Sequencing Stain and Finish Work

In a finish room, the cost of the next job depends entirely on the last one. EDGEBIC by User Solutions holds those costs in a setup matrix per work center, so a light-to-light changeover can be zero, a light-to-dark can be an hour, and a dark-to-light can be four hours, and the schedule uses the right number every time. If the underlying idea is new, start with what a setup family is.

The Pain: One Setup Number for Every Changeover

Every routing field for setup time assumes a single value. For a mill or a saw that is usually close enough. For a spray booth it is not close to anything.

Ask a finish room supervisor what the changeover costs and you will get a table, not a number. Same color, next job: no changeover at all. One light stain to another: rinse and go. Light to dark: mask, purge, extra coat. Dark back to light: solvent flush, color match, test panels, and most of the morning.

A scheduling system holding one number per routing step is wrong twice. It charges 30 minutes for a changeover that costs nothing, and it charges 30 minutes for one that costs four hours. The second error is the one that hurts, because it is the one that makes the plan a lie.

The consequence is familiar in almost every furniture shop. The finish room is always behind, everybody knows the finish room is always behind, and no report ever shows why. The general shape of this problem across the industry is covered in the furniture and woodworking scheduling overview.

Scenario A: What a Flat Setup Actually Produces

Take one booth on a day shift, three jobs sorted by due date, and a flat 30-minute setup on the routing.

With 30 minutes of setup each, the plan reads:

08:00  first  (1.5 h)  09:30
09:30  second (2.0 h)  11:30
11:30  third  (1.25 h) 12:45

Three jobs, done before lunch, plenty of shift left.

The floor pays something else entirely. Cold start into white: 30 minutes, which the plan got right. White to black: 60 minutes. Black to white: 240 minutes. Add the spray time and the day runs to roughly 11 hours.

The plan did not miss by a margin. It missed by half a day, in the same direction, every single week. And because the miss lives inside setup rather than inside run time, no efficiency report catches it.

Scenario B: The Matrix, Same Sequence

Now the planner configures the booth's matrix. Products are grouped into families first, because that is what keeps the data maintainable.

The family matrix for that booth, in minutes:

Run the same three jobs in the same order. At each step the engine asks what the booth ran most recently and looks up the cell.

08:00  first  (1.5 h)   09:30
09:30  second (2.5 h)   12:00
12:00  third  (4.75 h)  16:45

The plan now runs to 16:45 and spills past the shift. Setup total: 330 minutes, which matches what the floor actually pays.

That is not a worse plan. It is the first honest one. The planner can now see the problem in the schedule rather than discovering it in the booth, and can decide what to do about it.

Scenario C: Sequencing Against Real Costs

Reorder to run the lights together before the dark job.

08:00  first  (1.5 h)   09:30
09:30  third  (0.75 h)  10:15
10:15  second (2.5 h)   12:45

All three jobs inside one shift with more than three hours to spare. Setup drops from 330 minutes to 90, a 73 percent reduction, purely from the order of the queue.

The comparison across all three scenarios is the argument in one table:

Scenario A and Scenario C show the same finish time. Only one of them is real.

How the Lookup Resolves

The engine walks a fixed chain and takes the first hit.

1. Cold start. No previous product is known, so there is no changeover, but the machine still needs its initial preparation. The routing's setup value is used, not zero.
2. Same product. Zero. Running the same finish back to back costs nothing.
3. Product-level cell. A specific from-and-to product pair, if one has been entered for that work center.
4. Family-level cell. The families of both products are looked up, and the family pair's cell is used.
5. No cell found. The routing setup value, or the work center default.

Matrix values are entered in minutes, which is how finish room supervisors actually talk, and converted to hours internally. The resolved value is also what the schedule uses for the operation's real setup, and it feeds the overlap calculation if the step also streams work downstream. Full mechanism detail is in the setup matrix explained.

Building the Matrix Without Drowning in Cells

The reason to work at family level is arithmetic. A shop with 100 finishes has close to 10,000 ordered product pairs per booth. The same shop with eight families has 64 cells, and 64 cells is an afternoon.

A practical build order:

1. Name the families the way the floor does. Lights, darks, clears, stains, glazes, specials. If your supervisor already sorts a whiteboard into groups, those are your families.
2. Assign every finish to a family. This step is the one people skip, and it has a consequence worth knowing: the family lookup requires both products to have a family. If only one is assigned, the lookup silently falls through to the routing default and you never find out. When the matrix seems not to be firing, unassigned products are the first thing to check.
3. Fill the family grid, per booth. Different booths have different physics. A booth with a dedicated dark gun does not pay the same purge as one without.
4. Enter the zero cells deliberately. Two different products inside the same family need an explicit family-to-family cell set to zero if you want zero. Without it the routing default applies.
5. Add product overrides sparingly. One notorious pair that behaves differently from its family is worth an override. A hundred of them means your families are wrong.

Negative values are rejected, and the engine clamps defensively, so a typo cannot produce a negative changeover.

Changeovers on Jobs Already Running

A finish room reschedules constantly, and the matrix has to behave when it does.

The rule is that a changeover already paid is historical fact and is never recomputed. When a job is rescheduled, operations already logged as complete keep the setup they actually consumed. What the engine then needs is an anchor: for the first remaining operation on each booth, what did that booth physically run last?

That anchor comes from the last completed operation on the booth rather than from a projection. So a booth that finished a dark job on Tuesday afternoon starts Wednesday's plan with dark as its previous product, and a light job scheduled next correctly carries the full dark-to-light changeover. Without the anchor the engine would either treat the reschedule as a cold start, undercharging the change, or reuse a stale value from a plan that no longer applies.

Two smaller behaviors follow from the same design. The resolved changeover for a step is cached for the duration of a run, so if that step also overlaps work downstream, the overlap calculation uses the same figure rather than re-deriving a different one. And a machine mirroring another in lockstep inherits the parent's changeover by definition rather than resolving its own, which is correct: two chambers running as one process share one changeover.

What Changes Once the Numbers Are Real

Three things, in order of how quickly you notice them.

The plan stops lying. Finish room dates become dates you can quote from. That alone usually repays the configuration effort.

Sequencing becomes a measurable decision. Grouping like finishes together stops being an instinct the supervisor applies when they have time and becomes a visible tradeoff between changeover hours and due dates.

The optimizer has something to work with. Once each transition carries a real cost, minimizing total setup across a queue becomes a well-posed problem rather than a guess. See optimizer goals and presets.

The same booth logic travels to any shop where finish decides the schedule; scheduling musical instrument manufacturing around finish cure and hand work pairs it with the cure time and the few people qualified to do final setup.

One boundary worth stating. Grouping by finish family and hitting due dates pull in opposite directions, and no configuration removes that tension. What the matrix does is make the tradeoff explicit so a human decides it, rather than the plan hiding it and the floor absorbing it.

Where to Start

Pick your worst booth. Build four or five families, fill one grid, and reschedule. Compare the new finish room dates against last month's actual performance rather than against last month's plan, because last month's plan is the thing being corrected.

For make-to-order timing on the same jobs, see scheduling custom furniture to a promised date. For the industry picture, see the furniture and woodworking scheduling overview and the furniture manufacturing software page. For the rest of this series, see EDGEBIC by industry, and for the product, EDGEBIC.