Why
lines stop.

Why lines stop.

A field guide to the hours your line loses, and how you win them back.

Contact: team@fabrico.io · fabrico.io

Contents

Your worksheets, by page

The guide on one page.

Your OEE number is what is left after three kinds of loss. Here is one example shift, in minutes: where its 480 planned minutes went. In chapter 4 you measure the same four parts on your own line.

Example values: 480 planned minutes, 60 minutes of stops, 380 parts made at 1.0 minute a part, 361 of them good. Chapter 4 works through it.

What a stopped hour costs

Your line stops at 10:40. A technician walks over, finds a jammed sensor and cleans it. By 11:10 you are running again. The maintenance log gets one line: thirty minutes, one sensor cleaned, no parts. On paper, it was a cheap stop.

It was not. For thirty minutes your line made nothing you could sell. Six of your operators waited. The first parts after the restart went into the scrap bin. Thursday's shipment was already tight, so someone booked a Saturday shift. None of that is in the maintenance log.

In this chapter you put a number on one stopped hour on your own line. It takes you about twenty minutes and one sheet of paper.

The bill you see, and the one nobody adds up

A stop causes seven kinds of cost. Your maintenance log sees two of them.

You see the first two because they have their own budget line. The other five are spread over payroll, quality, freight and sales. Each looks small where it lands. Nobody sees the sum.

Five terms price your hour

You do not need to guess. Someone in your plant already has each number. Ask the right person and write it down.

Add the five and you have the cost of one hour on that line. Leave repair labour and parts out: they are already in your maintenance budget, and you would count them twice.

One line, worked through to the euro

Here is one line over twelve months. It lost 150 hours to 180 unplanned stops.

One stopped hour on this line costs €2,880. The year costs €2,880 × 150 = €432,000. Your log sees only the part below: the part you do not see is eighteen times bigger.

Example values, not typical ones. Log: repair labour and parts, €24,000 ÷ 150 hours = €160 an hour. Use your own numbers.

Hours leak out three ways, not one

Breakdowns are the stops you remember. They are only one of three ways your line loses hours:

• Stopped. Breakdowns, waiting for material, changeovers that run long.
• Running slow. Short stops and slow cycles.
• Making scrap. Hours spent on parts you throw away or rework.

Take one 480-minute shift. A 40-minute breakdown goes into the log. Sixteen short stops of two minutes each, 32 minutes in total, do not: each is cleared in seconds, so they just look like slow running. Price the slow and scrap hours like downtime. They cost you the same margin.

Example values.

Four traps your controller will catch

• Revenue instead of margin. Use price minus variable cost. Material you did not use is not a loss.
• Margin and overtime for the same parts. If overtime made up the output, count the overtime, not the margin.
• Every line as the bottleneck. A line with spare capacity can catch up. Its lost hour costs labour and scrap, not always margin.
• Idle labour that was not idle. If your operators moved to other work, do not count them.

A fifth trap is the most common: lost hours from memory. You remember the long breakdowns and forget the short stops. Use your stop log, and say so.

Rank your lines by what they really cost

Line C has the dearest hour. Line B loses the most hours. Line A leads on neither, and it ranks first. That is where you start.

Example rows.

One page for finance

Your finance team does not need your worksheets. It needs one page: every line with its lost hours, its cost an hour and its cost a year, your three biggest losses in hours, and one ask. Keep it honest: the same twelve months and the same source of lost hours for every line, every figure with the name of the person who gave it, and one ask in one sentence: fund this, for this amount, to attack this loss, which costs this much a year.

The same, run from the machines

The weak number on this page is your lost hours. Fabrico takes them from the machines: every stop with its reason, per line and shift, through PLC signals, IoT sensors or cameras with computer vision. Your operator picks the reason at the line, during the stop.

Request a demo

Where plants really stand

Sooner or later someone in your plant says it: world class is 85%, we are at 60%, so we are bad. It is the wrong conclusion from the wrong comparison. This chapter shows you what the 85% means and what plants that measure honestly find.

Where 85% comes from

The 85% figure goes back to Seiichi Nakajima's work on total productive maintenance in the 1980s. It was set as world class: the level of the best plants. It was never the level of a typical plant, and it was never a first-year target for a line that has just started to measure.

So if your plant is at 60%, you are not failing a standard. You are somewhere on a long scale that very few lines reach the top of. And if you see 85% after a few months of measuring, check the definition first. Chapter 3 shows you how a number gets there without anyone lying.

Source: Nakajima, S. (1988). Introduction to TPM: Total Productive Maintenance. Productivity Press. The usual source given for the 85% world class figure.

What plants find when they measure

Two published studies give you a sense of the range. Both come from Chalmers University in Sweden, from different decades with different methods. Read them as reference points, not as a league table.

Sources: Ljungberg, Ö. (1998). International Journal of Operations & Production Management, 18(5), 495 to 507. Hedman, R., Subramaniyan, M., Almström, P. (2016). Procedia CIRP, 57, 128 to 133.

Read the second study twice

In the study of 884 machines, almost half of the recorded losses could not be classified, because the loss categories were missing or poorly described. Even with automatic measurement, half the lost time had no usable reason. A number without reasons tells you how much time you lose. It tells you nothing about where to start. That is why chapter 3 spends as long on reason codes as on the formula.

Source: Hedman, R., Subramaniyan, M., Almström, P. (2016). Analysis of critical factors for automatic measurement of OEE. Procedia CIRP, 57, 128 to 133.

If your number does not match

If your line reports far above these ranges, check your definition before you celebrate. Check these four first:

• Changeovers booked as planned stops, so they never count as a loss.
• A "standard" cycle time lowered after a bad month, so speed losses vanish.
• Short stops that never reach a manual log.
• Startup scrap and rework left out of the reject count.

If your line reports far below, check the shift calendar. A line measured against 24 hours when it is only staffed for 16 looks worse than it runs.

A sensible first target

Do not set an OEE target in your first year. Set a target in hours. Take the top three losses from chapter 6 and aim to win back part of their hours on one line in 90 days, on the same definition before and after. OEE will move as a result, and nobody can say the gain came from a changed definition.

Compare your line with itself first: this month against its baseline, its best shift against its worst. The gap between your best and worst shift, on the same machines with the same products, is the most convincing target you will find. Your plant has already proven it can run that way.

Where your line stands

The same, run from the machines

Comparing your line with itself only works if both sides use the same definition and the same data. Fabrico calculates availability, performance and quality per line and shift from the machines, with every stop and a reason picked at the line. Your best shift, your worst shift and your baseline come from one source.

Request a demo

Measuring honestly

If you measure OEE, you may have a number you do not quite believe. Nobody lies to you. Every small choice in how your number is built just leans the same way: toward a better number. A changeover booked as a planned stop. A speed target lowered after a bad week. Warm-up scrap binned before anyone counts it. Together, these choices make your line look better than it runs, and they hide the losses you most need to fix.

You do not fix OEE. You fix the six losses under it.

OEE is availability × performance × quality. You cannot "improve availability" on a Tuesday morning. You can act on a specific loss. The six big losses, from Nakajima's work, give every lost minute on your line one place to sit.

The six big losses: Nakajima, S. (1988). Introduction to TPM. Productivity Press.

One rule sits above all six: you capture losses from your machine, not from memory. A manual log keeps the long stops and loses the short ones, so your small losses disappear first.

The six, one at a time

The 18-stop shift is an example.

Five ways your OEE number gets inflated

Read your own OEE definition against this list. Every yes means part of your number is not real.

None of these needs bad intent. Each is a default somebody chose once and nobody revisited. Write your definition down, on one page, and review it with the people who build your number.

Reason codes that stay clean

The six losses only help if every stop carries a reason, and reasons only help if your list stays clean. Six rules:

• Five to eight codes per line. Give your operators a long list and they pick the first one.
• Words your operators use. "Feeder empty", not "material flow deviation".
• The machine finds the stop, a person adds why. Start, end and length never come from memory.
• One "Other", read every week. Re-code it, or add the code that is missing.
• A review every month. Merge the codes nobody uses, split the one everyone uses.
• Every code maps to one of the six losses, so your Pareto chart rolls up to the loss tree.

The six-losses map for your line

The same, run from the machines

Measuring honestly is easier when your raw data does not depend on memory. Fabrico counts every stop from the machine, through PLC signals, IoT sensors or cameras with computer vision, and your operator picks the reason at the line while the stop is happening. Downtime shows by cause, line, machine and shift, and a breakdown can become a maintenance work card in the same system.

Request a demo

Days 1 to 7: a baseline on one line

Every improvement starts with one question: better than what? If your starting number is soft, every later number is an argument. In this chapter you get a baseline for one line in one week that nobody in your plant will argue with, on paper if that is all you have.

Pick one line. Not your best and not your worst: the one whose losses your people talk about most. You learn more from one line measured properly than from ten measured roughly.

Five inputs every OEE number is built from

Get these five right in week one and many OEE arguments never start. Write each one down, with a name next to it.

The one you will be tempted to skip is the ideal cycle time. If you take it from the "standard" rate your line was set to after a bad month, performance looks fine for ever. Chapter 3 called this the second way an OEE number gets inflated.

The formula, and the shift that settles arguments

OEE is availability × performance × quality. Availability is run time ÷ planned time. Performance is ideal cycle time × total count ÷ run time. Quality is good count ÷ total count. Here is one shift, worked through.

Example values with round numbers, not a typical value.

Here is a cross-check that never lies: OEE also equals good count × ideal cycle time ÷ planned time. Here, 361 × 1.0 ÷ 480 = 75.2%. If your two answers differ, one of your definitions is wrong. Better to find out in week one than in week ten.

As a waterfall, your losses become visible: 12.5 points to stops, 8.3 to speed, 4.0 to quality.

Your week, day by day

The trap in a paper week

A paper tick sheet gives you a good baseline for availability. It will understate your short stops, because a stop of fifteen seconds often goes unmarked, and it then shows up as slow running. That is fine for a first week, as long as you know it. Write one sentence under your baseline: "short stops under one minute are not in this count". Chapter 5 is about catching them.

Two sheets for your week

The same week, from your machines

On paper, your week costs your operators a few minutes every hour, and it still misses the short stops. That is the trade you make for starting this week without buying anything. When you are ready to stop making that trade, this is what changes.

The same, run from the machines

Fabrico records the machine states and counts for every line and shift from the machines, and your operator picks the reason at the line while the stop is happening. Availability, performance, quality and OEE come per line and shift in real time, on one definition. Your day 7 sums are there every morning, without a calculator.

Request a demo

The stops paper never sees

Ask your shift lead how often your line stopped yesterday and you hear about the long stops: the breakdown at ten, the changeover after lunch. Ask your machine and you get a different answer, because many of its stops were too short to write down. A jam cleared in four seconds. A sensor wiped. A part pushed back into its guide. Each one is nothing. Together they are often one of the biggest losses on your line, and on paper they leave no trace.

What a four-second stop costs

A filling machine stops for four seconds when a bottle tips over. Your operator stands it up and your line runs again. It happens about once every two minutes.

Example values, not typical ones.

A full shift every week, lost in pieces of four seconds, and not one of them is in your shift log. Paper misses it for a simple reason: writing down a four-second stop takes longer than the stop. So shift logs use a threshold of a few minutes, and everything under it is cleared and forgotten. Nobody is hiding anything. The tool cannot see that small.

Slow running makes no noise

A stopped line is loud. A slow line is quiet. You walk past a machine that runs at 54 parts a minute instead of 60, because it is running.

Example values: ideal rate 60 parts a minute, actual 54.

Your line never stopped, so no stop log shows anything. And if the "standard" rate in your system was lowered to 54 after a bad month, performance reads 100% while your line gives away 42 minutes a shift.

Lost speed has causes. A line is turned down after a jam so the jam does not come back. A tool wears. A material runs badly. A new operator plays safe. You can fix each one, but only once you see the gap.

Find the gap on paper first

Before you install anything, one shift and a calculator show you whether there is a gap worth hunting:

Example: 420 minutes at 60 parts a minute is 25,200 parts. Your log shows 35 minutes of stops, so 2,100 parts. Your total count is 21,000. That leaves 2,100 parts: 35 minutes that nobody wrote down.

Example values.

The stopwatch hour

To split your gap into short stops and slow running, watch your machine. One hour per shift, with a stopwatch, for one week. Count every stop, its length in seconds and its cause, and count the parts in the same hour. The stops give you the short stop share. The rest of the gap is speed. On three shifts, five days a week, that is fifteen hours of watching. Use them to name your first three causes. The first one is often fixed at your machine, not in a meeting.

Three ways to count without your operator

The paper hunt proves the gap and finds your first causes. It cannot run every shift. For that, the count has to come from your machine, and you have three ways to get it.

PLC signals are the cleanest source where they exist, but you need the people who look after your machine's control, because they know which signal means what. A sensor added from the outside changes nothing in the control. A camera fits where neither can see. Many of your lines will end up with a mix: the method follows your machine, not your plant.

Whichever you pick, check its first week against your stopwatch hour. If the two counts disagree, trust neither until you know why.

The question your crew will ask

The first question on your floor is often not technical. It is "is it filming me?", and it deserves an answer before the camera goes up, not after.

What it points at

The process area: the parts, the conveyor, the station.

What it records

Video, recorded continuously in short segments linked to what your line was doing, so you can review a stop next to its machine data.

Where it is kept

Fabrico runs on AWS in an EU region. The video itself can be kept in a private cloud for your plant alone, or on your own server.

Before installation day, ask your vendor two more questions in writing: how long is the footage kept, and who in your plant can see it? Give your crew all the answers in the same words, and let them see what the screen shows.

Then put one person on the floor who can answer the question when it comes up again. It will come up again, usually on the night shift.

What to say at the line, in three sentences

Your short stop hunt

The same, run from the machines

Fabrico counts every stop from the machine, through PLC signals, IoT sensors or cameras with computer vision, so your short stops are counted, not estimated. Slow cycles count as performance loss, not downtime. With hardware, installing and connecting the devices takes about a month, and the full rollout takes three to four months.

Request a demo

Choosing the top three losses

After your baseline week you have a list of every stop and its reason. You will be tempted to attack all of it. Your programme can die right there: ten actions, nobody quite owning any of them, and in six weeks your line looks the way it did before. In this chapter you choose three, and only three.

Rank losses in hours, not in percent

"Availability is down two points" does not tell your supervisor what to do on Tuesday. Hours do. An hour is something your supervisor can see, count and win back. Take two weeks of stop data, add up the hours lost per loss type, and sort.

Example values. The top three hold 42.6 of the 54.0 hours.

A few loss types carry most of your time, and they are often not the ones your people complain about most loudly. Three rules keep your ranking honest:

• Hours from stop data, not shift notes. Notes remember the dramatic breakdown and forget the forty short stops.
• Top three only. A fourth action splits your team and nothing moves.
• Two weeks, not one. One bad day can dominate a week. Two weeks show the pattern.

One owner and a first action for each

Before you rank, check that every hour sits on the right branch of the loss tree: breakdowns and setup under availability, minor stops and reduced speed under performance, startup and production rejects under quality. If a big share of your hours sits under "Other", fix your reason list before you trust the ranking.

A loss without an owner is a statistic. For each of your three, name one person, not a department, and one first action that can start this week.

Then run one countermeasure per loss every two weeks. Nothing counts as done until the loss hours fall in your data. A countermeasure that "should have worked" and shows nothing in the hours did not work.

What usually goes wrong

• Chasing your number. The talk drifts back to OEE points. Keep it on loss hours.
• Moving definitions. Someone reclassifies a stop as planned halfway through. Your gain on paper is not a gain on your line. Keep the chapter 4 definition fixed.
• Reviews without decisions. A review with no owner and no date is a report. Chapter 7 fixes that.

Your loss register

The same, run from the machines

Your ranking is only as good as the stop data under it. Fabrico shows downtime by cause, line, machine and shift, from the machines, for the current shift, last week or a past month, so the ranking is there without a spreadsheet. From a stop, a technician can open a work card for its owner in the same system.

Request a demo

The ten-minute daily meeting

In many plants the morning meeting is where yesterday gets argued about. Whose numbers are right, why your line stopped, who should have called whom. It runs long, it ends without decisions, and by ten o'clock your day has already gone the way it was going to go.

In this chapter you run a different meeting: ten minutes, standing, at your line, about today. It works because it uses the same loss hours as your Monday review in chapter 10, and because it decides instead of discusses.

One board, one agenda, every day

Your meeting has one board and one agenda, and both stay the same every day. When the agenda stays the same, your people prepare the numbers, not a speech.

Who speaks, and when an action is done

Who speaks

Your shift lead runs your meeting. Maintenance and quality are there, because many actions need one of them. You, the plant manager, listen, and speak only to decide what is blocked. If the most senior person in the room talks most, your meeting becomes a report to that person, and the people who know what happened stop saying it.

When an action is done

An action is done when the loss hours it targets fall in your data, not when someone says it is done. "We fixed the sensor" closes when the stops on that sensor stop appearing. Until then it stays on your board. That is not a judgement on anyone. It is what keeps your board honest.

Ten minutes a day is fifty minutes a week. Compare that with the hours you saw in chapter 1. If your meeting takes back even one stopped hour a week, it is the best-paid fifty minutes in your plant.

Your board and your agenda card

The same, run from the machines

Your meeting is only as good as yesterday's numbers, and paper numbers often arrive late. In Fabrico, the meeting-room screen shows yesterday's downtime by cause, line and shift, and a TV by the line shows the live state, so your whole crew sees a long stop while it is happening.

Request a demo

From stop reason to work card

When your line stops, two records start. Production writes down that your line stopped, for how long, and maybe why. Maintenance writes down what they repaired, with which parts, in how many hours. Both are correct. They are kept by different people, in different places, with different words, and no record shows which repair fixed which loss. That gap is one reason the same failure comes back. In this chapter you close it.

Two records, two stories

Look at one breakdown from both sides.

Example records.

Both are true. Neither tells you that this is the third bearing on the same drive this quarter, because your stop log does not know about bearings and the work order does not know about lost hours.

The handover happens at the stop

The fix is a handover at the stop, not at the end of the shift.

Repeats: the same failure three times

A single breakdown can be bad luck. The same failure three times is missing history. Once your stops and repairs share a record, three questions become a quick search instead of a meeting:

• Which failures repeat? The same machine, the same cause, more than twice in a quarter.
• What do they cost in hours? The lost hours of every stop behind those repairs, added up.
• Did the last repair hold? The time from one repair to the next stop with the same cause.

The repeats are where a preventive task belongs. If the same bearing fails every six weeks, you change it every five, on a plan, before it stops your line.

The spare part is part of the stop

The other half of many long breakdowns is the part. The machine is down, and the part is "somewhere in your store". Two habits shorten those stops: write off every part against the work card that used it, so your store knows what left and why, and give the parts behind your repeats a minimum stock, so the next failure does not wait for an order.

Your stop-to-repair handover card

How to use the card

One record, from the stop to the fix

This is the chapter where production and maintenance become one story. On paper, the handover card does it by hand. In a system, the stop and the work card are the same record from the first second.

Check your own paper cards against this. If the same line, machine and start time are written by three people in three places, you have found the hours your handover loses.

The same, run from the machines

Fabrico keeps production and maintenance in one system. A stop is counted by the machine and its reason is picked at the line. When it needs a repair, a technician opens a work card from the stop and takes the job on a phone or tablet, with the machine and its priority. Technicians write off the parts they use against the work card, with a notification when stock falls below its minimum, and a preventive plan with recurring tasks can cover the failures that keep coming back.

Request a demo

Changeovers

Your plant has two changeover times. The one on the schedule, and the one on a stopwatch that runs from the last good part of the old product to the first good part of the new one. Nobody lies about the first. "Done" just gets said when your machine starts, not when the first good part comes out, and the minutes in between belong to nobody.

In this chapter you time the second number, and you take the first ten minutes out without buying anything.

Time it end to end

A changeover starts at the last good part and ends at the first good part. Everything in between counts: fetching tools, cleaning, the tool change, the first runs that come out wrong, the adjustment, the second adjustment. If your stop log closes the changeover when your machine starts again, every minute of adjustment moves from setup into "slow running" or "startup scrap". The loss quietly changes branches on your loss tree.

• The good count marks the end. Not the green light, not your operator's word. The first part that passes ends the changeover.
• Book it as a changeover, never as a planned stop. A changeover booked as planned drops out of OEE, and a loss you do not count is never attacked.

Film one changeover

The standard method goes back to Shigeo Shingo's single-minute exchange of die. You do not need the whole method to start. You need one phone, one changeover, and one hour afterwards with your people who did it.

You will be surprised by what most teams find: how much of your stop is walking and waiting.

The first three moves that pay

An example, worked through

A filmed changeover takes 45 minutes from the last good part to the first good part.

Example values, not typical ones.

On your line, with two changeovers a shift, that is 40 minutes of running time a shift. Over a five-day week on three shifts, it is ten hours.

40 minutes × 15 shifts = 600 minutes, from the example above.

Your changeover worksheet

The same, run from the machines

Fabrico records a changeover as a stop, with the reason picked at the line, so changeover minutes show by line and shift next to every other loss, and the stop minutes before and after each move show without a stopwatch. The machine's stop ends at the restart, so time the run-up to the first good part by hand, as above.

Request a demo

The 90-day plan and the business case

The first nine chapters were about one line: what its stops cost, how you measure them honestly, how you pick three and attack them. In this chapter you put it together into two things you can hand to other people: a 90-day plan your plant can run, and a one-page case your finance team can sign.

Part one: 90 days on one line

OEE rises when every loss has an owner, not when your number is on a screen. Your plan has four phases over thirteen weeks, and one meeting every Monday keeps it moving.

Two weeks per countermeasure is a pace, not a law. If a fix waits on a part, start on the next loss while you wait, and keep the Monday review either way.

The Monday review: 30 minutes, same agenda

The engine of your plan is one review every Monday, with the same data and the same agenda. In the room: you, and your production, maintenance, quality and continuous improvement leads.

Day 90: what done looks like

• The same OEE definition on every line in scope.
• Baseline and current OEE published per line.
• Your top three losses down, proven in your data.
• Owners and actions reviewed every Monday.
• Fixes written into standard work.
• The next three losses chosen, with owners named.

Part two: the case finance can sign

Your finance director signs money and risk, not OEE points. Your case has three steps: hours, measured rather than estimated; money, with each hour priced once; and months: when the cash comes back, and what waiting costs.

Five drivers turn recovered hours into money

Capacity only counts if your market takes the extra output. If it does not, value the same hours as labour and energy saved.

Seven numbers to ask for in week one

Send this list in week one, and give every number an owner and a source:

Price each recovered hour once

Three rules keep your case honest. Count each hour once: an hour sold as output is not also saved overtime. Your measured loss hours are the ceiling: you cannot win back more than you lose. And use the low estimate: a case that survives it is easier to sign.

One example line, one year. It loses 600 measured hours. The case claims 250 of them and leaves 350 unclaimed.

Example values, not typical ones.

Cost the whole project, not only the software

Ask every vendor for one-time and yearly cost, in writing, and check that the total covers five items:

• The subscription for your lines and users in scope.
• Hardware and connectivity on your shop floor.
• Integration with ERP and other systems, if you need it.
• Internal time for IT, engineering and supervisors during setup.
• Training for your operators, supervisors, maintenance and new starters.

Shrink the risk, then ask for one decision

And show the cost of doing nothing. The example line loses 600 hours a year: 50 hours for every month the decision waits. In money, every month of waiting costs one month of net value.

One slide, six boxes, one decision

If you can say it in one breath, your slide is ready: we lose this many hours a year; winning back these is worth this much; it pays back in this many months.

The same, run from the machines

Both halves of this chapter rest on one number: lost hours, measured. Fabrico takes the stops from the machines, and your operator tags each stop with its cause, so you see downtime by cause, line, machine and shift. With hardware, installing and connecting the devices takes about a month, and the full rollout takes three to four months, so your first baseline may be taken by hand. Hosted on AWS, ISO 27001 certified by DNV.

Request a demo

Choosing an OEE and MES system

Everything in this guide works on paper first. When you want it to run every shift without the paper, you will look at systems. If your choice goes wrong, it will be on scope and data, not on features. This chapter gives you three steps: know what you are buying, weight twelve questions before any vendor presents, and make them show it live.

Three kinds of system are sold under one name

Pick your scope first. Each column usually includes the one to its left. Score vendors only against the scope you chose.

Take the steps in order. If you skip the first one, you score vendors against the wrong scope, and your best score can still be the wrong system.

Twelve questions, weighted before the first demo

Agree the weights with production, maintenance and IT before any vendor presents: 3 for a must-have, 2 for important, 1 for useful. Then score each answer from 1 to 5.

Weight first, score second, then multiply

Two vendors, same raw score, different result

The most you can score is 5 × 27 = 135. Vendor A: 103 ÷ 135 = 76.3%. Vendor B: 90 ÷ 135 = 66.7%. Their raw scores tie at 44. Your weights decide.

Example scores, not real vendors.

Ten things to see live, on screen, not in slides

Send this script to every vendor one week before the demo. Ask for each item on a real screen.

After each demo, score the twelve questions while it is fresh. After the last one, fill your decision grid on the worksheet page.

Red flags, and what IT needs before anyone signs

Listen for these answers in every demo. A red flag scores 1 on its question. On a weight 3 question, that ends the vendor's chances.

Seven answers IT needs

Collect all seven before your shortlist. A slide is not an answer. Ask for each one in writing and file it with your scorecard, so the answer you scored is the answer you sign.

Where Fabrico fits

When you want this method to run every shift without the paper, this is where we fit. Fabrico is a cloud-based MES and OEE platform for manufacturers, with maintenance management (CMMS) built in. Your machine data, your stops with their causes, your work cards and your reports live in one place.

Score us with your sheet from chapter 11

Sources

Every published figure in this guide comes from the sources below. Worked examples use round numbers and say so on their page. If you disagree with a number, disagree with the source.

About the examples and the screens

• Every example line, shift, vendor and record in this guide is constructed to show the method. None is a customer.
• The screens are real Fabrico screens from a demonstration plant, with sample data.
• The scoring method in chapter 11 is our own; use it as a starting point and change the weights to fit your plant.

Why Lines Stop · published by Fabrico · October 2026 · team@fabrico.io