Software
Shop Floor Data Collection
Why crews stop recording, what recording actually costs in shift time, the capture methods compared, and how to pick your granularity before you buy.
Quick answer: shop floor data collection decides whether your job costing is real. The failure is rarely the technology. It is that recording competes with the work, the person recording gets nothing back, and by month three the terminal is a coat hook. Choosing how much detail to capture, honestly, matters more than choosing the tool.
What It Has to Capture
Six things, and each one answers a question the office cannot otherwise answer.
| Captured | Answers |
|---|---|
| Who, on which job | Whose time went where |
| Which operation | Where within the job the hours went |
| Start and stop | How long it actually took against the estimate |
| Material consumed | What the job really used, including drop |
| Scrap | What was lost and at which step |
| Rework | Whether this was the first attempt |
The last two are the ones most implementations drop to keep things simple, and they are the two that explain the margin. A job that took forty hours tells you nothing. A job that took thirty hours plus ten hours of rework at the weld stage tells you what to fix.
Why Crews Stop Recording
This is the part vendor material skips, and it decides whether any of this works.
Recording competes with the work. A welder who stops to clock an operation spends those twenty seconds away from the arc. And critically, the person doing the recording gets nothing back from it. The data goes to the office. It produces a report the welder never sees, which may later be used to ask why a job took so long.
Treating that as laziness is the standard mistake and it leads to the standard failure, which is more enforcement, more resentment, and worse data.
Two things genuinely fix it:
Reduce the cost of recording. Fewer interactions, larger buckets, badge scan instead of typing, a station people already walk past. Every interaction you remove is one the system will still have in month six.
Give something back. A crew that can see their own numbers, and see the estimate they were working against, engages differently from a crew feeding a black box. Shops that post job results where the floor can see them get better data than shops that do not, for no extra software.
What It Actually Costs
Worth calculating instead of guessing, because the fear is usually larger than the number.
| Value | |
|---|---|
| Clock events per person per day | 8 |
| Seconds per event | 20 |
| Time per person per day | 2.7 minutes |
| Working days | 240 |
| Hours per person per year | ~11 |
| Six shop staff | ~64 hours |
| Cost at an $86 shop rate | ~$5,500 a year |
Now weigh that against what it reveals. On $1.5 million of revenue, finding and fixing two points of margin is $30,000. The trade is usually not close, and it is worth stating in those terms to a floor that suspects recording is bureaucracy for its own sake.
The number that would change the answer is interactions per day. At eight events it is a rounding error. At thirty it is a real cost and people will stop, which is the argument for fewer, bigger buckets.
Choose Your Granularity Before You Buy
This is the decision, and almost nobody makes it deliberately.
| Job level | Operation level | |
|---|---|---|
| What crews do | Clock on and off a job number | Clock each operation within the job |
| What you learn | The job made or lost money | Which step ran over |
| Discipline required | Low | Moderate and sustained |
| Estimate against actual | By total | By cost line, which is the useful one |
| Realistic for | Most shops immediately | Shops with a reason and a champion |
There is no wrong answer, and there is a wrong way to arrive at one. Buying operation-level capability and getting job-level compliance means paying for detail you never receive and building reports on data that is not there.
Our metal fabrication guide makes the same point from the costing side: a shop that will record at operation level gets operation-level costing, and a shop that will not gets job-level costing. Decide which you are before anyone signs an estimate.
A practical rule. Fewer buckets recorded reliably beats more buckets recorded unreliably. Four operations captured consistently is worth far more than twelve captured when someone remembers. Start narrow, prove it holds for a quarter, then add detail if the floor has absorbed it.
The Capture Methods
| Method | Strength | Watch for |
|---|---|---|
| Paper travellers | Costs nothing, works everywhere, no training | Transcription delay and error, and the data arrives too late to act on |
| Fixed kiosk or terminal | Fast, rugged, and works where crews already pass | Queues at shift change, and useless for crews who do not pass it |
| Barcode or QR scanning | Fastest per interaction, few mistakes | Labels have to exist and survive the shop, including heat and grinding dust |
| Mobile or tablet | Suits spread-out crews and off-site work | Must work without signal, which matters in a shop with thick walls |
| Automatic from machines | Zero effort, genuinely accurate | Only covers machine time, and says nothing about fitting or fit-up |
Two notes worth having before a demo. Automatic machine capture is excellent and partial, because it never sees the hand work where fabrication hours actually concentrate. And offline capability is not a nice-to-have in an Alberta shop with steel walls, so test it disconnected on real hardware in the actual building.
Odoo provides a dedicated shop floor interface reachable from a work order, documented under its manufacturing setup. Most competent systems have an equivalent. What differs between them is far less than what differs between shops that sustain recording and shops that do not.
How to Roll It Out Without Losing the Floor
- Start with one cell or one crew. Going shop-wide on day one is the classic error.
- Pick the smallest set of operations that would still answer your question.
- Tell people what it is for, specifically. “So we quote the next one properly” lands better than “for costing.”
- Show them the output within the first month, including where the estimate was wrong as well as where they were slow.
- Fix the friction fast. If the station is in the wrong place or the login is slow, that is your problem to solve, and solving it in week one buys years of compliance.
- Only then widen, either to more crews or to more detail.
The failure mode is going wide and detailed on day one, hitting resistance, and concluding the software was wrong. The software usually was not.
Related
- Metal fabrication ERP for Alberta job shops covers the six places fabrication margin leaks
- How to calculate your shop rate covers the rate this data gets valued at
- Operations software covers the wider category
- Spreadsheets vs an operations platform covers where manual capture stops being enough
Frequently Asked Questions
- What is shop floor data collection?
- Capturing what actually happened on the floor as it happens: who worked on which job, at which operation, for how long, what material was consumed, what was scrapped, and what was reworked. It is the input that decides whether job costing is real or a reconstruction, and it is the part of a system implementation that most often fails quietly.
- Why do crews stop recording?
- Because recording competes with the work and the person recording gets nothing back for it. A welder who clocks eight events a day is slower than one who does not, and nothing in their day improves as a result. Treating that as laziness is the mistake. It is a rational response to a system that takes from them and gives nothing, and the fix is to reduce the cost or return something useful.
- How much does recording actually cost us?
- Less than people fear and more than nothing. Eight clock events a day at twenty seconds each is under three minutes per person per day, which is roughly eleven hours a year each. Across six shop staff that is about 64 hours, or around $5,500 of sellable time at an $86 shop rate. The question is whether what it reveals is worth more than that, and in most shops it is by a wide margin.
- Should we record at operation level or job level?
- Decide before you buy, because it determines what your costing can ever tell you. Job level tells you a job made or lost money. Operation level tells you the fitting was fine and the welding ran double, which is the finding that changes the next quote. Operation level costs more discipline, so pick the one your floor will genuinely sustain instead of the one that demos well.
- Is a barcode kiosk better than an app?
- It depends on whether people move. A fixed kiosk is fast and rugged and works when crews pass a station anyway. Mobile suits crews spread across a yard or working off-site, and it needs to work without signal in a shop with thick walls. The method matters far less than whether the number of daily interactions is small enough that people keep doing it in month six.