Why nobody quite trusts the job profitability number.

You can probably produce a job profitability number. The question is whether you'd bet a decision on it. Ask what a job made and you'll get a number and a shrug in the same breath, because you know what's underneath it. The usual advice is to buy a better report. A better report sitting on soft data is still soft.

I owned a machine shop for a decade, so this isn't a theory I read about. You quote a job at a number, the floor runs it, something goes sideways, a setup takes twice as long, you eat a re-run on scrap, material comes in over. Then accounting closes the month and a number appears. Nobody's really asking what the report says. They're asking whether to believe it.

Job costing is only as good as the labor capture underneath it

Your ERP almost certainly has the fields already. JobBOSS, E2, Global Shop, M1, even a tight QuickBooks and spreadsheets setup will hold labor against a job and material against a PO. The fields aren't the problem. What goes into them is.

Job costing rests on people remembering to record what they were doing, at the moment they were doing it, all day, every day, while also doing the work. That's the whole foundation, and it runs on people remembering. So it fails the same handful of ways in nearly every operation:

• Clocking on and off every task isn't how the work feels. A machinist runs one job while a second is in setup and answers a question about a third. Popping in and out of the time clock at every switch doesn't match that, so it gets approximated.
• Time gets reconstructed, not recorded. Paper sheets filled in at the end of the day, or worse, Friday afternoon for the whole week. Nobody remembers Tuesday accurately on Friday. They remember roughly, and roughly is what lands in the ERP.
• Being tracked feels like being watched. If the system is understood as a way to check up on people, any excuse not to use it will get used. That's not a character flaw. It's what happens when the tool looks like a way to check up on people.
• There's no clean place to put indirect time. Setup, waiting on material, fixing somebody else's problem, a machine down. It gets buried in whichever job happened to be open. That job now looks worse than it was, and another looks better.
• Material and outside processing land late, or against the wrong job. The receipt hits after the job closed, so the cost never reaches the part it belongs to.

Every one of those makes the report wrong in a way it can't detect. It totals the numbers it was given, correctly, and hands you a confident answer built partly on a Friday afternoon guess. That's why the number gets doubted, and the doubt is rational. Your people know exactly how that data got there.

What actually fixes it

The data has to get captured without anybody having to remember to record it. You've probably worked that out yourself after getting burned once or twice. Every costing setup I've seen that holds up works that way. The ones that don't rely on discipline nobody sustains for long.

In practice that means taking capture off people wherever the work already leaves a trace, and making it nearly frictionless where it can't:

• Take the timestamp from something that's already happening. A job gets scanned at a station, a program gets loaded at the control, a first article goes to inspection. Those events already occur and already carry a time. Reading them beats asking somebody to remember them.
• Make whatever entry is left one action, at the machine. Not a login, a menu and a job search. Badge and a barcode at the point of work, or it will be reconstructed later.
• Give indirect time a real home. If waiting on material is a button, you find out how much of the week goes to waiting on material. If it isn't a button, that time hides inside job costs and makes good jobs look bad.
• Close the loop on material and outside processing. Receipts and vendor invoices have to reach the job while the job is still open. That's a systems problem, not a discipline problem.
• Put the number in front of people the next morning, not at month-end. When the read is same-day, a bad entry is visible while everyone still remembers the day. Three weeks later nobody can correct it even if they want to.

Do that and the number stops being an argument. Not because the report got smarter, but because what feeds it stopped being a memory exercise.

What this looks like once it's running

For a mid-size manufacturer I work with, the daily version of this is an owner's morning brief. Overnight it reads the ERP and the books and emails one page: cash-flow forecast, backlog, sales month-to-date. No logging in, no running three reports and stitching them together in Excel. The value isn't the email. It's that "where are we right now" gets asked and answered every day, so a wrong number surfaces while it can still be fixed.

Alongside it sits a monthly KPI briefing: on-time delivery, quality, scrap, quoting and sales, with the trend, filling the spreadsheets that used to eat a chunk of somebody's month-end.

There's a limit to this, and it's a real one. None of that fixes capture by itself. A brief reads what your operation records. If labor is being reconstructed on Friday, the brief will faithfully show you a reconstructed number, just sooner. What it buys is speed of detection, and detection is what makes the capture work worth doing. Anybody telling you a reporting layer alone will hand you a true margin is selling you the part that was never broken.

Closing the quote-to-actual loop

Once capture is honest, the actuals are worth something beyond the job they came from. This is where operations leak the same margin twice, once when the job runs over and again when they re-quote the same part off the estimate that was already wrong.

For the same manufacturer, a feedback loop I call the Shop Brain compares quoted against programmed against actual times, pulled from the ERP and from Mastercam, so quoting learns from what the floor actually did. The estimator stops working from memory and starts working from what the part really took last run. Quote the job, run the job, see the variance, let the variance reach the next quote. That's the loop that should have been closing all along, and it only works if the actuals are real.

If you want to go a level deeper on getting clean reporting out of a specific ERP, the guide on getting real reports out of JobBOSS covers the report-writer wall and what's actually possible without living in spreadsheets.

Common questions

Do I need to replace my ERP to fix this?

Almost never. The fields already exist in JobBOSS, E2, Global Shop, M1 or a QuickBooks and spreadsheets setup. Replacing the ERP hands you a new set of empty fields with the same capture problem underneath them. Fix capture first and most operations find the system they had was adequate.

Our people push back on time tracking. Is that fixable?

Usually, and not by insisting harder. Push-back is almost always a reaction to friction or to feeling watched. Cut the number of actions to one at the point of work, give indirect time an honest place to go so nobody has to lie about a slow morning, and use the output to fix quoting rather than to grade individuals. When people see the data coming back as better estimates instead of as a report card, most of the resistance goes with it.

Where do we start if the data is a mess right now?

Pick one thing that already leaves a trace somewhere and read that instead of asking a person for it. It's usually the cheapest honest win, and it tells you quickly how far the recorded numbers sit from reality. That gap is the useful finding, and it's normally the argument that gets the rest of the work funded.

How long does something like this take to stand up?

It starts with a paid diagnostic. Then a build that's quoted up front, so you know the price before it starts, with no hourly billing. I host and run it for you on managed infrastructure that works on top of the systems you already run, so nothing moves to a new platform. An ongoing care plan after the build keeps it running: hosting, monitoring, bug fixes, and small adjustments as your work changes; term and price by agreement, and the ongoing cost is simply what keeps the tool running and improving. No hourly billing and no per-seat license.

Where the numbers come from. If your ERP and your books don't talk, the profitability question is being answered twice by hand: JobBOSS QuickBooks integration. And the reporting side of the same system is JobBOSS custom reports. Once the capture is honest, the delivery question is the manufacturing KPI dashboard nobody opens. On engineer-to-order work it is harder again, which is software for machine builders.