What data to collect from a machine every day — without unnecessary spreadsheets

Where the problem starts

A shift can run without accidents, yet output still falls short. The machine is on, the spindle spins, the operator is busy, but by the end of the day there are fewer parts than expected. Usually this is noticed too late, after the plan has already slipped.

Equal running time does not guarantee equal results. Two CNC lathes might operate for 10 hours each, but one produces 180 parts while the other makes 120. The cause is rarely a "bad day." More often it’s small losses nobody recorded: prolonged setup, waiting for a tool, a long time to the first good part, or a stop after inspection.

The problem begins when the shop only looks at whether the machine was running. The phrase "it wasn't stopped" sounds convincing, but it's not enough for output. What matters is not that the machine was switched on, but how much time it actually spent cutting metal, how much was spent on re-setup, and where the shift broke into short pauses.

In practice the argument usually looks familiar: the supervisor says the machine was loaded all day, the operator recalls a long setup, the process engineer insists the program is fine, and procurement blames delayed blanks. Everyone has their version, and without numbers the discussion quickly becomes subjective.

Losses almost always hide in simple things: setups taking too long, repeated re-setup of the same operation, waiting for blanks or inspection of the first part, short stops that nobody counts as downtime, and one overloaded operation where a queue builds.

Because of this, the plan and reality live separately. On paper the shop is evenly loaded, but in reality one machine is the bottleneck, another waits often, and a third loses time to re-setup. In serial metalworking this is especially visible. Extra 15–20 minutes several times per shift quickly turn into a serious shortfall in output.

Daily accounting quickly brings the discussion back to facts. You don't need a thick report. It's enough to record a few simple metrics each day: cutting time, setup time, reasons for downtime, count of good parts, and loading by operation. Within 3–5 shifts you can see where time is lost and which operation slows the whole shop.

In short, the problem doesn't start with the machine but with the lack of an accurate picture of the shift. Until that picture exists, the shop argues about causes. When numbers appear daily, the argument reduces to one question: exactly where did the minutes go?

Five metrics for every day

You don't need a long report for shift accounting. To understand what's happening at a machine, five numbers and a few short notes are enough.

• Cutting time during the shift.
• Setup and re-setup time.
• Downtime with a clear reason.
• Number of good parts and scrap.
• Time waiting for material, tool or operator.

This set is enough to start making decisions the next day instead of collecting data just for archive.

The most honest metric is cutting time. If an operation should take 6 hours by calculation, but actual cutting lasted 3 hours 40 minutes, the problem isn't spindle speed. Losses usually come from waiting, setup or frequent stops between batches.

Count setup separately from downtime. Otherwise everyone argues over wording and the meaning is lost. If a CNC lathe was switched three times from one part to another, the re-setup might have eaten more time than the machining itself.

Don't look at scrap separately from time. Five ruined parts in a large series may be tolerable, but in a short run that's a serious hit to output. That's why it's useful to note, alongside scrap count, how many minutes it took to get the first good part after setup.

Keep the recording format simple: one event—one line. For example: 09:10–09:35, waited for inserts; 12:40–13:05, re-setup; 15:20–15:35, first part failed inspection. From such notes the supervisor quickly sees what most often prevents the shop from running smoothly.

If the shop runs different orders, daily metrics help spot overloaded operations before schedules slip. Two weeks of such records is usually enough to reveal recurring bottlenecks and eliminate some obvious losses.

How to measure time without arguments

Arguments about time don't arise because of numbers but because of different rules. One operator writes "setup," another labels the same event "waiting," and the supervisor later doesn't understand where 40 minutes went.

So first you need a single dictionary for all shifts. Then numbers can be compared between people, machines and days.

The rules are simple. All shifts should use the same list of stop reasons. Setup, repair and waiting must be recorded in separate lines. Count micro-stops separately from long stops by choosing one threshold, for example up to 5 minutes and over 5 minutes. Each re-setup should have a precise start and end. And one more important rule: a trial run is not part of serial production. The series hasn't started until the first confirmed good part is produced.

Fix the micro-stop threshold once and don't change it every month. If today a short stop is 3 minutes and tomorrow 7, the report quickly loses meaning.

For re-setup use an even stricter boundary. Start—when the machine finishes the previous job and the operator begins changing a tool, fixture, program or setting a part. End—when the first good part of the new order is obtained. Everything before that belongs to setup.

Don't hide trial parts inside production. If the operator changed jaws, adjusted the tool offset and made two test parts in 18 minutes, those 18 minutes should not be recorded as production time, even if the spindle was turning. Otherwise the cycle will later look too long when the series hadn't actually started.

The same rule prevents mixing setup with waiting. If the operator is ready but the blank hasn't been delivered, that's waiting for material. If a mechanic replaces a sensor, that's repair. After the repair, setup continues from that moment, not merged into a single block.

Usually a short cheat-sheet next to the machine and a single time source for everyone is enough. Then the supervisor has fewer disputes at shift end, and the shop gets numbers they can trust.

How to keep records without extra tables

Don't start with a big form. Begin with one short sheet used by all machines on the shop floor. Otherwise one operator writes "setup," another "re-setup," and a third doesn't note anything. You can't compare such records.

For a lathe shop a few fields are usually enough: work start, work end, setup time, stop, reason for stop and parts produced. That's already enough to understand where the shift went.

You also need a specific person to log events. Most often the operator records them and the shift supervisor checks them. If everyone is supposed to record, in practice no one is accountable.

Roll out the system step by step. First choose one form for all machines. Then assign who records and who verifies. After that agree on common event names and stop reasons. Only then run a test for one shift for 5–7 days.

The most common mistake is simple: people fill downtime from memory at the end of the day. Then 7 minutes become 20, and two short stops merge into one long one. When the operator records events immediately, you get live data, not an estimate "by eye."

What to simplify from day one

Don't ask people to write everything. Long forms get filled negligently. Keep only the fields you actually use to make decisions. If no one uses a field like "shop temperature," remove it.

An easy check: open a week's records and see which columns are almost always empty. If a field doesn't help explain downtime, setup time or operation loading, it only gets in the way.

Usually a short test in one section with repeating parts is enough. After 5–7 days it's clear where the form is inconvenient: people confuse stop reasons, forget start times or use very different words. Fix this while the system is still local.

When the form is short and the same for everyone, downtime accounting stops being extra work and becomes part of the shift. That's the goal.

Example of one shift in a lathe section

Take an 8-hour shift, or 480 minutes. This example shows how a few numbers give a real picture without long tables.

During the shift the machine actually cuts for 340 minutes, or 5 hours 40 minutes. Another 55 minutes go to re-setup for a new batch. The shop loses 35 minutes waiting for blanks. Those three figures alone are enough to understand where time is lost.

As shares of the shift the picture becomes even clearer. Cutting takes about 71% of time. Re-setup takes roughly 11%. Waiting for material is another 7%. The remaining 50 minutes usually come from loading/unloading parts, measurements and short stops.

In this example the problem isn't the machining itself. The machine is busy most of the shift, which is normal for a section that changes batches. Output falls because of material supply. If blanks arrive late, the machine stands idle even though program, tool and cutting mode are ready.

This changes how you analyze the shift. There's no point immediately tweaking cutting parameters or hunting for extra seconds in the cycle if 35 minutes were lost waiting for blanks. First find out why the material didn't arrive on time: did the warehouse delay delivery, did the operator wait for a lift, were blanks not prepped, or did the batch lack labeling?

If you only look at part count you can draw the wrong conclusion. There may be fewer parts than planned, but the cause isn't the machining cycle. The bottleneck may be upstream, in logistics inside the shop.

From one such day you can already make a sensible decision: don't try to speed up cutting at all costs, but arrange blank delivery to the machine by the time the current batch ends. Those 35 minutes will then return to productive work without risking quality or overloading the operator.

Where the numbers break most often

Numbers don't break in the table but at the machine. The error is usually not the metric set, but how people name the same event.

The most common confusion is around setup. It's convenient for an operator to record almost any stop as "setup": changed a tool, adjusted a dimension, waited for a program, chased the first good size. But these are different losses. If you lump them together you won't know whether time was spent on real re-setup, adjustment after scrap, or waiting.

The second problem is too-general reasons. Entries like "technical stop" or "shop downtime" explain nothing. If a machine stood for 27 minutes you need to know why: no blank, waiting for a technician, changing the chuck, searching for a tool, or re-checking a dimension after a deviation.

Another weak point is end-of-day memory entries. By shift end people remember noise, not facts. Short stops of 4–7 minutes simply disappear. Often those short stops add up to an extra hour per shift.

Scrap is often recorded separately from time. The inspector counts parts, the supervisor counts minutes, and there's no link between them. As a result a shop might see 12 defective parts but miss that they cost 46 minutes of machine time and another 18 minutes for readjustment. For a CNC lathe that’s a noticeable loss.

The problem is worsened when different machines use different labels. One records "setup," another "re-setup," a third "correction," and then no one can fairly compare shifts or equipment.

Simple measures help: one handbook of stop reasons for all machines, clear separation of re-setup, minor adjustment and waiting, record events immediately, link scrap to lost minutes and forbid vague entries without explanation.

If you follow these rules for a week without gaps, the picture clarifies quickly. You can usually see which machine loses time waiting for tools, which section often shifts size, and where the problem is organizational rather than equipment-related.

Quick check at the end of the shift

In 5–10 minutes before shift handover you can catch errors that later spoil the whole picture. Start not with new graphs but with a short factual check.

First verify whether times add up. The shift has a fixed length and all of it should be divided into clear parts: cutting, setup, waiting, inspection, downtime, maintenance. If 25–30 minutes are missing in an eight-hour shift, those figures can't be reliably compared.

Then ask five simple questions:

• Does the sum of all machine states match the shift time?
• Is there a clear reason for each downtime?
• Was there a repeated setup on the same part?
• Did one operation take too much time?
• Does output match cutting time?

Here’s a simple example. The machine worked 480 minutes. Of them 290 minutes were cutting, 40 minutes for setup, 55 minutes waiting for blanks, 25 minutes for tool change and 70 minutes for small stops and inspection. On paper everything adds up, but output is below plan. That means you should look for problems in blank supply and how often the operator is distracted by measurements, not in cutting parameters.

This check isn't for the report. It quickly shows where real losses are: downtime, repeated setups or one overloaded operation. Do it daily and disputes become much rarer.

What to do with the numbers next

When daily accounting is in place, don't try to fix everything at once. Numbers are for one clear action. Start with the most frequent downtime cause over the last 5–7 shifts.

The first goal is often low-tech: the operator waits for a technician, a tool is hard to find, the machine lacks a blank, or the first part takes too long to inspect. Pick one cause and remove it first. If you scatter efforts, the shop gets bogged down in talk.

A practical rule: one problem, one person responsible, one week to check. That’s usually enough to see whether the solution helped.

Also compare setup time for identical batches. If the same part took 20 minutes to set up on Monday and 45 on Thursday, the cause is almost always concrete: different tooling, searching for a tool, a repeat touch-up, extra measurements or waiting for a program. A week’s comparison quickly shows where the process drifts.

The same applies to operation loading. One turning operation might constantly build a queue while a neighboring resource sits idle. That’s not about people but about work distribution. Sometimes moving part of the batch to another machine helps. Sometimes it’s better to split roughing and finishing between two resources if the part allows.

Another common mistake is recording "machine downtime" for issues actually related to shop supply. If the machine waits for a chuck, holder, insert or blank, the cause lies in shift preparation.

Before major conclusions check four things: is there enough tooling for the whole shift plan, are required tools gathered in advance, are blanks placed near the workstation, and is there a bottleneck at one operation?

After that it becomes clear where order is needed and where the shop truly hits equipment limits. If a queue persists at one operation despite correct setup and no tool shortages, then look deeper at the resource. In such cases EAST CNC can help match a CNC lathe, commission it and provide service. But consider that step only after removing obvious losses.

Then numbers start working for output, not just piling up in a table.