Reliability and uptime
For the person who owns the availability number for the year, but not the crews, the budget lines or the work that produce it.
The same failures come back, and every one of them was closed properly.
Take the stoppages that cost you most this year and count how many were the second or third time that machine stopped for that reason. Then look at how those jobs were closed: completed, on time, by someone competent, signed off. Nothing in the closing of a job asks whether the cause was removed, so a repeat and a first occurrence are written the same way, and cost you the same way twice.
The Availability Gap
Your plant keeps a complete account of every failure after it has happened, priced to the minute. What it does not keep, anywhere, is where the next one will come from. The records sit at the level of machines; the money leaves at the level of the year.
Cost of Unplanned Downtime
You are already paying for unplanned downtime; it is just spread across budget lines that never say so. It is the overtime crew that exists because the work could not be planned. The buffer stock held because the failures cannot be predicted, not because the process needs it. The critical spare on a shelf, tying up cash, waiting for a machine nobody can name. And the crisis meeting of your most expensive people that forms early and dissolves by noon.
None of that appears in a line called unplanned downtime. It appears as overtime, as the cost of holding stock, as express freight, and as the senior engineer who spent Thursday in a meeting instead of on the problem he was hired for. It has been funded for years, and it will be funded next year unless something changes upstream of it.
Then there is the production itself, plus the restart. In a continuous process the restart is usually the bigger half: a kiln that has cooled, a lining that has set, a line that takes hours to bring back to rate. That half is booked as ramp-up rather than as loss, which is why nobody argues about it and nobody counts it.
Missed Warnings
The warning was there. It nearly always is. It sat in a repair note from March, written by someone who has since moved to another team. It sat in a reading nobody trended, because trending it would have meant knowing in advance that it mattered. And it sat in one fitter's memory of the last time this happened, which is the only one of the three that ever gets consulted, and only if he is on duty that day.
Count what each of those knows on its own. The note is a note. The reading is a reading. The memory is a story. The finding is in none of them; it is the relation between them, and a relation cannot be filed. Buy a bigger database of readings and you get more readings. You do not get the relation.
After the machine is down, a competent engineer assembles that relation in a few hours, at overtime rates. Before it goes down, nobody assembles it at all. Holding three unrelated documents in mind, for every machine, every day, is not a job a person can do. It is not that your people are failing at it. Nobody has ever done it anywhere.
Underneath that sits a second problem. Faults travel, so the reading you trend is often pointing at the wrong machine. A seizing roller shows up on the belt motor. A blocked chute shows up on the drive above it. One electrical supply problem shows up on every machine hanging off that panel. The strongest signal in the plant frequently comes from a healthy machine that happens to be best placed to feel somebody else's problem, and that is where the crew gets sent.
And there is a reason this number, of all your numbers, is the easiest to report without moving. Reliability is an outcome; everything that produces it is measured as activity. The planner is measured on the plan being met, the crew on jobs completed, engineering on projects delivered. All three can hit their target in a year when reliability did not move, and none of them is doing their job badly. The availability number does what you want, and the plant does not change. It rises in a quarter with a light production plan, when a machine is run below its rate, when a stoppage is booked to a category that does not count.
Unified Maintenance Record
Three joins, and none of them needs a purchase order to begin. It is not new information; it is the information you already hold, held together.
Pre-Failure Diagnosis
The machine is read from its motor, and the plant's scattered records are held as one account that can be reasoned through the way your best engineer does: backwards, from symptom to cause. Neither half is enough on its own, and most tools stop at the first.
Threads takes what the plant already writes down and holds it as one account of what is true at your plant, in the equipment names your maintenance system already uses. Web is the knowledge graph built out of Threads, and it is where the joining happens: the failure to the finding, the repair to what came after, the machine to the others like it. Spiders are the agents; each is good at one task, and they reason backwards from a symptom to a cause across the Web, saying which fact is missing and which document would close it when the record does not answer. Workflows are the path from where the plant stands today to the availability number it is measured on for the period; the work of every role is filed into it from the day it starts, your own procedure and its gates are how it is walked, and the spiders work together underneath to make it happen.
Every finding carries the evidence used and the check that would settle it, so a person can walk from the conclusion back to the sources without the author in the room.
Compounding Reliability
By repeats, per machine, against the availability number you are measured on for the period. A single correct diagnosis does not change how a machine is maintained or what it costs. A year of them, counted in one place against that number, does.
Work is counted from the day it is raised, not the day it is finished. A report that only counts finished work is describing a month that has already gone, and a budget that has already been spent.
When three settled jobs on one machine share the same signature, the system proposes it as a condition-based check (a reading to watch, instead of a date on a calendar) and as a step in the procedure. A person decides, and that is the only route by which a repeat breakdown becomes a standing check.
It is the direction every piece of work is asked about: out of breakdown, into preventive maintenance on a calendar, into maintenance based on the machine's condition, into problems designed out altogether. Each step to the right is a step out of the overtime line and the spares shelf.
Return-to-Service Authority
Every finding on this page is a proposal until a named person accepts it. The system can rank, explain and recommend; it cannot return a machine to service, and it never will.
Availability Commitments
Three things change for the person who owns the number.