Mean Time to Repair is one of the most honest metrics in manufacturing maintenance. It does not care about intentions or effort. It measures one thing: how long it takes your team to get a machine back online after it fails. For most US plants, that number is higher than it should be — and the gap is not where most managers think it is.
What MTTR Actually Measures
MTTR is calculated by dividing total downtime by the number of failures in a given period. If your plant had 12 unplanned stoppages last month and they totalled 36 hours of downtime, your MTTR is 3 hours. Simple enough. But the number hides a lot.
Most maintenance managers assume the bulk of that time is spent on the actual repair — replacing a bearing, rewiring a drive, swapping a sensor. In practice, research from the Society for Maintenance and Reliability Professionals shows that diagnosis accounts for 40 to 60 percent of total repair time on complex equipment failures. The wrench time is often the shortest part of the job.
That is the insight that changes how you approach MTTR reduction. If you want to cut your average from 3 hours to under 90 minutes, you do not start by training technicians to work faster. You start by cutting the time it takes to identify the fault.
Where the Time Goes on the Plant Floor
A typical unplanned failure on a mid-complexity piece of equipment follows a predictable pattern. The machine stops or throws an alarm. A technician is dispatched. They arrive, assess the situation, and begin troubleshooting. They may consult a manual, call a more experienced colleague, or work through a mental checklist built from years on that specific machine.
On equipment the technician knows well, this goes quickly. On equipment that is less familiar — or on a fault pattern they have not seen before — it can take an hour or more just to narrow down the cause. Then there is parts availability, which adds another variable. And if the fault requires a specialist or an OEM call, you are looking at hours or days.
The plants that have made the biggest reductions in MTTR have attacked this problem at the diagnosis stage. They have found ways to get technicians to the right answer faster, regardless of their individual experience level with that specific machine.
The Experience Problem Is Getting Worse
One of the structural challenges in US manufacturing right now is that the most experienced technicians are retiring. The people who knew every quirk of every machine on the floor — who could diagnose a fault by sound alone — are leaving, and their knowledge is leaving with them.
The Deloitte and Manufacturing Institute skills gap study projects 2.1 million unfilled manufacturing jobs by 2030. A significant portion of those are in maintenance. The plants that are managing this transition well are the ones that have found ways to make institutional knowledge accessible to less experienced technicians in real time.
This is not about replacing experienced people. It is about making sure that when a newer technician encounters a fault pattern that a veteran would recognise immediately, they have access to that same pattern recognition without having to track down the veteran or wait for a shift change.
Three Things That Actually Move MTTR
Plants that have reduced MTTR by 40 percent or more in the last two years have generally done it through a combination of three things.
First, they have made their maintenance documentation searchable and usable in the field. This sounds basic, but most plants still have manuals in binders, PDFs on shared drives that nobody can find, and OEM documentation that exists somewhere but is not accessible from the floor. When a technician can ask a question in plain English and get a relevant answer from the machine's own documentation in under a minute, diagnosis time drops sharply.
Second, they have connected fault history to the diagnosis process. Every machine has a pattern of failures. The same fault codes tend to appear for the same underlying reasons. Plants that have made their work order history searchable and connected it to current fault diagnosis are giving technicians a significant advantage. Instead of starting from scratch on every failure, they can see what fixed the same symptom the last three times it appeared.
Third, they have reduced the friction in getting to the right answer. This means fewer phone calls, fewer manual lookups, fewer trips back to the maintenance office to consult a colleague. The goal is to get the technician to a confident diagnosis while they are standing in front of the machine.
The Role of AI in MTTR Reduction
The reason AI has become relevant to this problem is that it can do something that static documentation and spreadsheets cannot: it can synthesise information from multiple sources simultaneously and surface the most relevant answer for a specific fault in a specific context.
When a technician describes a fault — the symptoms, the error codes, the conditions under which it occurred — an AI system trained on that machine's documentation and work order history can cross-reference all of that in seconds and return a ranked list of probable causes with the supporting evidence. That is the kind of pattern matching that used to require a senior technician with years of experience on that specific machine.
The plants seeing the biggest MTTR reductions are not using AI to replace their technicians. They are using it to make every technician on the floor as effective as their best one. For a deeper look at how this connects to overall equipment effectiveness, see our guide on predictive maintenance for manufacturing.
Measuring Progress Accurately
If you are going to reduce MTTR, you need to measure it accurately first. Many plants track downtime in broad categories that do not distinguish between diagnosis time, parts wait time, and actual repair time. Without that breakdown, you cannot tell which part of the process you are improving.
Start by breaking your MTTR into components. Track how long it takes from fault detection to diagnosis confirmation. Track how long from diagnosis to parts in hand. Track how long the actual repair takes. Once you have that breakdown, you will know exactly where to focus.
For most plants, the diagnosis window is the biggest opportunity. And it is the one that technology can address most directly, without requiring changes to staffing levels, parts inventory, or repair procedures. Our post on machine downtime tracking covers how to set up that measurement framework in practice.
A Realistic Target for Your Plant
What is achievable? Plants that have focused specifically on the diagnosis window and given technicians better tools for fault identification have reported reductions from an average MTTR of 45 minutes to under 4 minutes for common fault patterns. For complex or novel faults, the improvement is less dramatic but still significant — typically a 40 to 60 percent reduction in diagnosis time.
The key is not to set a single target for all fault types. Simple, recurring faults should be diagnosable in minutes. Complex, first-time faults will always take longer. The goal is to make sure that the time spent is on genuine problem-solving, not on searching for information that already exists somewhere in your documentation.
For plant managers building the business case for this kind of investment, the ROI calculation is straightforward. Take your current average MTTR, multiply by your cost per hour of downtime, and model what a 50 percent reduction in diagnosis time would save over a year. For most plants running more than two or three unplanned stoppages per week, the number is significant enough to justify the conversation.
Ready to put this into practice?