ROI Case File No.501 'The Factory Where Stopped Machines Were Invisible'
The Factory Where Stopped Machines Were Invisible
Chapter One: The Stopped Time Was Never Visible
"We can see that the equipment has stopped. We just can't see for how long."
Kohei Kishida, Manufacturing Division Manager at TechnoCraft, opened the factory layout map as he spoke. Press machines, cutting machines, assembly lines—more than thirty pieces of equipment in operation. "Each machine has an operation light that turns red when it stops. But nobody records when it stopped and when it was restored."
"Who responds to equipment stoppages?" Claude asked.
"Three maintenance staff on the floor," Kishida replied. "When they spot a red light, they rush over, investigate the cause, and fix it. Then on to the next. Records go onto paper daily reports by hand, but nobody aggregates them afterward."
"So equipment stoppage data is piling up on paper," I confirmed.
"Exactly," Kishida said. "I've been saying we need DX for three years. But I don't know where to start. Equipment upgrades, software adoption, talent development—they all look necessary, and I can't prioritize them. The people on site are flat out handling incidents. Looking medium to long term, our in-house engineers are aging too, and knowledge transfer is also a problem."
"You can't prioritize because there's no decision axis," I replied. "Let's redesign this with ROI."
Chapter Two: ROI Asks—Return on Investment as the Decision Axis
"This case calls for ROI."
Claude wrote three letters on the whiteboard. R, O, I.
"ROI stands for Return on Investment, a metric for measuring the return rate against the investment," I explained. "Where ROAS (return on ad spend) is an indicator specific to advertising, ROI is a universal indicator usable for any investment decision. In a DX-investment situation where you cannot decide where to begin, ROI becomes the common yardstick. Once you can line up the investment efficiency of upgrades, software, and development side by side, the priority order becomes self-evident."
"First, let's measure the current cost," Gemini said, opening ROI Polygraph. He entered the equipment operation data Kishida had provided.
"The monthly equipment stoppage cost has come out," Gemini read aloud. "Lost production opportunity from equipment downtime averages 4.2 million yen per month—calculated from estimated downtime against hourly gross profit. On-site dispatch and root-cause investigation work by three maintenance staff totals 300 hours per month at 3,200 yen per hour, or 960,000 yen monthly. Because paper reports are not effectively aggregated, recurrence prevention isn't applied and the same root causes lead to repeated stoppages, costing about 700,000 yen monthly. Equipment upgrade requests from the floor are stalled due to lack of prioritization, an opportunity cost averaging 500,000 yen monthly. Total: 6.36 million yen per month. Annualized: roughly 76.3 million yen."
Kishida stared at the figures. "I had no idea equipment stoppage losses were stacking up this much."
"Now let's design with ROI," I continued.
[ROI Layer 1—List the Investment Candidates]
"First, we exhaustively list the DX investment candidates," Claude said. "'Operation visualization investment' attaches IoT sensors to equipment and automatically collects operating data. 'Predictive maintenance investment' analyzes that data to detect abnormalities in advance. 'Talent development investment' documents the maintenance team's tacit knowledge and transfers it to younger staff. 'Upgrade decision platform investment' uses numerical data to set the priority order of equipment upgrades. These four are the candidates."
"It looks like we should do all of them," Kishida confirmed.
"All of them, but the order matters," I replied. "Once ranked by ROI, the first investment becomes clear. Starting them all at once disperses resources, and nothing finishes properly."
[ROI Layer 2—Compute Each Investment's ROI]
"We'll test the four investment candidates with ROI Proposal Generator," Gemini continued.
"Operation visualization investment—initial cost 3.2 million yen (IoT sensor installation, data collection infrastructure, visualization dashboard). Monthly effects: 1.2 million yen from faster recovery via real-time downtime awareness, plus 500,000 yen from accumulated recurrence-prevention data. Total monthly: 1.7 million yen. Payback: about two months."
"Predictive maintenance investment—initial cost 4.8 million yen (anomaly detection algorithms, additional sensors, operational platform). Monthly effect: 1.8 million yen from reduced unplanned stoppages. Payback: about three months. However, this assumes operating data is already accumulated."
"Talent development investment—initial cost 2.4 million yen (training programs, knowledge documentation, OJT design). Monthly effects: 400,000 yen from improved junior response capabilities, 300,000 yen from veteran workload distribution. Total: 700,000 yen. Payback: about four months."
"Upgrade decision platform investment—initial cost 1.8 million yen (priority algorithm design, floor request management system). Monthly effect: 400,000 yen from resolving the upgrade request backlog. Payback: about five months. However, only by first having operating data can the priorities be quantified."
[ROI Layer 3—Decide the Investment Order]
"By ROI alone, operation visualization is the top priority, but there's another critical perspective," Claude continued. "Investment dependencies. Predictive maintenance and the upgrade decision platform cannot function without operating data. In other words, operation visualization is the foundation for the other three. The rational sequence: start with operation visualization, and once roughly three months of data is accumulated, run predictive maintenance and the upgrade decision platform in parallel. Talent development runs in parallel throughout."
Kishida took notes. "It's not just ROI on its own—the order only becomes clear once you build in the dependencies."
"ROI is a decision axis, not a decision-maker on its own," I replied. "Once you have the numbers, the structural discussion is what matters."
[ROI Layer 4—Test the Integrated Payback]
"Here is the consolidated estimate across the four investments," Gemini continued.
- Total initial cost: 12.2 million yen (sum of four investments)
- Monthly cost: 220,000 yen (combined ongoing IoT and predictive maintenance system fees)
- Total monthly effect: 4.8 million yen (sum of all four monthly effects)
- Net monthly effect: 4.58 million yen
- Payback period: 12.2 million yen ÷ 4.58 million yen ≈ 2.7 months
"Payback under three months," Gemini summarized. "Of the current annual loss of 76.3 million yen, approximately 55 million yen is structurally recoverable. The remainder—items like genuine equipment aging—requires separate investment judgment."
Kishida looked over the numbers. "Everything seemed equally urgent, but the moment you give it an order, it feels possible to move forward."
"ROI is the indicator that moves stalled decisions," I replied.
Chapter Three: An Investment Plan With an Order
"Here's the implementation plan," I said, standing at the whiteboard.
"Months 1–2: install IoT sensors, build the data collection infrastructure, launch the operation dashboard. Months 3–4: accumulate operating data, analyze maintenance response patterns. Month 5: begin predictive maintenance algorithm development; in parallel, build the quantification base for upgrade requests. Month 6: pilot predictive maintenance; launch the upgrade decision platform. Month 7 onward: run the talent development program in parallel throughout, document veterans' tacit knowledge, and reflect it in junior OJT."
"Only the talent development runs in parallel?" Kishida confirmed.
"Because development takes time," Claude replied. "Systems stand up in a few months, but talent development is a year-scale effort. Postpone it, and you'll have systems that no one can fully operate. We grow people in parallel."
Kishida closed the document. "I understand now why DX has been stalled for three years. We were trying to see the whole picture all at once."
Chapter Four: The Day Operation Lights Became Data
Nine months later, a report arrived from Kishida.
Three months after the operation visualization system launched, equipment downtime fell 32% versus the prior baseline. The moment a stoppage occurred, a notification reached the dashboard, and maintenance dispatch time was cut by an average of seven minutes. "The time we spent walking the floor checking operation lights disappeared entirely," Kishida wrote.
The predictive maintenance algorithm entered pilot operation in month 6. Over three months of running, roughly 40% of unplanned stoppages could be detected in advance. "From the combination of vibration and temperature data, we can identify bearing degradation before failure. That's fundamentally different from responding after a stop," the report said.
The most surprising change appeared in the team's mindset. With operating rates now visible numerically, the three maintenance staff began proposing improvements on their own. "Each person now checks how their own response affects the following week's graph. Data changed behavior," Kishida wrote.
The talent development program ran in parallel, and two junior staff reached a level where they could independently handle first-line response. Documentation of veteran maintenance knowledge reached about 40% completion. The remainder was folded into the upgrade process as an ongoing initiative. "Development never finishes, but the first step has been taken," the report said.
With upgrade priorities now numeric, the backlog was resolved. Of about 20 requests per month, those with the highest ROI were addressed first. "The voice from the floor asking 'when will my upgrade happen' has disappeared," Kishida wrote.
The final line of the report read: "Three years of stalled DX began moving the moment ROI gave it an order. The option of 'doing everything' was, in effect, the same as doing nothing—I see that now."
On the morning that the time spent walking the floor for red operation lights disappeared, a new factory had begun, the report said.
"The difficulty of DX investment lies in having too many options. Upgrades, software, development—all look necessary, and there's no way to know where to start. ROI asks for a common yardstick. Line up the numbers, and priority emerges. But ROI on its own isn't enough. Only by incorporating the dependencies between investments does the order become clear. In a factory where equipment stoppage data was piling up on paper, the day the operation light turned into a number, what had been stopped wasn't the equipment—it was the decision."
Related Files
Tools Used
- ROI Polygraph — Visualizing equipment stoppage losses, maintenance work hours, and stalled upgrade costs
- ROI Proposal Generator — Payback simulation across four DX investment areas
