Part 3 listed the ten handoffs a job needs to flow without a person, and step 7 — in-process truth — is the one every other step leans on. You can't sequence what you can't see, can't trigger purchasing for a job you think is still in receiving, and can't answer the customer's "where is it?" without a walk.
Here's the thing: your shop already has a system of record for this. It's on a clipboard, riding along with the parts.
The traveler already knows
Pick up any traveler on the floor. It has the job number in marker, the routing, and a column of initials and dates as each operation was done. Sometimes a stamp. Sometimes a note: "short 4, see Mike." That sheet of paper is the data — who did what to the job, in what order, when. Most shops run on it. It works.
What it can't do is be in two places. The owner wants to know where job 1042 is; the traveler knows, but it's in the finishing bay under a part. So the owner walks. Or calls. Or the office keeps a whiteboard that's a copy of the travelers, updated when someone remembers, which means it's wrong by lunch.
Making the traveler digital means one thing: the information on the sheet is available away from the sheet, as it happens. It does not mean replacing the sheet with a system that's slower to use than a Sharpie. If the digital version costs the operator more than an initial and a date, they will stop using it inside a month, and they'll be right to.
Three questions, not an MES
"MES" (manufacturing execution system, the big floor-control software category) is a word that gets attached to this, and it drags a lot of weight with it — machine monitoring, scheduling optimizers, machine-uptime dashboards. Most of that is several steps ahead of where this post is. Floor truth is three questions:
- Where is every job right now?
- What's been done to it, and what's next?
- Who did it, and when?
That's it. A station tablet or a scanner on a lanyard; a scan on the way in, a scan on the way out. The traveler still prints — the floor likes paper, and paper doesn't need WiFi — but it prints from the record instead of being the record. The initials go into a screen with one tap, and the system knows what the whiteboard was guessing.
Rules learned from the floor
Every one of these was learned by watching a floor ignore a system that broke it:
- One scan, no typing. If the operator has to pick from a dropdown, you've already lost. The barcode on the traveler is the job; the station is known from the device. Scan, done.
- Works offline. WiFi dies in the back bay, behind the big mill, in the paint booth. The app queues scans and syncs later. A scan that fails because of signal teaches the operator the system is unreliable, and they're right.
- Built for gloves and distance. Big targets, high contrast, readable from arm's length with a part in the other hand.
- The print is one tap away. The current rev of the drawing, on the station screen, is the single feature the floor will thank you for. Make it the easiest thing on the screen.
- Operators undo their own mistakes. Scanned the wrong job? Tap undo. If every miss-scan needs the office to fix it, the operator will stop scanning rather than ask.
- The number goes on a screen they see. WIP (work in process) at the station, jobs due today, what's stuck. If the only person who sees the data is in the office, the floor is scanning for someone else's benefit, and they'll feel it. Put it on the wall by the station.
What it unlocks, rung by rung
Floor truth is the rung that makes the earlier ones automatic and the later ones possible:
- Part 1's clipboard becomes live. Arrivals, completions, and WIP per station without anyone counting. The WIP cap can turn red on a screen.
- Part 2's order knows where it is. The structured work order now has a position, so the office can answer "where is it?" from a chair.
- QC capture lands where the work happens. A measurement, a photo of the part, a note — attached to the operation, by the operator, at the time. Not reconstructed at the end of the week.
- Part 3's handoffs get their trigger. Material gets reserved when the job is actually released. Shipping gets booked when the last operation actually scans out. The chain can start to run without someone watching it.
And it changes the customer call. "Where's my order?" becomes a glance, or — further along — a status email the system sent before they asked.
What we measure
- Jobs whose position the system knows without a walk. Usually starts near zero. The sprint commits to a percentage.
- Time to answer "where is it?" From minutes and a walk, to seconds.
- Stale travelers. Hours since a job's last scan, per job. Stale is either stuck or unscanned, and both are worth knowing.
- Scan compliance by station — honestly tracked, because a station that stops scanning is telling you the system is costing them something. Fix the system, not the operator.
The honest limits
A six-person shop with one bay probably doesn't need this. The walk is thirty feet and the owner can see every job from the door. The trigger is when the walk takes longer than the work it interrupts, or when the customer asks for status more often than the office can find out.
And the failure mode is real: a floor that's asked to scan for data nobody uses will quietly stop, and the data will rot, and the system will be declared a failure. The scan has to pay the operator back — the print one tap away, the undo, the number on the wall — or it isn't floor truth, it's monitoring the floor instead of helping it.
Next: the thing that holds all of this together commercially — how a sprint's number gets chosen, quoted, and proven, and what we do when it misses.