" of downtime: lost revenue from goods not produced, wages paid to staff who cannot work, labor for the people fixing the problem, premiums for rush-ordered replacement parts, and contractual penalties for late delivery. For smaller suppliers it adds a further risk: missing on-time, in-full (OTIF) delivery can cost them their status as a supplier.
If you want to tie these losses to equipment availability, use the loss structure behind overall equipment effectiveness (OEE), explained in OEE and the 7 Big Losses vs. Six Big Losses in TPM. Of those losses, predictive maintenance mainly reduces breakdown loss.
How Big Maintenance Budgets Are: Survey Data From Japan
Before pricing the benefits, it helps to know how much plants spend on maintenance in the first place. In Japan, a widely cited reference is the Maintenance Survey that the Japan Institute of Plant Maintenance (JIPM) has run since fiscal 1992. Approvers at a Japanese head office may know these figures, so they are useful context for your proposal.
In the fiscal 2022 survey, total maintenance cost averaged 3.38% of the value of manufactured goods shipped. Applied to all Japanese manufacturing, that gives an estimated JPY 11.17 trillion in maintenance spending. The ratio varies by industry: 4.38% for food, 3.96% for chemicals, and 2.95% for steel.
About a Quarter of Maintenance Spending Goes to Predictive Maintenance
The same survey splits maintenance cost into six categories. In the fiscal 2024 report, the actual fiscal 2023 shares were 28.6% for preventive maintenance, 25.8% for predictive maintenance, and 22.5% for breakdown maintenance.
Preventive maintenance, which covers statutory inspections and scheduled servicing, is the largest share. Emergency repairs after a failure still account for over a fifth.
Source: JIPM, "FY2024 Maintenance Survey Report" (Japanese). In this survey, predictive maintenance cost means inspections such as equipment diagnosis, plus repairs done in a planned shutdown after an abnormality is found.
Note the direction of the forecast: the predictive share drops by 2.1 points in fiscal 2024, while the preventive share rises by 2.4 points. The report itself says that predictive maintenance costs, expected to grow with IoT and AI, are trending down instead.
One reading is that the technology exists, but it has not yet changed how Japanese plants allocate maintenance money. At the same time, breakdown maintenance still takes more than 20% of the total.
Repairs done in a hurry after a failure tend to cost more because of rush parts orders and weekend or night work, and they come with lost production on top. When you build an ROI case, this breakdown share is the first cost you want to cut.
How Much Changing Maintenance Strategy Can Save
For the cost difference between maintenance strategies, a common reference is the "O&M Best Practices Guide" from the U.S. Department of Energy's Federal Energy Management Program (FEMP). It mainly covers building and facility equipment, but it gives a sense of how far apart the strategies are.
The savings depend on your starting point. Plants that mostly fix things after they break have the most room to gain.
Source: U.S. DOE FEMP / PNNL, "O&M Best Practices Guide." These figures cover maintenance cost only and do not include avoided downtime.
The takeaway: at a plant where preventive maintenance already works well, maintenance cost savings alone are hard to stretch far enough to pay for a predictive program. To show a return, you have to count avoided downtime as well as maintenance savings.
For how time-based and condition-based maintenance differ and when each fits, see TBM vs CBM: Time-Based vs Condition-Based Maintenance.
Four Benefits to Put a Number On
Split the benefits of predictive maintenance into the four items below and you are less likely to miss any. In a ringi-sho, the same four items also show approvers exactly where the money comes from.
Avoided downtime (1) is usually the largest. Items 2 to 4 depend on the equipment.
The formulas show the logic only. Replace them with figures from your own maintenance records and cost data.
Price Avoided Downtime at Contribution Margin, Not Revenue
If you price downtime at revenue, you overstate the benefit. While a line is down, you also stop spending on variable costs such as materials and power, so what you actually lose is revenue minus variable cost: the contribution margin. Your accounting team can often give you this figure by product or by line.
Also check whether you can make up the lost output later. If demand allows it and you can recover with overtime or weekend shifts, the loss is no longer the margin but the cost of that recovery: overtime labor and other premiums.
On a line running at full capacity, or for a product with tight delivery dates, any output you cannot recover is lost contribution margin.
Price Repair Savings as Emergency vs. Planned Cost
The same bearing replacement costs more when the machine stops in the middle of the night than when you fit it into next week's planned shutdown. In an emergency you order parts by express, pay contractors a premium, and call in maintenance staff on nights or weekends.
Sort your past repair work orders into "unplanned" and "planned," and the cost gap at your own plant becomes visible. If your records do not separate them, start by picking two or three typical failures and asking what each one cost at the time.
Longer Replacement Intervals and Avoided Secondary Damage
On equipment where parts are replaced on a fixed schedule, some parts come out while they still have life left. If you can time replacement by condition, the longer interval saves both parts and labor. This does not apply to parts whose inspection or replacement intervals are set by law or regulation.
Secondary damage is hard to price, yet it can be the largest item per event. Catch a bearing problem early and you replace only the bearing; run it until it seizes and the shaft and housing are damaged too.
Keep uncertain amounts like this out of the main calculation and show them separately as upside. That keeps the core numbers credible to approvers.
What Goes on the Cost Side: More Than the Initial Quote
A common mistake is to judge the investment by the sensor and software quote alone. Predictive maintenance pays off only when someone reviews the data, makes a call, and turns it into maintenance work. Compare the program's full cost, including running costs and the staff time it requires.
List not only what appears on the quote, but also the yearly running costs and staff time.
Actual amounts vary widely with the number of machines, monitoring points, network type, and analysis method. Confirm them with quotes from more than one vendor.
Staff time is the item most often missed. Every alert means a trip to the machine, and checks that find nothing wrong are most frequent right after startup. Plan for extra checking work during the first few months, while you tune the alarm thresholds, so nobody is surprised by how much effort the system takes.
On the other hand, if you have people in-house who can take and interpret vibration measurements, you can rely less on outside services and keep costs down.
If you are adding sensors to older Japanese machines that have no network connection, the hardware and installation choices are covered in Retrofit Machine Monitoring on Older Japanese Machines.
Worked Example: A Line-Critical Drive vs. a Pump With a Standby
Now put the items together. Every number below is a fictional example to show the method. None of them come from a real company or machine. Amounts are in Japanese yen; the method works the same in any currency.
Example A: No Standby, and a Stop Halts the Whole Line
Take the main drive motor and gearbox on an extrusion line. There is no standby unit, so when it stops, the whole line stops. Three years of maintenance records show an average of two unplanned stops a year, each lasting 10 hours including the wait for parts.
The line's contribution margin is JPY 250,000 per hour, and because it runs at full capacity, lost output cannot be recovered. An emergency repair costs JPY 1.5 million; the same repair done in a planned shutdown costs JPY 600,000.
Assume predictive maintenance turns one of the two yearly stops into a planned shutdown, and the planned work is done on a non-production day with no impact on output.
Subtract running costs from the yearly benefit to get the net benefit, then see how many years it takes to recover the initial cost.
All figures are fictional. Interest and tax are not included. For a more rigorous multi-year comparison, calculate net present value (NPV) with a discount rate.
In this example, the JPY 4 million initial cost is recovered in about 1.7 years, and the program is JPY 3.2 million ahead after three years. Put another way, preventing about two unplanned stops in three years pays for it.
"How many stops do we need to prevent to break even?" is a framing senior managers grasp quickly, and it works well in a ringi-sho.
Example B: A Cooling Water Pump With a Standby Unit
Now put the same system on a cooling water pump that has a standby. When it fails, operators switch to the standby and production keeps running, so avoided downtime is close to zero. The only benefits left are the repair cost difference and longer replacement intervals.
Say the pump fails once a year, an emergency repair of JPY 500,000 becomes a planned repair of JPY 300,000, and a longer bearing replacement interval saves JPY 50,000 a year. The yearly benefit is JPY 250,000. If running costs for this one pump are JPY 300,000 a year, running costs alone exceed the benefit, and the initial cost is never recovered.
The only difference between the two examples is where the system went, not how good the sensors or software were. In practice, the ROI calculation is less a yes-or-no test for the whole program and more a way to decide where to start.
Which Equipment Pays Back First
As the examples show, the return on predictive maintenance depends mostly on the target equipment. The usual approach is not to cover the whole plant at once, but to start with equipment that pays back quickly, build a track record, and then expand.
Two questions sort equipment: how much it costs you when it stops, and whether it shows warning signs before it fails.
Start with equipment that costs a lot when it stops and shows warning signs before it fails.
Vertical: impact of a stop (higher = larger) / Horizontal: warning signs before failure (right = clearer)
Whether warning signs are clear depends on whether the time from the start of deterioration to failure (the P-F interval) is much longer than your inspection or monitoring interval.
Conditions for Top-Priority Equipment
Put equipment that meets several of these conditions first. The more conditions it meets, the more likely it is to pay back quickly, as in Example A.
- No standby unit, so a stop halts the whole line or downstream processes
- At a bottleneck process, with little room to make up lost output
- Has had unplanned stops in roughly the past three years
- Uses custom parts that take days to arrive
- Mostly fails in ways that give warning, such as bearing wear or rising vibration
Equipment that tends to fail suddenly, without warning, does not benefit much: predictive maintenance cannot prevent failures it cannot see coming. The P-F curve is the clearest way to think about how early warning signs appear; the TBM vs CBM guide also explains it.
In Process Plants, Whole-Process Monitoring Is Another Option
In process industries such as chemicals and oil refining, another approach is to watch process data such as temperature and pressure for early signs of trouble across a whole process, rather than monitoring each rotating machine. Even so, monitoring a single rotating machine can be worth a lot in these plants.
Fuji Electric describes a case in which it fitted IoT sensors to fan motors at a chemical plant so their condition could be diagnosed. By its own estimate, avoiding failure stops is worth JPY 30–70 million a year in prevented losses.
Presenting the Case for Ringi Approval
Once the numbers are done, shape them into something approvers can review. In a ringi, several people read the same proposal in turn, and each tends to ask about precedent, track record, risk, and alternatives (see Ringi in Japan: Why Approvals Are Slow and How to Help). Because the benefit depends on how many stops you prevent, a single number invites the question "Can you really prevent that many?"
Show a Range With Three Scenarios
A good answer is to show three scenarios that vary the share of stops you prevent. For the Example A equipment, if 30%, 50%, or 70% of unplanned stops become planned shutdowns, the payback periods look like this:
Showing that even the conservative case pays back within four years makes the decision easier.
Fictional example using the Example A conditions: 2 unplanned stops a year, 10 hours each, contribution margin JPY 250,000/hour, repair cost difference JPY 900,000 per stop, initial cost JPY 4M, running cost JPY 1M/year.
If even the conservative case pays back well within the equipment's remaining service life, the investment is sound. If only the optimistic case pays back, either rethink the target equipment or run a small trial first to measure how many stops you actually prevent before expanding.
Record Your Baseline Before You Start
To show results later, you need "before" numbers. For the target equipment, record unplanned downtime hours and count, mean time between failures (MTBF), the number and cost of emergency repairs, and the share of maintenance cost spent on breakdown repairs.
The benefit of predictive maintenance is a failure that did not happen, which you cannot show without a baseline. One to three years of records before the program starts become your yardstick for measuring results. If your records are thin, start by standardizing how you record failures at the same time you install the system.
Three Reasons Programs Fail to Pay Back
Some programs never deliver the return the calculation promised. The usual causes are these three:
- Stuck at the pilot: The pilot proves the idea on a few machines, but the program never expands or gets a working process around it.
- No one acts on alerts: Nobody owns the check or the decision to schedule a shutdown, so notifications get ignored.
- No decision criteria: Data piles up, but there is no threshold for when to replace a part, so the team still replaces on a fixed schedule.
These are less technology problems than workflow problems: the system was never built into everyday maintenance work. That gap likely helps explain why the predictive share of maintenance cost in the JIPM survey is not growing.
One-Page Summary to Attach to a Ringi-sho
Approvers at a Japanese head office may read your proposal in English, in translation, or through a Japanese colleague. A one-page table with the assumptions spelled out travels better than a long narrative. Fill in this template with your own figures.
| Item | What to enter | Where the number comes from |
|---|---|---|
| Target equipment | Machine, line, and why it was chosen (no standby, bottleneck, past stops) | Asset list, Figure 7 criteria |
| Baseline | Unplanned stops per year, hours per stop, MTBF, emergency repair count and cost | Maintenance records and work orders, past 1–3 years |
| Loss per hour of downtime | Contribution margin per hour, and whether lost output can be recovered | Accounting / production control |
| Benefits (main case) | Avoided downtime + repair cost difference (+ longer replacement intervals if you can estimate them reliably) | Figure 4 formulas |
| Upside (shown separately) | Uncertain items: secondary damage, quality, safety | Past failure cases |
| Costs | Initial cost; running cost per year; staff hours per month | Vendor quotes (two or more), internal estimate |
| Scenarios | Payback at 30% / 50% / 70% of stops prevented; stops needed to break even | Your calculation (Figures 6 and 8) |
| Precedent and track record | Similar installations at other group plants or by the vendor at Japanese customers | Vendor references, sister plants |
| Risks and stop conditions | False alarms, staff time, data security; what result ends or expands the pilot | Your plan |
| Owner and process | Who reviews alerts, who schedules shutdowns, how results are reported | Maintenance organization |
Checklist: Records to Pull Before You Calculate
- Unplanned stops on the target equipment over the past one to three years: date, duration, cause
- Repair work orders split into emergency and planned, with parts, labor, and contractor costs
- Contribution margin per hour for the affected line, confirmed with accounting
- Whether lost output can be recovered, and what recovery costs (overtime, weekend shifts)
- Lead times for critical spare parts, especially custom or imported parts
- Parts on a fixed replacement schedule, and whether any interval is set by law or regulation
- Running costs from at least two vendors, including how fees change as you add monitoring points
- Who will review alerts, and how many hours a month you expect it to take
Where This Method Has Limits
- The cost and survey figures above come from specific reports with their own scope. The Siemens figures are based mainly on large manufacturers, the FEMP figures mainly on facility equipment, and the JIPM figures on Japanese plants. Use them for scale, not as inputs to your own calculation.
- How a ringi is routed, which amounts need which approver, and what format the head office expects all vary by company. Ask your Japanese counterpart which template and level of detail they want.
- Statutory inspections and legally required replacement intervals stay in place regardless of what condition data shows. Check local rules and your company's own standards.
- Some companies require NPV or internal rate of return (IRR) instead of a simple payback period. Confirm the required method with your finance team before you build the case.
FAQ
How do you calculate predictive maintenance ROI?
Add up the yearly benefits (avoided downtime priced at contribution margin, plus the difference between emergency and planned repair costs), subtract yearly running costs, and divide the initial cost by the net result to get the payback period. Add longer replacement intervals where you can estimate them reliably, and show uncertain items such as avoided secondary damage separately as upside.
Should downtime cost be based on revenue or profit?
Use contribution margin (revenue minus variable costs), not revenue. While a line is down you also stop spending on materials and power. If you can recover the lost output later, use the cost of recovery, such as overtime, instead.
Can maintenance cost savings alone justify predictive maintenance?
At a plant that already runs a good preventive program, it is difficult. The FEMP guide estimates 8–12% maintenance savings when you add predictive to preventive maintenance. To show a return, include avoided downtime as well.
Which equipment should get predictive maintenance first?
Equipment with no standby, at a bottleneck, with a history of unplanned stops, and with failure modes that give warning, such as bearing wear or rising vibration. Equipment with a standby unit often does not pay back on its own.
How do I present a predictive maintenance budget to a Japanese head office?
Use a one-page summary with stated assumptions, three scenarios for stops prevented, the number of stops needed to break even, a baseline from your records, and precedent from similar installations. Ask your Japanese counterpart which ringi format they use.
Sources
- Downtime cost estimates: Siemens, "The True Cost of Downtime 2024"
- Savings by maintenance strategy: PNNL, "O&M Best Practices: Maintenance Approaches" (U.S. DOE FEMP, "O&M Best Practices Guide")
- Maintenance cost breakdown: JIPM, "FY2024 Maintenance Survey Report" (September 2025, Japanese)
- Maintenance cost ratio and national estimate: JIPM, "FY2022 Maintenance Survey Report, Summary" (September 2023, Japanese)
- Chemical plant fan case: Fuji Electric, "Cost reduction through predictive maintenance and IoT use cases" (Japanese)
Original article in JapaneseThis guide is adapted from a MONOSCOPE article written for readers in Japan: 予知保全の費用対効果の出し方｜停止損失と保全費で見る投資判断
About the editor
MONOSCOPE EditorFOUNDER & EDITORSpent about ten years at a plant equipment maker and a plant engineering company, proposing, designing, and commissioning equipment, then moved into web marketing. Now a web consultant for B2B manufacturers, helping them explain their products and technology. Licensed in Japan as an Energy Manager and a Third-Class Chief Electrical Engineer.
Has run an engineering blog since 2017 (500+ articles) and a Japanese YouTube channel with about 20,000 subscribers. Each English guide is written from the Japanese shop-floor point of view and checked against the cited standards.
