Paper is cheap but slow, tablets add structure but still rely on people, sensors capture time and counts on most machines, and PLC links give the richest machine data where you can reach it. No single method covers everything, so use machines for time and counts, and people for causes, quality decisions and context.
Start with the data, not the method
Shop floor data covers far more than machine stops. Think of orders started, quantities, scrap, labor hours, material lots, inspection results and handover notes.
Each type has a natural source. Some facts exist only in a machine, and others exist only in somebody's head.
If your goal is specifically an OEE number, read OEE data collection methods. This page takes the wider view.
What each method can capture
| Data | Paper | Tablet | Sensor | PLC link |
|---|---|---|---|---|
| Exact stop times | Poor | Fair | Good | Good |
| Short stops | Poor | Poor | Good | Good |
| Part counts | Fair | Fair | Good | Good |
| Machine alarms | Poor | Poor | Limited | Good |
| Cause of a stop | Good | Good | No | Partly |
| Scrap reason | Good | Good | No | Partly |
| Material lot | Fair | Good, with a scanner | No | Rarely |
| Handover notes | Good | Good | No | No |
Read the table by rows, not columns. Sensors win on times and lose on causes, so the best answer mixes methods.
Paper
A sheet on a clipboard needs no IT and no training, and it can start tomorrow. It also keeps details no system asked for, like a margin note about a bad batch.
- Tradeoffs: figures arrive late, handwriting can be hard to read, and someone must type them in again.
- Best for: a first look at one line, or as a fallback during an outage.
- Failure signs: neat round numbers, and every sheet filled in with the same pen at the end of the shift.
If you start here, give people a fixed form, such as our downtime log template. A fixed form is much easier to total than free notes.
Tablets and terminals
A tablet at the line keeps the human knowledge and removes the retyping. Drop-down lists stop spelling drift, and each entry gets a time as it is made.
- Tradeoffs: entries still depend on people finding time, and a long menu is a chore that slows the line down.
- Best for: stop causes, scrap reasons, order start and end, and lot scans.
- Failure signs: one reason code taking most of the entries, because it sits at the top of the list.
Short lists work better, and downtime reason codes shows how to build one. Moving away from forms as a program is covered in paperless manufacturing.
Sensors
External sensors read a machine from outside, without its controller. A clamp-on current sensor gives state, and part counting sensors give output.
- Tradeoffs: each one needs mounting, power and thresholds, and it sees only what it is pointed at. Anything inside a cabinet is work for a qualified electrician.
- Best for: older machines, and quick pilots where no one wants to open the control logic.
- Failure signs: counts that drift from hand counts, and stops that appear when a bracket is knocked.
PLC links
A direct link reads what the controller already knows, including alarms, modes, speeds and counts. The data is rich, and it needs no new sensors on the machine, though you may need a gateway.
- Tradeoffs: it needs someone who knows the program, as well as network access and care with old controllers.
- Best for: newer machines, and lines with many stations.
- Failure signs: a run bit that stays on during a jam, or counters that reset without warning.
Choosing signals is covered in PLC signals for OEE. Protocol routes are in Modbus machine data and IIoT gateway.
The cost you do not see: time spent recording
Paper looks free because the paper is cheap. The cost sits in the minutes people spend writing and retyping.
Example: 10 operators each spend 12 minutes a shift on sheets, and a clerk spends 40 minutes a day typing them up, over 5 shifts a week.
Run this sum with your own numbers before you compare methods. You also have the delay to consider: figures typed the next day cannot help the shift that made them.
How to choose, one data type at a time
Who will act, and on what? Data nobody acts on is waste.
Stop times, counts, scrap reasons, lots or labor hours.
Machines for time and counts, people for causes and judgement.
Compare with a hand record for a shift, and fix what disagrees.
Add machines and data types step by step, starting from what the team uses.
Three mixes that work
- Getting started: paper or a tablet, with a single part counter on the bottleneck. You will have a first number and a list of main stops.
- Older mixed plant: sensors for state and count on each machine, plus a tablet for causes and scrap.
- Newer automated line: PLC links for state, count and alarms, a tablet for causes and a camera on the worst stop point. See computer vision for OEE.
A mix needs a common language. Use the same machine names, the same reason codes and the same shift times across every method, or your data will not join.
How data collection goes wrong
- Collecting everything because you can. More fields mean less care per field, so start with what someone will use.
- Asking operators to record what a machine can record. Save their attention for what only they know.
- Never checking the records. Any method can be wrong, so compare it with a person watching the line now and then. Downtime data accuracy shows how.
- Never discussing the data. People stop filling in data that nobody discusses. Use it in the daily production meeting.
One line, one question
Choose one line and one question, such as "what stops this line most often?". Then pick the simplest method that will answer it within a week.
When you outgrow spreadsheets, what is MES explains the software layer that holds all this data together.
Machine facts and human knowledge both need a home. Fabrico, a light MES for OEE, takes stops, speed and counts live from PLCs or IoT sensors, with add-on AI cameras counting output where a machine has no usable signal, while operators tag the cause at the line. See how Fabrico collects line data.
What people ask next
What is shop floor data collection?
It means capturing what happens in production as it happens: machine states, counts, scrap, labor, materials and notes. The source can be paper, a tablet, a sensor or a machine controller. Most plants use a mix.
Should I replace paper with tablets first or with sensors first?
Ask which is the biggest gap. If you do not trust stop times and counts, sensors or a PLC link will fix that. If times are fine but causes and scrap reasons are poor, begin with tablets.
How is this different from OEE data collection?
OEE needs a narrow set of inputs: time, counts, good parts and stop causes. Shop floor data can also include labor, lots and quality records. For the OEE side, see OEE data collection methods.
How do I keep operators entering data?
Make entry fast, with a short list and few taps, and let machines record whatever they can. Then show people the data in a regular meeting and act on it. Operator downtime logging has more detail.
What software do I need to collect shop floor data?
A spreadsheet can hold paper data for one line, and our OEE tracking template is a place to begin. Once several lines and machine signals are involved, you usually need a monitoring tool or MES.
Fabrico is a light MES for OEE. It takes stops, speed and counts from PLCs or IoT sensors, with add-on AI cameras counting output where a machine has no usable signal, and operators tag the cause at the line.
Related guides: OEE Without Software: Paper and Spreadsheets · OEE Calculation for a Machine Shop, by Example · Machine Downtime QR Scanner: Setup and Limits