Plant operations
What the daily plant meeting really costs — and how to measure it at your plant
Ask a plant manager what is wrong with the morning meeting and you will usually hear about the meeting: it runs long, too many people attend, it drifts. Those are real complaints and mostly cosmetic. The expensive problem is not the thirty minutes in the room. It is the sixteen hours afterward, when what was agreed quietly fails to happen.
A daily coordination meeting is a plant's fastest control loop. It is where a kiln upset overnight becomes a decision about feed rate this morning, where a rising baghouse differential becomes a bag change on Friday's outage list, where a hazard someone noticed on a walkdown becomes a work order with a name attached. When that loop closes, the plant runs on what it actually knows. When it does not, the plant runs on what people remember, and the same conversation happens again next week with more urgency and less time.
The gap is hard to see because nothing fails visibly. No one refuses to do the work. The action simply never acquires the properties it needed — an owner who agreed, a date, a definition of done, evidence that it happened — and it decays. Six weeks later a bearing fails and someone says, with complete sincerity, we talked about that.
Five ways the loop leaks
These are failure modes, not people problems. Each one is a specific property that went missing between the room and the record.
1. The owner who was not in the room
Work gets assigned to whoever is named, not to whoever agreed. Purchasing is not at the 6:30 meeting, so the action reads purchasing to expedite the bearing and nobody in purchasing knows it exists. The action looks complete on paper — it has a verb, a noun, and a department — but it has no accepted owner, and an owner who has not accepted is not an owner. This is the single most common way an action dies, and it is invisible in every set of minutes that records assignment rather than acceptance.
2. The number that changed shape
Someone says the kiln held 211 tph overnight. By the time it reaches the minutes it is about 210, and two weeks later, in a different meeting, it is we were running 215 before the fan problem. None of these people is lying. The number lost its basis — was that an instantaneous reading, a shift average, wet or dry, before or after the cooler fan came back — and a number without its basis cannot be compared to anything, so it drifts toward whatever the current argument needs.
The same erosion hits units, times, and equipment identity. The tail pulley bearing is a precise object if everyone in the room shares one mental map of belt 561-BC-04 and an ambiguous one the moment a contractor or a new engineer reads it.
3. The decision nobody wrote down as a decision
Plants make good decisions in these meetings constantly and record almost none of them. What gets written is the outcome — holding at 210 — not the decision: what problem was being solved, what options were considered, who owned the call, what was explicitly accepted as risk, what would trigger a review. Six weeks later, when someone asks why the plant was running conservatively, nobody can reconstruct it. Worse, the dissent disappears. The reliability engineer who said I would run 215 and watch the amps was right or wrong, and either way that judgment is now unavailable to the next decision.
4. The question that dissolved
Does the grating repair have a work order? Nobody knows. The meeting moves on, because meetings move on. An open question with no owner and no due date is not tracked anywhere in most plants — it is not an action, so it does not enter the action list, and it is not a decision, so it does not enter the minutes. It simply evaporates, and the hazard it was attached to stays open with a temporary control and no scheduled repair.
5. The action that closed without evidence
The status turns green because someone said it was done. Whether it was done to the standard anyone had in mind is unknown, because the standard was never written. Write up the bag section change-out for the outage list can mean a line item, a scoped job package with parts staged, or a verbal mention to the planner. All three get marked complete. Only one of them survives contact with the outage.
Measure the gap at your own plant
You will find no shortage of published statistics about the cost of bad meetings. Nearly all of them are extrapolated from surveys of office workers and none of them describe your plant. The number that will actually move your leadership is your own, and two weeks of honest observation will produce it.
Run this alongside the meeting you already hold. It requires no software and no change to how the meeting is conducted — changing the meeting first would contaminate the baseline you are trying to take.
The four measurements
Re-litigation rate. Each day, count the agenda items that are substantially the same as an item from a previous meeting in the window. Not routine status — genuine repeats, where the plant is re-deciding or re-discovering something it already handled. Divide by total items. This is the clearest single indicator that the loop is not closing, and in most plants it surprises people.
Administrative time. Time the whole tail: how long the facilitator spends writing minutes, how long distributing them, how long anyone spends reconstructing what was decided when someone asks later. Multiply by the loaded hourly rate of the people doing it. This is the one number that is trivially easy to collect and almost never collected.
Action field completeness. Take every action recorded in the window and score four properties, one point each: does it have an owner who accepted it; a due date or an operating trigger; a stated definition of done; and an agreed form of closure evidence. Report the percentage scoring four out of four. In most plants the first run comes in low enough that people assume the scoring is wrong. It usually is not.
Carry-forward count. How many actions appear in the window that were also open at the start of it? Which are on their third appearance? A recurring carry-forward is not a slow action; it is an action missing one of the four properties above, and the specific missing property tells you which failure mode you have.
Turning it into a number
Two of these convert to money directly. Administrative time is a rate times hours. Re-litigation is the meeting's cost per minute times the minutes spent on repeat items, and the meeting's cost per minute is simply the loaded rate of everyone in the room divided by sixty — a figure that tends to end arguments about whether the meeting is worth improving.
The other two are leading indicators rather than costs. Low field completeness and a growing carry-forward list predict the expensive events — the deferred repair that becomes an unplanned outage, the hazard that stays open past its interim control, the expedite freight on a bearing that was discussed six weeks earlier. You cannot responsibly assign a dollar figure to those from a two-week sample. You can show the trend, and the trend is usually enough.
Do not skip the baseline in order to get to the improvement. A plant that cannot state its starting numbers cannot tell whether anything it changes afterward helped, and every vendor it talks to will be happy to fill that vacuum with a case study from somebody else's plant.
What closing the loop actually requires
Once you have the numbers, the fix is unglamorous and structural. Four properties, enforced rather than encouraged:
An owner who accepted, not an owner who was named. Acceptance is a distinct event from assignment and has to be recorded as one. If the owner was not in the room, the action is not ready to leave the meeting.
Critical fields captured with their basis. Numbers keep their units, their basis, and the time they describe. Equipment keeps its tag, not its nickname. What was ambiguous stays visibly ambiguous instead of being silently rounded into a false precision.
Open questions tracked as first-class items. A question with an owner and a date is a manageable object. A question without them is a rumor with a countdown on it.
Closure that requires evidence. The standard is agreed when the action is created, not argued about when someone wants to mark it done. If closure evidence cannot be defined up front, that is a signal the action is not yet well enough understood to assign.
What no tool fixes
Be skeptical of anything — including software — that claims to solve this by summarizing better. A summary of a meeting where nobody accepted ownership is an accurate record of nobody accepting ownership. Structure exposes the gap; it does not fill it.
The judgment stays human, and should. Deciding to hold the kiln at 210 rather than 215 is an operating decision with real consequences, made by people accountable for them. What can be moved off people is the clerical burden: capturing the fields, tracking what is open, resurfacing the unanswered question tomorrow, and refusing to let an action close without the evidence everyone agreed to. That work is mechanical, and it is exactly the work that gets dropped first when a plant is busy — which is precisely when the loop matters most.
If you run this measurement and want to compare notes, or you are considering a governed pilot on the meeting you already hold, CementOps AI is selecting a small number of design-partner plants: jlarkin@planthuddle.com.