Weekly Report for Dispatcher
- Workspace
- Patient Transport, Mu Hospital
- Week
- Monday 3 to Sunday 9 August 2026
- Written
- Monday 10 August 2026, after the night’s run
- Owner
- Thanicha Mahaworakij, Logistics Manager
Mu Hospital does not exist. It is the test hospital Mu was built and trained on: about five months of porter jobs, from April to August 2026, shaped like a real hospital’s, with no real patient or porter in them.
Everything else on this page is real. Every figure is what Class 26’s pipeline produced when it was run by the week on that data, and what the Skills built beside it found. Nothing was typed in to make the story read better.
Recap
Most of this week went the way its own past said it would. Of its 168 hours, 139 were Under Controlled or Elevated: 63 and 76. Their jobs were answered as fast as, or faster than, the same hours have been answered before. 29 hours were not. 17 were Stressed and 12 Critical.
Every hour is graded by the share of its jobs answered within their bar. Over 75% is Under Controlled, 50 to 75% Elevated, 30 to 50% Stressed, under 30% Critical.
The bar is the pace 60% of jobs met in the most similar hour of the past. So an hour that runs exactly as its past did lands near 60%, which is Elevated. Elevated is not a warning. It means the hour kept pace with itself. Under Controlled means it beat its own past.
Below, the week’s 168 hours, Monday at the top, midnight on the left.
Under Controlled 63Elevated 76Stressed 17Critical 12
The week turned on Monday morning. From 09.00, jobs arrived faster than they were accepted, and by 10:40 there were 45 waiting. It was the longest queue I have seen since May, and it took until 12:57 to clear. Monday 10.00 was the hardest hour of the week: three of its 75 jobs were answered within their bar.
The line is the number of jobs waiting to be accepted, minute by minute, from 08.00 to 13.30. The steps beneath it are each hour’s own limit, the queue that hour’s Baseline day stayed at or under for three quarters of its minutes.
It crossed its limit at 09:11 and did not come back under it until 12:57.
Waiting to be acceptedThe hour’s own limit
It began an hour earlier than it looked. At 09.00 nine porters faced 62 jobs, almost seven each, past the point where this team starts to slow. Jobs were handed to porters who were still on the one before, and waited behind it. Three more porters came at 10.00, but they came into a queue of 22 that was still growing. 10.00 carried the same load as the Monday I compare it with, 6 July, and on 6 July jobs waited three minutes to be accepted. On 3 August they waited 31, because 10.00 inherited 09.00.
The same hour, 11.00, was Critical on Monday, Tuesday and Wednesday. After that the week settled. On Tuesday at 09.00 nine porters answered 43 of 45 jobs within their bar. Of the hours that asked as much of each porter, none did better.
Service Responsiveness
| Measure | P50 | P80 | Bar (P60) | Within the bar | Past Best (P40) |
|---|---|---|---|---|---|
| Response | 4:58 | 12:44 | 5:17 | 61.5% | 3:53 |
| Completion | 13:31 | 21:08 | 14:09 | 54.9% | 11:34 |
| Assignment | 0:46 | 1:55 | 0:56 | – | 0:34 |
| Cycle | 10:45 | 17:15 | 11:54 | – | 9:29 |
Times in minutes and seconds. I judge every job against its own hour’s past, not against one target for the whole week. For each hour I find the day in the past year that looked most like it: the same weekday and hour, carrying about the same load. The bar is the time 60% of that day’s jobs were answered within.
For each hour I look back over the past year for the same weekday and hour. Of those, I keep the one whose load was closest: jobs for each porter on duty. Where loads tie, I take the closest in jobs, then in pressure, then the most recent.
The day I find is the hour’s Baseline. Its jobs set the bar: the pace 60% of them met. Monday 10.00 was matched with 6 July 10.00.
| Monday 3 August 10.00 | Its Baseline, 6 July 10.00 | |
|---|---|---|
| Jobs | 75 | 74 |
| Porters | 12 | 12 |
| Jobs per porter | 6.25 | 6.17 |
| Chosen by | Closest job load | – |
| Bar for its jobs | 10:14 | – |
So a week that runs exactly like its past lands near 60%. This one reached 61.5% on Response, a little better than its own history, and 54.9% on Completion, a little worse.
A fixed target, say every job answered within ten minutes, judges a quiet night and a full Monday morning by the same line. A team can meet it at 03.00 with nobody waiting and miss it at 10.00 doing the best work of its week.
So I judge each hour against its own past instead. Because the bar is the pace 60% of the past hour’s jobs met, a week that repeats its past passes about 60% of its jobs. More than 60% means the week beat its own history; less means it fell behind it. The number to watch is not whether it reached 100%, but which way it moved.
Where a contract sets a minimum the bar never goes below it, so a team that is already fast is not held to an impossible line. This week’s typical bar for a busy hour was 6:00.
Half of all jobs were answered within 4:58. The slowest fifth waited longer than 12:44. Past Best, 3:53, is the pace of the fastest 40% of everything before this week. It is there to show what this team has already done, never to grade it.
Assignment and Cycle have no share: neither is held to a bar. Assignment stayed quick all week, under a minute for half of all jobs. The time lost on Monday was not in dispatching. It was in the wait after a job reached a porter who was still busy with the one before.
Workforce Performance
| Measure | This week | The eight weeks before |
|---|---|---|
| Jobs completed | 4,227 | – |
| Cancelled | 177 (4.0%) | – |
| Porters who worked | 23 | – |
| Jobs per porter per day | 35.5 | 33.7 to 35.2 |
| Effort, average porter | 65.6% | – |
| Offers declined (simulated) | 2.6% | 2.4% to 2.6% |
| Porters on duty, average staffed hour | 8.0 | – |
23 porters completed 4,227 jobs. Each carried 35.5 a day on average, more than in any of the eight weeks before.
Effort is the share of a porter’s time on duty spent on jobs. The average porter’s was 65.6%. Hour by hour it was lighter than that sounds: in half of the week’s staffed hours, Effort stayed under 50%, and in three quarters under 75%.
Effort is the minutes porters spent on jobs, from accepting one to closing it, over the minutes they were on duty. Only porters who touched a job in the hour count as on duty: the test hospital has no roster, so a porter waiting for work with nothing to do is not seen.
A porter who takes the next job before closing the last counts both, so an hour’s Effort can pass 100%. It did once this week, Monday at noon. Counted once per minute, that hour was below 100%.
Effort, each hourPast 100%: one porter's jobs overlapping
The test hospital records when a job reached a porter and when it was accepted, but never a decline, so the rate of declined offers is my simulation. A porter declines more often the busier their last hour was, set so that the weeks before decline about 2.5%, as real Dispatcher hospitals do. This week came to 2.6%, the most of the nine, and Monday from 09.00 to 12.00 to 3.2%. Nearly every job that waited long had gone to a porter still on another one: they were not saying no, they were still busy.
The heaviest hour was Monday 12.00: 12 porters, 54 jobs, Effort 117%. Effort only passes 100% when one porter’s jobs overlap, and on Monday at noon they did. That was the hour the queue began to clear.
Demand Management
| Measure | This week |
|---|---|
| Typical queue (P50) | 3 jobs |
| Controlled ceiling (P75) | 5 jobs |
| Peak queue | 45 jobs, Monday 10:40 |
| Controlled hours (queue) | 134 (80%) |
| Stressed hours (queue) | 25 (15%) |
| Critical hours (queue) | 8 (5%) |
| Times past the hour’s own limit | 6 |
| Recovery, typical (P50) | 10.5 min |
| Fastest recovery | 7 min, Saturday 11:16 |
| Longest recovery | 125 min, Monday from 10:51 |
The queue is every job that has been asked for and not yet accepted. I read it hour by hour, by the most that waited at once. In a typical hour it reached three. Three hours in four kept it to five or fewer, and that is its controlled ceiling. On Monday at 10:40 it reached 45.
Each hour also has its own limit: the queue its Baseline day stayed at or under for three quarters of that hour. Held to it, 134 hours kept their queue under control, 25 went over, and eight went at least twice over. Six times this week the queue rose past its own limit to 10 jobs or more, at least twice the limit. Five of them were back under it within half an hour of their peak, and half within 10.5 minutes.
The sixth was Monday. The queue passed its own limit of seven at 09:11, reached 45 at 10:40, and was at 45 again at 10:51. It took another 125 minutes to come back. Nothing else this week came close.
Key Moments
A habit: 11.00, three days running
11.00 was Critical on Monday, Tuesday and Wednesday. When the same hour goes hard three days in a row, I stop calling it bad luck. On Tuesday and Wednesday the queue passed its own limit inside that hour, at 11:09, 11:50 and 11:25, though never for long. I will watch whether 11.00 holds its habit next week.
The hardest morning: Monday, 09.00 to 12.59
| Monday 10.00 | This hour | Its Baseline, 6 July 10.00 |
|---|---|---|
| Jobs | 75 | 74 |
| Porters | 12 | 12 |
| Jobs per porter | 6.25 | 6.17 |
| Waiting to be accepted, average | 31 min | 3 min |
| Response, average | 40 min | 11 min |
| Within the bar | 3 of 75 | – |
10.00 was no heavier than its past. It was heavier than its past with a queue already standing. At 09.00, 62 jobs met nine porters, almost seven each where this team bends at five. Nearly every job that waited long had gone to a porter still finishing another. From 09.00 jobs arrived faster than they were taken. From 10:30 the two ran level, so the queue held between 30 and 40 instead of falling. At noon, as the jobs coming in grew shorter, the porters accepted 45 in half an hour and it cleared by 12:57.
I replayed the morning with its own jobs and its own porters, each taking a job as quickly as a free porter usually does. The queue still reached 41. Monday was not short of willing hands. It was short of hands.
A replay runs a real day again: its own jobs, arriving when they arrived, and its own porters, on duty when they were. Each porter takes the next job as soon as they are free, and each job takes as long as it really took. On the July and August days it was not tuned on, its daily peaks follow the real ones closely.
It answers what-ifs a report usually cannot. Here, Monday as its porters came, then the same morning with the three who came on at ten starting at nine. A replay is held against a replay, never against the day itself.
Replayed, the porters as they cameThree of them starting at 09.00
The fastest recovery: Saturday, 11:16
On Saturday the queue reached 13 against its own limit of three, and was back under it seven minutes later. A small queue on a quiet day, but it shows the same team clearing a pile-up within minutes when nothing holds acceptance back.
The best hour: Tuesday, 09.00
| Tuesday 09.00 | This hour |
|---|---|
| Jobs | 45 |
| Porters | 9 |
| Jobs per porter | 5.0 |
| Response, median | 4:18 |
| Within the bar | 43 of 45 |
Five jobs a porter is where this team bends: above it, by day, response begins to slow. Tuesday 09.00 sat right on that line, and its median response was 4:18, faster than the week’s 4:58.
I should be honest about one thing. Its bar was easy. The day I compared it with, 21 April, had been a slow morning, so 43 of 45 flatters it. Its own pace is the better measure, and its pace was good.
A Baseline is the most similar hour of the past, not the best one. If that hour was slow, its bar is slow too, and an ordinary hour clears it easily.
Tuesday 09.00 was matched with 21 April 09.00, a slow morning. Its bar was 18:39, against 6:00 for a typical busy hour this week. Its own median, 4:18, is the fairer measure.
| Tuesday 4 August 09.00 | Its Baseline, 21 April 09.00 | |
|---|---|---|
| Jobs | 45 | 39 |
| Jobs per porter | 5.00 | 4.88 |
| Bar for its jobs | 18:39 | – |
| Median response | 4:18 | – |
Outlook: Monday 10 to Sunday 16 August
| State | My forecast | The past eight weeks |
|---|---|---|
| Under Controlled | 20% | 45% |
| Elevated | 67% | 40% |
| Stressed | 10% | 9% |
| Critical | 3% | 5% |
I expect a week much like this one, with Monday morning at its centre again. I rate five hours Critical: Monday 09.00, 10.00 and 11.00, and Wednesday 11.00 and 15.00. 11.00 keeps its habit. The calmest busy hours should be Thursday afternoon.
The forecast is the model Class 26 trained. I ran it this morning on every job before today, so it has already seen 3 August. It predicts how many jobs each hour will bring and how long they will take, and I hold each hour to the bar it will face.
The plain count beside it agrees on the day if not the hours. A weekday and hour that went hard on at least half of its past eight is likely to go hard again, and five did: Monday 09.00, 12.00, 13.00 and 15.00, and Tuesday 08.00.
Recommendations
This week and the weeks before it keep showing pressure gather on weekday mornings between 09.00 and 12.00, and it starts earlier than the hour that shows it. I tried each of these in a replay of this hospital’s own days, with its real jobs and its real porters.
Start Monday’s late-morning porters at 09.00. Three porters came on duty around 10.00 on 3 August, after the queue had already begun. In the replay, the same three starting at 09.00 hold Monday’s queue to 27 instead of 41, with nobody added. My forecast expects next Monday 09.00 to 11.00 to be the hardest hours of the week.
Keep one porter in reserve from 10.00 to 12.00 on weekdays. On the 30 mornings since May when the queue passed 10 in those hours, one more porter would have brought 17 back under 10. A morning like 3 August needs more than one: it took four extra in the replay to hold the queue near 10.
Leave the morning’s requests where they are. Spreading routine moves out of the peak sounds like the obvious fix, but in the replay it saved none of the hardest mornings since May, this Monday among them. Discharges and ward-to-ward moves arrive when they arrive.
Hospital operations rarely slow down, and many important moments of a day become invisible once they pass. This week showed again how much coordination, patience and constant adjustment it takes to keep an operation moving. I hope this report makes some of those efforts a little more visible.