Closing Out a Finished Job: From the Last Punch to the Variance Review

Closing out a job is two acts, not one: you confirm the work is genuinely finished and lock the record, then you read the hour variance while the details are still fresh enough to explain. This walkthrough follows one order through both halves in EDGEBIC by User Solutions, from the last kiosk punch on Thursday afternoon to a routing correction filed Friday morning. It is one of the worked examples we publish so the closing routine is visible rather than assumed.

Most shops do the first half and skip the second. The second half is the one that pays.

The Job

JOB-4507, 120 hydraulic brackets for a repeat customer, five operations, planned at 46.5 hours. The last operator punched out of inspection at 15:40 Thursday.

The job is physically done. The record is not, and the difference matters.

Step 1: Verify Before You Close

Open Job View and select the job. Read down the operation grid.

Four steps show an actual end. Step 30, deburr, does not. Its logged hours read 6.0 against a 6.5 hour plan, so it satisfies neither completion test: no end is stamped, and the logged hours do not cover the plan.

This is the most common reason a finished job refuses to report finished. An operation counts as done when its actual end is stamped, or when its logged hours fully cover its planned hours. Half an hour short on both tests leaves it open.

Two honest fixes, and they are not equivalent.

If the deburr genuinely took 6.0 hours, stamp the actual end and let the record say the step ran half an hour under plan. If the operator forgot to punch the last stretch, log the missing time first and then stamp. Do not inflate the hours to force the step closed, because that hour lands in the variance you are about to read and quietly corrupts it.

Halden's supervisor confirms the deburr really ran short. Double-click the row's Actual End cell, use End = Now or type the real finish, and save. If you change values that were reported before, the save asks for a reason and stores it in the logging history, which is why "operator finished 30 min early, no rework" is worth typing instead of "fix".

Step 2: Stamp the Job Complete

With every operation done, click Complete this Job on the Job View tab. The Mark Complete dialog asks for the completion date and an optional reason. Click OK.

Three things change immediately. The header reads that the job is completed on that date. The Log Actual and Complete this Job buttons lock, with a note that you can no longer log actuals on this job. The job's bars take the completed shade on the Gantt.

One rule to know before you click. If completion is recorded without an explicit end time, the actual end snaps to the end of the last day that has logged hours, at 23:59 of that day. A job whose hours stop on Wednesday closes on Wednesday, no matter which day somebody pressed the button. That rule is what makes it impossible for a same-day job to finish before it started.

If you close a job too early, Reopen Job reverses it and re-enables actuals entry. Nothing is lost; the banner clears and you can correct and close again. The procedure is in how to mark a whole job complete.

Step 3: Read the Variance Before You Forget

This is the half most shops skip, and it takes eleven minutes.

Open Reports and run Work Center Progress, filtered to the job. One row per routing step with net hours, actuals, pieces made, and percent complete. The table above is what it shows.

Read it as a distribution, not a total. The job ran 52.0 against 46.5, an 11.8 percent overrun. Stated that way it sounds like the whole estimate is optimistic and the natural reaction is to pad every routing.

The step breakdown says something completely different. Saw, deburr, weld and inspect land within half an hour of plan, which on a 120 piece order is noise. The mill step alone ran 29.0 against 24.0. Five of the 5.5 hours of variance sit on one operation.

Now ask the specific question. Twenty-four planned hours on the mill is 1.2 hours of setup plus 0.19 hours per unit. The operator says the setup on this fixture is closer to three hours since the fixture change last quarter. Three hours minus 1.2 is 1.8, and the remaining 3.2 hours spread across 120 pieces is about 1.6 minutes per part, which is inside normal drift.

The finding is not "our estimates are 12 percent optimistic". The finding is "the mill setup on this routing is roughly two hours light, and it has been since the fixture change". That is one number to correct on one step, and it improves every future job on this product.

Step 4: The Efficiency Read

Run Earned Value for a second opinion on the same job.

At completion, SPI settles at 1.00 by definition, because earned value equals planned value once everything is done. Mid-job the two indexes can point opposite ways, which is the case worked through in earned value mid-job, ahead of schedule and over budget at once. The meaningful number at close is CPI, which is earned value divided by actual cost in hours: 46.5 divided by 52.0 gives 0.89.

A CPI of 0.89 means the job earned 89 hours of plan for every 100 hours it burned. Read across a run of closed jobs, that number is the honest answer to "are we quoting our own capability". One job at 0.89 with a known single cause is a routing fix. Ten jobs at 0.89 with ten different causes is an estimating problem.

If your work centers carry an hourly rate, the report adds dollar columns automatically. At Halden's mill rate of 78 dollars an hour, five overrun hours is 390 dollars on one order, and the same understated setup across the 22 orders of this product last year is a number worth taking to the routing review. Every column's definition, formula and example values are one click away under Column Details, which is the fastest way to stop two people reading SPI differently.

Step 5: Close the Loop

Three actions come out of the review, and they take longer to describe than to do.

1. File the routing correction. Raise the mill step's setup from 1.2 to 3.0 hours on the master routing. Running jobs are unaffected, because each scheduled job carries its own routing snapshot and re-plans against that rather than against a master edit.
2. Check the audit trail once. The Job Audit Trail report reconstructs the whole story of one job: its original schedule, every reschedule, work center swaps, daily actuals, and routing versions. Thirty seconds here answers "why did this job move twice in June" before a customer asks.
3. Do nothing else. A closed job with a known cause and a filed correction is finished. Resist reopening it to make the numbers prettier.

What the Closeout Is Worth

The completion is bookkeeping. Any system can do it.

The variance review is the part that compounds. A shop that reads five closed jobs a week is feeding real hours back into its routings continuously, which means next quarter's promise dates are computed from what the plant actually does rather than from estimates somebody typed in once. That is the same discipline as logging daily instead of weekly, applied at the other end of the job.

For the daily version of this rhythm see a day in the life of a planner, for the hours that feed it see how to enter actual hours for yesterday, and for the metric itself see what earned value means in manufacturing.