# Late Start / Early Finish — what our own reports already record **Source:** 1,267 vessel productivity reports, 2024–2026, as held in the productivity dashboard dataset (`shifts.db`, 16,460 crane-shift records across 7,680 vessel-shifts). **Prepared:** 28 September 2026 --- ## The central point **"Late Start / Early Finish (Hrs)" is not a new concept and not an opinion. It is a standing column in all four of our productivity report templates** — Break-Bulk, Bulk, RoRo and Rev.04 — verified against the source files: | Template | Column I header | |---|---| | Break-Bulk Productivity | `Late Start / Early Finish (Hrs)` | | Bulk Productivity | `Late Start / Early Finish (Hrs)` | | RoRo Productivity | `Late Start / Early Finish (Hrs)` | | Productivity (Rev.04) | `Late Start / Early Finish (Hrs)` | It is filled in shift by shift by the people on the quay. We hold **16,460 crane-shift records** of it. The structured observation has been running for years — it has simply never been read back to us as a total. --- ## What it shows (2024–2026) | Measure | Value | |---|---| | Vessel-shifts with late start / early finish recorded | **54.9%** | | Crane-shift records showing it | 47.5% (7,811 of 16,460) | | Total time lost | **1,658 crane-hours** | | Equivalent to | **207 full eight-hour crane shifts** | | Share of all recorded delay time | **12.1%** (of 13,687 h) | | Average where it occurs | 12.7 minutes per crane-shift record | ### By cargo type — where it hurts most | Template | Rank among delay causes | Share of that template's delays | Hours | |---|---|---|---| | RoRo | **#1 of 9** | 47.4% | 63 h | | Break bulk | **#2 of 9** | 27.4% | 651 h | | Rev.04 (general) | #4 of 12 | 8.8% | 645 h | | Bulk | #6 of 9 | 7.8% | 299 h | On break bulk, 71% of crane-shift records carry a late start or early finish — the highest of any cargo type. ### By year | Year | Records | % with late/early | Hours | Share of delays | |---|---|---|---|---| | 2024 | 4,912 | 63.6% | 759 h | 17.5% | | 2025 | 6,470 | 37.1% | 456 h | 9.1% | | 2026 | 5,152 | 45.1% | 447 h | 10.2% | 2024 was materially worse. 2025 improved. **2026 is trending back up** (37.1% → 45.1%), which is the argument for acting now rather than assuming the 2025 improvement holds by itself. ### By shift Effectively flat — day 48.0%, night 50.3%, afternoon 45.3%. **This is not a night-shift problem.** It is systemic across all three shifts, which points at handover and gang allocation rather than at any one crew. --- ## Honest limitations These matter, and should be stated rather than discovered later: 1. **The column merges late start *and* early finish.** The dataset cannot separate the two. The procedure's value is partly that it will. 2. **This is a floor, not a ceiling.** The figure depends on tally clerks recording it. The same culture in which a late start passes without comment also under-records it. The true number is very likely higher. 3. **Hours are crane-hours,** not gang-hours or vessel-hours. A late start affecting four cranes counts four times — correctly, since four cranes stood idle, but it is not the same as "1,658 hours of vessel delay". 4. **Records outside 2024–2026 were excluded** (a small number of rows carry corrupt dates such as 1900 and 2028). Filtered range: 2024-01-01 to 2026-12-31. 5. **"Delay due to CSO" is the largest category overall** in three of four templates. Late start / early finish is not the single biggest delay we have — it is the biggest one that sits **entirely inside our own control** and requires no counterparty to fix. --- ## What this means for the procedure The objection that the start-and-stop-times work is "based on personal opinion and not on observation from the gemba" is answered by our own data: - The observation exists, is structured, and has 16,460 records behind it. - It has been collected at the gemba, by the people doing the work. - It shows the problem in more than half of all vessel-shifts. - It is worst on break bulk and RoRo, and it is getting worse again in 2026. The procedure does not replace observation. **It makes the observation usable** — by separating late start from early finish, attaching a reason code to the gap, and putting the number in front of the people who can close it, in the same week rather than a year later.