Injection moulding of crates, lids and closures for food and logistics customers, two shifts. 45 staff. 5 machines studied for 4 weeks (7 September to 4 October 2026).
Verdict: OEE 67% across 5 machines (availability 86%, performance 79%, quality 97%). The losses are worth £71,883 a year, £45,681 of it on machines whose extra output would sell; £64,972 a year is recoverable at the targets. Biggest loss: speed and small stops; top logged cause by value of lost time: Hydraulic leak. Modest improvements capture most of this: the limit is what would sell and the hours you could cut, not the machines. Speed losses of 231 hours aren't explained by logged stops: check the ideal cycle times and log short stops.
Lost hours are not the same as lost money. The crate presses lose capacity for about 222,998 crates a year, but the evidence (orders turned away since June) supports selling only about 36,000 more, so their lost time is valued only up to that. The lid presses' lost time is worth only the Saturday overtime it could remove (752 hours a year at most, not every lost hour). The closures press has spare capacity, so only its scrap costs money. That is why modest fixes capture most of the value: the limit is demand and overtime, not the machines. One cheaper option to test before any capital: Saturday overtime on a crate press earns about £94 an hour of contribution for about £32 of cost.
| Machine | Availability | Performance | Quality | OEE | Worth money because | Losses a year | Recoverable |
|---|---|---|---|---|---|---|---|
| Press 1 (350 t): stacking crates | 88% | 80% | 97% | 68% | Extra output sells: up to 18,000 units a year | £22,444 | £20,593 |
| Press 2 (350 t): stacking crates | 77% | 78% | 96% | 58% | Extra output sells: up to 18,000 units a year | £23,237 | £20,831 |
| Press 3 (200 t): lids | 88% | 82% | 98% | 71% | Hours cut: up to 376 a year | £12,702 | £11,366 |
| Press 4 (200 t): lids | 89% | 80% | 96% | 68% | Hours cut: up to 376 a year | £13,090 | £12,059 |
| Press 5 (120 t): closures, 8 cavities, counted in shots | 94% | 75% | 99% | 70% | Spare capacity: scrap only | £411 | £123 |
| Loss | Hours in the study | Share of planned time | £ a year | Aim to recover |
|---|---|---|---|---|
| Breakdowns | 76 | 6% | £12,108 | 30% |
| Changeovers | 82 | 6% | £11,511 | 50% |
| Waiting | 26 | 2% | £3,900 | 50% |
| Speed and small stops | 239 | 18% | £32,142 | 25% |
| Quality (rejects) | 27 | 2% | £12,221 | 30% |
Speed and small stops is the gap between how long the machines ran and the ideal time for what they made. Only 8 of its 239 hours were logged as small stops: the rest is slow cycling or stops too short to log. Before putting money on it, check the ideal cycle times against the machines' own cycle data and time a few hours of running to see where it goes: some of it may be measurement, not loss.
| # | Cause | Lost time, £ a year before limits | Fix |
|---|---|---|---|
| 1 | Hydraulic leak Breakdown · 10 stops, 40 hours | £44,204 | Replace Press 2 hoses and seals; weekly leak check (Maintenance lead, 11 Dec) |
| 2 | Mould change: hunting for tools and clamps Changeover · 20 stops, 26 hours | £17,364 | Tool trolley for each mould with clamps and fittings; shadow board (Shift leader, days, 13 Nov) |
| 3 | Mould change: waiting for the crane Changeover · 8 stops, 13 hours | £14,291 | Book the crane on the schedule; preheat the next mould while the last runs (Production manager, 27 Nov) |
| 4 | Sprue picker robot fault Breakdown · 11 stops, 12 hours | £13,368 | No fix yet |
| 5 | First-off approval wait Changeover · 26 stops, 20 hours | £11,646 | No fix yet |
| 6 | Heater band failure Breakdown · 7 stops, 15 hours | £10,755 | No fix yet |
Causes are ranked by the value of their lost time at each machine's full rate (contribution on the crate presses, overtime cost on the lid presses), before the sales and hours limits, so the figures overlap and don't add up to the total. The hydraulic leaks on the crate presses come first because every hour lost there is a crate the business could sell. Changeovers are the cheapest win: tools ready on a trolley and the crane booked on the schedule cost little and need no capital.
OEE is measured from 4.0 weeks of counts and a stop log, against each machine's ideal cycle time; a slow ideal cycle hides speed losses and a short study can miss rarer breakdowns. The year assumes 47 weeks like the study. Lost time is valued at contribution (price less materials and other costs that vary with output) only on machines whose extra output would sell, with evidence, and only up to the extra units a year that could be sold; elsewhere at the cost an hour you would actually save, up to the hours a year that could really be cut. Scrapped units also cost their material. The recovery targets are the team's assumptions, not forecasts. Valuing every machine at contribution would overstate the prize: extra output is only worth money if customers buy it.
Quillmoor chose the guided route. At the final session the agent was set up in an AI assistant the business already uses, with read-only access to the press counter export and the operators' stop log sheets. Every Monday at 7:00 it loads last week's counts and stops into a new dated copy of the study, runs the checks and sends the production manager a short note: each press's OEE against its study figure, the stops with no cause, and any fix past its due date. It uses the rules on Inputs: a press is flagged 5 points below its study OEE, a week when more than 5% of stop minutes have no cause, and a fix is judged 4 weeks after it is marked Done. Nothing left Quillmoor, and no decision changed, without one of its own people approving it. These events continue the fictional example.
| Date | What the agent found | What was decided |
|---|---|---|
| 26 Oct 2026 No flag | The first week: 66.1% across the five presses, against 66.6% in the study. | Nothing needed. |
| 2 Nov 2026 Flag | Press 2 at 49.8%, 8 points below its study figure: two hydraulic leak stops. | The hose and seal replacement was brought forward from 11 December to 21 November. |
| 9 Nov 2026 Flag | 11% of the week's stop minutes had no cause, all on nights on Presses 3 and 4. | The cause went on the night handover sheet; two stops turned out to be colour purges. |
| 16 Nov 2026 Flag | The dryer fix was past its 6 November due date and not started. | A new date of 27 November; dryer loading went on the night checklist. |
| 14 Dec 2026 No flag | Four weeks after the tool trolleys: 168 minutes a week hunting for tools and clamps, against 391 in the study. | Nothing needed. Trolleys for Press 5 next. |
| 21 Dec 2026 No flag | Four weeks after the new hoses: no leak stops on Press 2, and an average OEE of 63.9% against 57.8%. | Nothing needed. The weekly leak check stays. |
At the 30-day check-in on 16 November, Geoff and the production manager read the log together. No rule needed changing.
What the agent added. The method is unchanged: same spreadsheet, same checks, same valuation. The agent made sure OEE was measured every week, the same way as the study, so the Press 2 leaks were caught in their first week and each fix was checked against the study's own figures, not memory. Four weeks of better figures are a good sign, not proof: the next full study, on the same presses, is the test. No £ figure is claimed for the months after; the study's figures above are unchanged.