The Report Nobody Can Pull Names Your Blind Spot

The report that takes a week

A team asked a protein processor for twelve months of production data. The plant's operations lead did not argue about the value of the request. He argued about the calendar. Twelve months is a lot, he said, can you do year to date. Year to date it was.

Then the list got specific, and the list is where the plant told on itself. Plan versus actual attainment by line: no problem. One line runs steak, the other runs barbacoa, and the numbers already exist because someone reports them. Yield by cut: also fine, the ERP produces a trim loss report that sets standards against actual inputs and outputs. Daily finished goods output: "I think that's the only tricky one." The reply was blunter. "That one's hard. The rest of them are not."

The ask was never yield. It was rate. How long does it take to fill one combo of a given cut, and what does that cost in hours. The plant knew the answer swung wildly. In-process material that only needs opening and checking fills a combo in about twenty minutes. Material that has to be cut and chunked can take an hour and a half on the same station with the same crew. That is a four times swing in the unit of work, and the floor can run as many as seven different cuts in a day.

Every piece of the record that would price that swing exists somewhere. When did this run start. How many labels printed. What did the weights read. When did the run finish. The operations lead said as much: it goes somewhere, the question is whether any report pulls it together without being complicated. In their systems a batch does not equal a row, so the record has to be assembled by a person, and that person has three other things to get out that day.

A plant meters what it is held to

The reports that came back fast were not the easy ones technically. They were the owned ones. Attainment by line exists because scheduling answers for it. Yield exists because someone answers for give-away and trim. Nobody answers for hours per hundred pounds, so nothing counts hours per hundred pounds.

Look at how the day is governed and the reason is obvious. There is no shift schedule in the usual sense. The crew starts at seven and runs until the truck is full. Breaks are a fifteen and a thirty, and if the day runs long they get a second fifteen. Heavy pounds on a hard cut means more hours for fewer pounds. Light pounds and an easy cut means the parking lot at three thirty. Hours are the designed flex in the system, and the plan absorbs variation by letting the day get longer.

That is a defensible operating rule for a business that cannot control its own input, and this one cannot. The parent allocates the raw material, the purchaser is doing inventory management rather than sourcing, and pallets arrive commingled, so isolating a single cut would mean destacking and separating every load. Mix arrives, the floor cuts what shows up, and the clock closes the gap.

The trap is what happens next. The flex variable is exactly the variable that never enters a report, because reporting follows accountability and nobody is accountable for a number the operating rule treats as elastic. So the plant can tell you precisely how it performed against a plan it wrote, and cannot tell you what an hour of its own labor bought. Every improvement case that arrives after that, automation, crewing, a line rebalance, gets built on the half of the operation that leaves records.

The same hole shows up in the award file

This is not only a floor problem. A packaged foods brand ran a corrugated and folding carton bid, and the award file had to be rebuilt weeks after it should have closed. One group worth roughly $117K in award spend came out of the folding carton award because those same items already sat in the corrugated bid and were being bought from a different supplier. Material numbers had to be corrected before the duplicates were even visible.

Underneath it was the same failure. Nobody owned item-level volume, so the forecast handed to suppliers was filled in only for the items the team already knew about. Two suppliers came back and said, in effect, this is not what we signed up for, and asked to requote at current market levels. The bid did not break on price. It broke because the number that defined the bid had no owner, and unowned numbers get invented at the moment someone needs one.

Build the record before you build the case

Run the four-report test before you scope anything. Ask for plan versus actual by line, run start and stop times, output by hour, and daily finished goods output, all in one sitting, and time each one. What comes back in a day is what the business manages. What takes a week is where your variation is hiding, and it is the number your project will eventually need.

Then define the atomic record and get agreement on it in writing. For this plant it is one line per run: start time, cut, labels printed, weight per container, stop time. That record produces pounds per hour by cut, which produces hours per hundred pounds, which is the only honest denominator for a labor or automation case. Ask for the demonstrated rate file the current standards were built from and find out whether anyone has opened it this year. It usually exists, left over from last year's annual operating plan, sitting where one person put it.

Name the system of record before any trial starts, not after. On a film qualification at another protein plant, the baseline for run speed and downtime was pinned to one production data system before the first roll ran, and quality specification ownership was pushed to the client's quality team rather than left with procurement. That looks like paperwork. It is the difference between a trial that produces a defensible number and a trial that produces an argument.

What a measured operation looks like

Run start, run stop, labels printed, and weight per container come off a single report and land inside a day, not a week. Hours worked per hundred pounds is reported by cut every week, and the spread between the fastest and slowest cut is a figure operations can state from memory. Volume by item has one owner and one system, and any forecast sent to a supplier is pulled from it rather than typed into a spreadsheet. When a data request cannot be filled, the gap gets logged as a gap with a name against it.

The number nobody owns is the number you are trying to change

Every plant will hand you its plan compliance in an afternoon. Almost none will hand you what an hour actually bought, because the hour is where the plan hides its variation. Ask for the hard report first, and the answer to your capital question is usually sitting in the delay.