The short answer
- Schedules fail because they are built on rated speed, not measured speed
- Measure pieces per shift including setup, not prints per hour
- Plan to about 80% of capacity — the remainder absorbs the week's surprises
- Sequence by ink and garment colour to remove changeovers you do not need
- When it breaks, move the job with the most slack, not the smallest
Most decorating shops schedule by writing jobs on a board in roughly the order they arrived, then spend the week reacting. That works until two things collide, at which point somebody makes a decision under pressure and a customer who was promised a date finds out late.
A schedule that survives a bad week is not a more detailed schedule. It is one built on real numbers with room in it.
Why schedules fail
| Cause | What it looks like |
|---|---|
| Rated speed, not real | Plan says 320/hr, shop does 210 |
| Setup not counted | A 24-piece job takes as long as a 100-piece one |
| No slack | One late delivery moves everything |
| Approval time ignored | Job scheduled before the proof came back |
| Rush jobs inserted | Three other jobs quietly pushed |
| The dryer is the constraint | Press planned, dryer full |
The first row is the root of most of it. A press rated at 320 pieces an hour achieves that while running; across a shift including makeready, colour changes, reloading and the inevitable interruptions, the real figure is much lower. Planning on the rated number guarantees a schedule that cannot be met.
The last row catches shops that have upgraded a press without upgrading the dryer. A press that prints six times faster feeding a dryer that has not changed has simply moved the bottleneck, and the schedule still needs to be built on whichever machine is slowest.
A schedule built on rated speed is not optimistic. It is arithmetic that was never going to work.
Measure your real capacity
Not from a spec sheet and not from a good day. Take a fortnight of actual output and divide by shifts worked.
What to record for two weeks
- Pieces produced per shift, by process
- Number of setups per shift, not just jobs
- Time from job start to first good print
- Dryer throughput at the belt speed you actually run
- Hours lost to waiting — approvals, deliveries, breakdowns
Two numbers come out of that: pieces per shift and setups per shift. Those are what a schedule should be built from, because they include everything that actually happens rather than only the printing.
Do it separately for each process. Screen, DTF and embroidery have completely different setup profiles, and a single "capacity" figure across all three is meaningless. Water-based work in particular runs at roughly a third of plastisol throughput because of the cure time, and a schedule that ignores that will be wrong every time it appears.
Build in slack
Plan to about 80 per cent of measured capacity. The remaining fifth is not waste — it is what absorbs the late delivery, the reprint and the rush that will arrive.
A shop planned to 100 per cent has no capacity to absorb anything, so every surprise displaces a customer. A shop planned to 80 handles most weeks without anyone noticing, and the slack is used almost every week — it simply does not have to come out of somebody's delivery date.
| Planned at | A late delivery means |
|---|---|
| 100% of capacity | Another job moves; a customer is told |
| 90% | Recoverable with overtime |
| 80% | Absorbed; nobody outside notices |
Owners resist this because idle capacity looks like waste. It is insurance, and its value shows up as an on-time rate that does not move when something goes wrong.
Sequencing the week
Order matters as much as load. Two jobs needing the same ink or the same garment colour run far faster together than separated by something else.
Sequence by
- Ink colour — group jobs sharing a colour to avoid re-mixing
- Garment colour — all darks together, then lights
- Process — batch water-based work rather than interleaving it
- Mesh count, where screens can be reused across jobs
- Due date, as the constraint rather than the sort order
Due date should be the constraint, not the sequence. Sorting purely by due date produces a week of changeovers; sorting by ink within due-date groups produces the same deliveries with materially less setup.
Schedule maintenance as a job too. A machine blocked for a scheduled check is capacity you knew about; the same machine down unexpectedly is capacity you did not, which is why maintenance belongs on the board.
When it breaks anyway
It will. The question is who absorbs it, and the instinct — move the smallest job — is usually wrong.
Move the job with the most slack: the one whose due date is furthest away, or whose customer has the least at stake on the exact day. A small job due tomorrow for an event is a worse candidate than a large job due next Friday for stock.
When something has to move
- Identify which jobs have genuine slack, from the due dates
- Move the one with most, not the one that is easiest
- Tell that customer before they notice, with a new date
- Consider a partial delivery — often it solves their real problem
- Record what caused it, so the pattern becomes visible
Telling the customer early is the whole difference. A delay communicated three days ahead with a new date is a manageable inconvenience; the same delay discovered by the customer on the due date is a failure. That is the same discipline as any proactive update, applied to bad news.
The daily review
Ten minutes each morning, the same way every day, with whoever runs production.
Four questions
- What is due in the next three days, and is anything at risk?
- What is waiting on somebody else — approval, delivery, artwork?
- What is on each machine today, in what order?
- What arrived yesterday that has not been scheduled?
The second question is the one that finds problems earliest. Jobs waiting on a customer approval are invisible on a production board — they are not late yet — and they are the commonest cause of a comfortable schedule becoming a rush that was nobody's fault.
Software helps most here by flagging jobs at risk before they are late. That is the feature worth having in a scheduler — not the calendar view, but the list of things that will miss unless somebody acts today.
What to measure
| Number | Tells you |
|---|---|
| On-time rate | Whether the schedule is real |
| Pieces per shift | Your true capacity, updated |
| Setups per shift | How much of the day is changeover |
| Jobs moved, and why | Where the disruption comes from |
| Hours waiting on approvals | Usually larger than anyone thinks |
Review those monthly and adjust the capacity figure the schedule is built on. Real capacity changes — new equipment, new people, a different job mix — and a schedule still using last year's number will drift back into optimism without anybody deciding to. The rest of the production archive assumes your plan is built on measured numbers with slack in them.



