🔬 Case Study 34.2: Northwind — Eleven Hours Short Before Anybody Drew It
📜 Tier 3 — Illustrative. Northwind Semiconductor, the Cascadia Grand, and every figure here are constructed. The arithmetic is internally consistent on constructed inputs. The method transfers; the numbers do not.
Where Chapter 31 left it
400 attendees. Three days at the Cascadia Grand. $540,000, of which $121,400 is AV.
The agenda was rebuilt: general session 14 hours → 4, Meridian training 3 → 12 in sixteen groups of 25, and a 1.5-hour certification session that did not exist when the AV was quoted.
Chapter 31 handled the objective, the money, and the measurement.
Nobody had drawn the build.
The threshold check
| Load-in exceeds one working day | ✅ Two-day build |
| More than three suppliers at one access point | ✅ AV, staging, décor, freight, registration — five, one dock |
| Crew larger than the guest-facing team | ✅ 34 crew at peak against 11 event staff |
| Elements requiring other elements first | ✅ The whole build |
| A vehicle that cannot leave once arrived | ✅ The Cascadia's dock is a dead-end bay |
| Overnight occupation | ✅ Two nights |
| Hiring labour rather than booking suppliers | ✅ House crew, called and paid by the hour |
Seven of seven.
And until week thirty-one, the production document for this event was a delivery schedule with eleven rows.
The eleven hours
The dependency map was drawn in week thirty-one. It took fifty minutes.
Room handover (MON 18:00)
│
▼
Floor protection (1.5h) ──► Staging build (6h) ──► Trim height ──┐
│ │
└──► Power distro (2h, EXCLUSIVE) ──────────────────────────┤
│ ▼
│ Rig / LED (9h)
│ │
├──► Breakout AV ×8 (2 crew, 45 min/room) │
│ [parallel, different space] ▼
│ Focus (2h)
│ │
Décor (4h) ──► Dressing ◄──────────────────────────────────────────┤
▼
Programme (2h)
│
Sound check (1h)
│
Speaker rehearsal (3h)
│
DOORS (THU 06:00)
Alt-text: a dependency diagram for the Northwind build. Room handover leads to floor protection then staging build then trim height. Power distribution runs exclusively and feeds both the rig and eight parallel breakout rooms. The rig leads to focus, programme, sound check, and speaker rehearsal, ending at doors. Décor and dressing join the chain separately and are not on the critical path.
💰 Run the Numbers: the critical path, added up
Hours Floor protection 1.5 Power distro — exclusive, nothing else in the room 2.0 Staging build to trim height 6.0 Rig and LED 9.0 Focus 2.0 Programme 2.0 Sound check 1.0 Speaker rehearsal 3.0 CRITICAL PATH TOTAL 26.5 Plus: two crew meals +1.0 Plus: the truck window before floor protection can start +1.5 TOTAL REQUIRED 29.0 And the window available:
Room handover MON 18:00 Doors THU 06:00 Elapsed 60 hours Less: two overnight closures at 9 hours each — the Cascadia does not permit work 22:00–07:00 −18 Less: Tuesday and Wednesday crew turnaround constraints — no crew can work more than one 10-hour call per day −13 AVAILABLE WORKING HOURS 29 Twenty-nine required against twenty-nine available.
Which is not a plan. It is a coincidence with no margin in it at all.
🚨 And then the certification session.
Chapter 31's rebuilt agenda added a 1.5-hour certification on Wednesday afternoon in sixteen rooms.
The AV quote was written against the old agenda. Sixteen rooms need sixteen sets, not eight — eight more rooms at 45 minutes with 2 crew, plus a test pass.
Which does not sit on the critical path — breakouts are parallel — but it consumes the only crew who could have been doing anything else, and it adds 7 crew-hours on Wednesday, the day with the least slack.
The honest statement of the position, as written in the production meeting minutes:
"We are eleven hours short against a schedule with zero contingency, and we discovered it in week thirty-one because somebody drew a picture."
🎓 Advanced: two currencies, and confusing them is why the gap was invisible
The shortfall is in two different units and they do not add.
Path hours — the 29 required against the 29 available. Zero margin. A production manager will not sign a schedule with no contingency, and a defensible margin on a 26.5-hour path is about four hours.
Crew-hours — the certification session's eight extra breakout rooms are 7 crew-hours, and they land on Wednesday, which is the day with the least slack. They do not lengthen the critical path, because breakouts are parallel. They exhaust the only crew who could have absorbed a problem.
So the honest statement is: four hours of missing path contingency, and seven crew-hours of missing Wednesday capacity.
"Eleven hours short" is the shorthand the production meeting used, and it is two different elevens added together by people in a hurry — which is itself a small instance of this chapter's recurring error. The two numbers are both real and they need two different fixes.
What was done
Four changes, none of which cost the client anything.
⚡ 1. The room handover moved
The Cascadia's previous event was a lunch, out by 15:00, and the room sat empty until an 18:00 handover because that was the contract.
One call. Handover moved to 15:30.
+2.5 hours, for free, and it was available all along because nobody had asked.
⚡ 2. Breakout AV moved to Tuesday
Sixteen rooms × 45 minutes × 2 crew = 24 crew-hours. It had been scheduled for Wednesday because Wednesday was "the breakout day."
The breakout rooms were empty on Tuesday. They are a different space and are not on the critical path.
Moved to Tuesday afternoon, in parallel with the rig. +7 hours of Wednesday returned.
⚡ 3. The truck order was inverted
The original marshalling had the AV truck first because AV is the biggest supplier.
But the dependency map says floor protection precedes staging and staging precedes the rig — so the first vehicle should be the floor protection, then staging, then AV.
AV arriving first would have parked 9 hours of kit in a room that could not receive it.
+1.5 hours, and it avoided a yard problem: the Cascadia's dock is a dead-end bay, so the first truck in is the last truck out.
⚡ 4. The speaker rehearsal moved to Wednesday morning
It had been at 19:00 Wednesday, after sound check, which is the correct order and the wrong day.
Moved to Wednesday 08:00 — before the room is finished, using a partial rig, working set and a confidence monitor.
Which is worse. It is a rehearsal in an unfinished room and the lighting is not right.
And it is three hours off the critical path, and six executives are more likely to attend at 08:00 than at 19:00 on their second day.
💰 The recovered schedule — in both currencies, kept separate
Path hours:
Effect Earlier handover +2.5 available One phone call Truck order corrected +1.5 available Reordering Rehearsal off the critical path −3.0 required A worse rehearsal in a better slot Required 29.0 → 26.0 Available 29.0 → 33.0 PATH CONTINGENCY 0 → 7.0 hours Against a defensible target of ~4 Crew-hours:
Breakout AV moved to Tuesday +7.0 crew-hours of Wednesday capacity returned Wednesday's resource clash Resolved From zero margin to seven hours of path contingency and a resolved crew clash, at a cost of one phone call and three reordering decisions.
Nothing was bought. Nothing was cut. The client was never asked for anything.
The marshalling plan
🕐 Tuesday, the dock
The Cascadia's dock: one bay, 3.6 m, dead-end. First in is last out.
Vehicle Window Contents 06:00 Box van — floor protection 30 min Holds at the Ninth Street marshalling point 06:45 53-ft trailer — staging 90 min Called forward at 06:40 08:30 53-ft trailer — AV, general session 120 min 10:45 Box van — AV, breakouts 45 min 11:45 CONTINGENCY WINDOW — unallocated 60 min The slot that was used 13:00 Box van — décor 60 min 14:15 Box van — registration, signage, freight 45 min 15:30 Box van — furniture 45 min One named person releases every vehicle by radio from the Ninth Street point. Nobody arrives without being called.
And the contingency window was used — the staging trailer arrived at 07:10 rather than 06:45 and ran 25 minutes over. The whole day absorbed it and nothing downstream moved.
Infrastructure
| General session AV | 48 kW |
| Breakouts × 16 | 19 kW |
| Catering — Ch.31's F&B for 400 | 34 kW |
| Working light, offices, registration | 9 kW |
| Sub-total | 110 kW |
| × 1.25 diversity | 138 kW |
House power was adequate at 200 A three-phase. The decision that mattered was not capacity.
🚪 AV went on its own leg, and the reason is a story from a previous event at this hotel.
The Cascadia's kitchen and its ballroom share a distribution board. Two years earlier a different event's LED wall flickered visibly during dinner service, every time the walk-in compressors cycled.
The Cascadia knows this. It is not on any document. It was volunteered by the house electrician during the site visit, in a corridor, because somebody asked him what goes wrong here.
Chapter 33's question — "what do foreign planners get wrong here?" — has a domestic sibling, and it is "what goes wrong in this room?"
Comms: four channels — production, AV, catering and front of house, security. 38 radios, 5 spares, a channel plan printed on the production office wall, and a five-minute briefing at the first call that everybody attends.
The overnight resets
🕐 Wednesday night and Thursday night
22:15 Day ends 22:30 Reset crew calls — 6 people, fresh, called at 22:00, not the day crew 22:30 Ballroom cleared and re-set to the next day's configuration 23:00 Consumables restocked from the stores list — batteries, gaffer, water, coffee, printer paper, gloves 23:20 AV reset to the next day's configuration and tested 23:40 Waste out. Breakout rooms reset — 16 of them 00:20 One person walks the room with the next day's schedule in hand 00:35 Locked. Briefing note left for the 06:00 call The Thursday-night reset found two things: a breakout room whose projector had been set to the wrong input during the day and left that way, and the fact that the certification session on Friday needed table configurations nobody had changed.
Both would have been discovered at 08:15 on Friday, in front of people.
What went wrong anyway
🚨 Three things
1. The stores ran out of gaffer tape on Thursday afternoon. Ordered for three days at a rate estimated from a one-day event. A runner went to a hardware shop. $38 and forty minutes of somebody's day.
2. Two breakout rooms were double-booked by the hotel — a corporate lunch on Wednesday that the Cascadia's own sales team had sold after the contract. Resolved in twenty minutes because the production meeting had produced a room list the hotel had signed.
3. The crew break room was the same room as the speakers' green room on Wednesday morning, for forty minutes, because two documents said different things and neither was wrong.
Nobody was harmed and one executive made a joke about it that was not entirely a joke.
The result
| Build | On schedule. 7 hours of path contingency, 4 used |
| Crew at peak | 34 |
| Meal penalties | $0 |
| Overtime beyond schedule | 6 crew-hours, Tuesday |
| Sound check | Wednesday 18:00, as planned |
| Speaker rehearsal attendance | 6 of 6 — against 3 of 6 the previous year at 19:00 |
| AV budget | $121,400 quoted, $124,900 final — the eight additional breakout sets |
Discussion Questions
DQ1. The event satisfied seven of seven diagnostics and the production document was a delivery schedule until week thirty-one. How?
Consider
**Because nobody's job included asking, and because Chapter 31's work looked like the whole job.** **The objective, the agenda fight, the room block, the F&B, the measurement design — that is a large, difficult, high-value body of work**, and it was done well. **None of it touches the build.** **And the structural reason is the division of labour.** **The AV company had a production manager who knew exactly what the build required.** They were never asked, because they were engaged as a supplier quoting against a spec rather than as a partner scheduling a room. **Which suggests a cheap fix: the first production meeting happens as soon as the largest technical supplier is appointed**, not at −6 weeks. **The AV production manager could have drawn that map in week nine and it would have cost nothing.** **And the honest note the case study should carry: the map took fifty minutes.** **The reason it was not drawn is not difficulty. It is that nobody had it on a list.**DQ2. The handover moved from 18:00 to 15:30 with one phone call. What does that say?
Consider
**That contracted times are frequently defaults rather than constraints, and nobody asks.** **The 18:00 was in the contract because it is the Cascadia's standard.** The previous event was out at 15:00 and the room sat empty for three hours **because the question had never been asked by anybody on either side.** **And notice the return: 2.5 hours is 8.6% of a 29-hour schedule, obtained for free, at the moment when hours were the scarce resource.** **The general form is Chapter 31's *"ask for the three-year pickup"* and Chapter 13's access negotiation:** **the venue holds information and flexibility that it does not volunteer, not out of bad faith, but because nobody has ever asked and its standard terms are written for the median event.** **And there is a limit worth stating: this only works if you ask early.** **A handover change requested at −4 days is a favour; at −7 weeks it is a scheduling question.**DQ3. The speaker rehearsal was moved to a worse slot deliberately. Was that right?
Consider
**Yes, and the reasoning is worth extracting because it is counter-intuitive.** **The rehearsal at 19:00 Wednesday was in the correct sequence** — after sound check, in a finished room, with the real lighting. **It was also three hours sitting on the critical path of a schedule with zero margin.** **Moving it to 08:00 Wednesday makes the rehearsal worse and makes the event safer**, and those are not the same axis. **And then the surprise: attendance went from 3 of 6 to 6 of 6.** **A rehearsal at 19:00 on day two of a conference competes with dinner, drinks, and exhaustion. At 08:00 it competes with breakfast.** **So the "worse" rehearsal was attended by twice as many of the people it exists for**, which probably makes it a better rehearsal in the only sense that matters. **The transferable point: an element's quality in isolation is not its value in a schedule.** **Chapter 31 made the same move with the paddle-raise placement and Chapter 32 made it explicitly** — the best version of a thing, in the wrong slot, is worth less than a compromised version in the right one.DQ4. The house electrician volunteered the kitchen-compressor problem in a corridor. Design the process that captures that reliably.
Consider
**It cannot be fully captured, and the honest answer is that you increase the number of corridors.** **But three things measurably help.** **Ask the question, explicitly, of every venue-side person you meet: *"what goes wrong in this room?"*** — Chapter 33's technique, domesticated. **It is a different question from "are there any issues," which produces "no."** **Ask it of the people who do the work rather than the people who sell the room.** **A director of events knows the contract; a house electrician knows the distribution board.** They are rarely in the same meeting and the second one is never invited. **And write down the answer somewhere that survives.** **The Cascadia has known this for at least two years and it is on no document** — which means every planner rediscovers it or does not. **The deeper observation, and it is uncomfortable: a venue's institutional knowledge frequently lives entirely in individuals**, and a planner benefits from it only by accident. **Chapter 39's systems work should include a venue file that records exactly this kind of thing, per venue, forever.**DQ5. The gaffer tape ran out because a three-day quantity was estimated from a one-day rate. Generalise.
Consider
**Consumption is not linear in days and everybody estimates it as though it were.** **Day one uses a day's worth from a full stores.** **Day two uses more** — more repairs, more improvisation, more things that have started to fail — **and it draws from a stores nobody restocked.** **By day three the rate is higher and the stock is lower.** **§34.9 lists consumables as failure mode one for exactly this reason**, and the fix is the overnight reset's restock-from-a-stores-list line. **But the deeper error is the estimate itself.** **The right method is not "one day × three." It is a rate per crew-hour**, and this build had 34 crew at peak across three days — **substantially more crew-hours than a one-day event of the same guest count.** **And the general form recurs: a quantity derived from the wrong unit.** **Chapter 28's barn seated 89 for a dinner and was asked to hold a ceremony. Chapter 32's wine pull was scoped from a merchant's willingness rather than a guest count.** **Here, tape was scoped per day rather than per crew-hour.** **Three chapters, three unit errors, and the fix each time is the same: write the unit next to the number.**DQ6. The crew break room and the speakers' green room collided because two documents said different things and neither was wrong. What kind of failure is this?