GUIDE · OUTPUT
Shot counts vs pieces:
checking output on multi-cavity moulds.
Published · ProdCrest
Short answer. On a multi-cavity mould, pieces = shots × cavities that are actually running. Record the shot count and the active cavity count as two separate fields, keep the cavity count in the product master with a date, and test every shift’s reported output against the machine’s theoretical maximum. A ratio far below normal usually means shots were entered as pieces; a quiet overstatement usually means a blocked cavity nobody wrote down.
Why the counter and the boxes disagree
A moulding machine counts cycles, not parts. Each cycle (a shot) fills every open cavity in the tool, so an 8-cavity tool makes 8 parts per shot. Output goes wrong in three ordinary ways:
- Shots typed as pieces. Someone reads the counter and writes the number straight into the “pieces” column. Output is understated by the cavity factor.
- A blocked cavity that is not recorded. One cavity is closed off for a defect, but the sheet still multiplies by the full count. Output is overstated, and the boxes come up short later.
- An outdated cavity count in the master. The tool was modified or replaced, and the master still holds the old number. Every conversion after the change is wrong in the same direction.
The first one is large and easy to spot. The other two are small, steady and easy to miss, which is why the check needs a written method rather than a glance at the total.
The capacity check
For each shift row, compare what was reported with what the machine could have made:
theoretical pieces = run minutes × 60 ÷ cycle time in seconds × cavities
ratio = (good + scrap reported) ÷ theoretical pieces
Run minutes are planned minutes minus recorded downtime. A healthy row sits a little below 1, because real cycles run slightly slower than standard and short stops go unrecorded. Agree the limits before you look at the data. The ProdCrest demo treats a ratio below 0.5 as a likely shot count and above 1.1 as impossible output that needs an explanation.
A synthetic example: one machine, one week, entered in shots
The ProdCrest demo uses synthetic data: an invented moulding plant with four machines, three shifts and four weeks of records. Every figure below was produced by that dataset; none is a customer result.
Take one shift row from machine IM-02, running Cap 28mm on an 8-cavity tool with a 13-second cycle:
| Step | Value (synthetic) |
|---|---|
| Planned minutes − downtime | 480 − 15 = 465 run minutes |
| Theoretical pieces | 465 × 60 ÷ 13 × 8 ≈ 17,169 |
| Reported good + scrap | 1,642 + 405 = 2,047 |
| Ratio | 2,047 ÷ 17,169 ≈ 0.12 |
| Good pieces if the 1,642 were shots | 1,642 × 8 = 13,136 (estimate) |
Across the whole synthetic period, rows that were entered correctly sat between 0.78 and 0.95. Twenty-one rows fell below 0.5, and all 21 had the same pattern: machine IM-02, every shift, 10 to 16 August. Read as pieces, they added up to 26,234; converted by the cavity count, about 168,000. Those rows explain 7 of the 17 work orders where the spreadsheet and the ERP disagreed (see shift-log vs ERP totals).
A pattern like this points at a method (how that machine’s counter was read that week) rather than at a person. The useful question is what changed on IM-02 on 10 August, not who typed the numbers.
The blocked-cavity case
The capacity check does not catch a single blocked cavity, because the ratio stays in the normal band. Simple arithmetic shows the size of it: on a 12-cavity tool, 1,000 shots with one cavity blocked make 11,000 parts. A sheet that multiplies by 12 reports 12,000, which is 1,000 parts, or about 8%, that do not exist. Nothing in the totals looks strange; the shortfall appears later, at packing or dispatch.
The record has to carry it. Two habits close the gap:
- An active cavities field on the shift record, filled in at every shift start and whenever a cavity is blocked or reopened, with the time.
- A box count once per shift (or a weigh count with a confirmed part weight) compared with shots × active cavities. A gap of a few percent that repeats is worth a look.
The method
1. Record shots and active cavities separately
Let the sheet do the multiplication. If people write pieces directly, nobody can later tell which factor they used.
2. Keep the cavity count in the product master, with a date
Cavities, standard cycle time and part weight belong in one table, each with the date it was last confirmed. A tool change is also a master-data change.
3. Run the capacity check on every row
Compute the ratio per shift row, not per month: a monthly total hides a week of shot counts among three weeks of correct data.
4. Correct openly
When a row is clearly in shots, convert it, mark it as an estimate and keep the original value next to it. Confirm with the people who ran the machine before the figure goes into a report.
5. Look for the pattern, then fix the form
Group flagged rows by machine, shift and date. A cluster usually leads to a form, a counter display or a handover habit that can be changed so the error stops at the source.
Short version: “Twelve cavities, eleven running”, a 30-second fictional factory story on YouTube.
Limits
The ratio cannot tell a slow cycle from a blocked cavity when both keep it inside the normal band, and a correction by the cavity count is an estimate until the plant confirms it. The demo does not model cycle-time or cavity changes within the period. None of this explains why output was low; it makes sure the number you discuss is the right one.
See the same check in the interactive demo, or read the full list of checks for production records. ProdCrest runs checks like these as part of a production data review. Describe your situation.