How Variance Data Fixes the Routing That Was Always Wrong

Some routings have been wrong for years, and the reason nobody fixed them is that nobody could prove which ones: variance data proves it by showing which operations miss their standard in the same direction, run after run. EDGEBIC by User Solutions compares planned hours against logged actual hours per operation, per day, and per work center, so a bad standard stops being an opinion held by one supervisor and becomes a pattern in a report. The payoff is not accounting accuracy. It is a capacity plan that stops lying to you.

This post covers the routing-accuracy outcome. It sits under the EDGEBIC results guide. For the visual version of the same comparison, see the planned versus actual overlay.

The Routing Everybody Knows Is Wrong

Every shop has one. A part where the routing says 2.2 hours on the mill and everyone who runs it knows it is closer to three. The standard was set years ago, possibly on a different fixture, possibly by someone who is no longer there. The operators work around it. The supervisor pads the schedule mentally. Sales quotes off it anyway.

Nobody fixes it because fixing it requires evidence, and the evidence was scattered across timecards nobody reconciled at the operation level. What you had instead was an anecdote: a good machinist saying the number is low. That is true and it is not enough to change a standard.

What a One-Directional Gap Looks Like

The distinction that matters is between a bad day and a bad standard.

Only the third and fourth rows are routing findings. The first two are the shop being a shop. That is why one job's overrun proves nothing and twenty jobs' overruns prove a great deal, and it is the whole reason variance needs to be reported over a population rather than discussed job by job.

Three Reports, Three Grains

The comparison shows up at three levels of detail, and you use them in this order:

Daily Production is the scan. One row per day and work center: planned hours, actual hours, variance, adherence percentage, and jobs touched. You are looking for work centers whose adherence sits consistently on one side of 100 percent.

Shift Production is the same lens with operator context and a plain-language badge on each row: "on rate", "-1.5h", "+2.0h overtime". Chronic over-runners are obvious at a scan, which is exactly what you want before you commit to a deeper look.

Work Center Utilization, plan versus actual variance tab, is where you confirm it. This is the capacity authority in the report set, because it computes available hours from real shift calendars and per-day overrides rather than approximating them. If a station shows a persistent gap here, it is real.

Earned Value adds a job-level version of the same signal: CPI is earned value divided by actual cost, and a CPI persistently below 1.0 means the job is burning more hours than the plan earned it. A part number whose jobs all land under 1.0 is a routing problem wearing a job costume. See the report catalog for how the families fit together.

The Arithmetic That Makes It Worth Doing

Take one repeat part. The mill operation is standard 2.2 hours and actually runs about 3.0. You build it 40 times a year.

Thirty-two hours a year, on one operation, on one part number, that the capacity plan never reserved. Now find eight such parts across the shop and you are carrying roughly 256 hours of invisible load, which on a single-shift mill is about six weeks of capacity the plan believes it has and does not.

This is why the phrase "our schedule is optimistic" is usually not about the scheduler. The engine plans exactly what the routing tells it. A low standard does not create capacity; it hides consumption, and hidden consumption surfaces as unexplained lateness at the far end of the month.

Setup Deserves Its Own Look

Setup is where standards drift hardest, because it is the number most often estimated once and never revisited. Setup phases at the kiosk produce setup variance separately from run variance, which matters because they have different fixes. Run time over standard usually means the standard is wrong. Setup time over standard often means the sequence is wrong, and the fix is in changeover sequencing rather than in the routing number. See how sequence-dependent setup shapes a schedule for that distinction.

What Happens After You Correct It

Be ready for the schedule to look worse. It will, and that is the point.

Correcting eight standards upward pushes those work centers up in the Work Center Utilization report, possibly from GOOD into HIGH or from HIGH into CRITICAL. Jobs that looked feasible will show as at risk. Nothing about the plant changed; the plan simply stopped overstating what the plant could absorb.

The value arrives in what you can do next. An overload you can see six weeks out has options: resequence, use an alternate work center, add a shift, or move a promise date while it is still a conversation rather than an apology. An overload you cannot see has one outcome, and it happens on the ship date. See seeing next month's overload this month.

The Limits Worth Stating

Variance data cannot fix a routing on its own. It identifies candidates. Deciding whether the standard should be 3.0 or 2.8, and whether the operation should be re-methoded instead, is engineering work the report cannot do.

It depends entirely on logged actuals. Every one of these reports reads hours that operators logged. Hours entered in a weekly batch produce variance figures with no day-level meaning, and a blank report is almost always a logging habit rather than a bug.

A single large variance is not evidence. Resist changing a standard off one dramatic job. The pattern is the finding; the outlier is a story.

It says nothing about whether the operation should exist. A routing can have perfectly accurate hours for a step that should have been combined with the next one. That question is answered by the ranked routing suggestions report and by process engineering, not by variance.

Correcting standards does not add capacity. It reveals the capacity you actually have, which is often less than the plan claimed. That is more useful and less pleasant. For the wider version of that argument, see finite versus infinite capacity scheduling.

Want to see the variance on your own repeat parts? Bring a year of job history to a demo and we will run the comparison over your routings.