Kyalami Grand Prix Circuit, 28 to 30 August 2026. Prepared by One-eyed Jack.
The three-day report
On Saturday and again on Sunday, 135 afternoon seats went live at 12h30 for that same afternoon. Both days every one of them was gone before 12h31 — 56 seconds on Saturday, 51 on Sunday, timed from the moment they appeared. Across the three days the Audi Driving Experience offered 648 seats, booked 640, and delivered 667 drives to 628 people. Every figure here comes from the booking record.
The afternoon board did not empty. It vanished.
Saturday and Sunday each put 135 afternoon seats on the board at 12h30, bookable for that afternoon only. Saturday’s last of the 135 was taken at 12:30:56. Sunday’s at 12:30:51. Timed from the release, that is 56 seconds and 51 seconds — 270 seats across the weekend, none of which survived the minute it was released in.
Underneath those two windows sit six individual sessions. Measured the same way — from the moment a session was released to the moment its last seat went — the quickest filled in 21 seconds and the slowest of the six in fifty-six. The morning sessions, released at 07h45, are on the same scale for comparison.
The 07h45 morning release behaved differently, and usefully so. It took 266 of its 270 weekend seats, 98.5%, but the four sessions that filled took between three minutes and thirty-eight. Sunday’s morning window took 45 bookings in its first minute, then 31, then 20, then 19 — the shape of a queue draining rather than a door bursting.
Two release windows, both landing while the audience was already on the ground at Kyalami. That is the mechanic, and it is the part of this activation that carries directly into the next one.
The board carried 648 seats. The scanners at the cars counted 667 drives.
Every day finished ahead of the seats it had to give: 112 against 108 on Friday, 278 against 270 on Saturday, 277 against 270 on Sunday. Across the weekend that is 102.9% of the offered capacity actually used.
Those 667 drives were taken by 628 people. Thirty-eight of them drove twice or three times, and eighteen came back on a different day — they queued again, on a board that was by then almost entirely full, to do it a second time.
Seventy-nine of the 667 went to someone who did not arrive holding a booked seat: 56 taken at the door, 20 who were still in a queue when they were put in a car, and three who had cancelled and came anyway.
One caveat, and we would rather raise it than have it raised. Attendance is a scan attributed to a session, so a guest arriving late and rolling into the next slot reads as an over-capacity session rather than as extra throughput. The day-level figure is the defensible one: each day put more drives through the cars than it had seats for. Saturday 18h00’s 54 is a single reading and should not be briefed as a repeatable per-session rate.
The practical consequence is simple. Sizing 2027 against seats booked counts 640. The activation delivered 667 drives. Plan against drives.
Fifty-two booked guests never scanned in. Almost none of it was lost.
That is 8.1% of the seats taken, and it is normally a car standing still. Instead, 56 people arrived at the stand without a booking and were put into cars — every one of them scanned in — and further guests were called off the session queues as places came free. Walk-ins ran 4 on Friday, 25 on Saturday and 27 on Sunday.
Sunday shows it most clearly. Thirty-two booked guests did not arrive — a 12.0% no-show rate against Friday’s 3.8% — and the day still ran 277 drives through 270 seats.
Taken together, 56 walk-ins and 20 guests admitted straight off a queue more than covered the 52 seats the no-shows left behind. None of it was planned. Four walk-ins on Friday against twenty-seven on Sunday is a process finding itself over three days rather than starting the job on day one.
The record shows when each walk-in was created, not which empty seat it filled. Read it as capacity added at the door rather than a measured seat-for-seat recovery.
Fifteen of the eighteen queues hit a ceiling we set ourselves, not one the audience set.
Each session’s queue was capped at 15 places. Fifteen sessions filled that cap, including all six on Sunday, and fourteen were still full when the show closed. Once a queue was full, the next person was refused and no record was kept of them.
That puts 234 of the 247 recorded entries inside queues that had stopped counting. Only three sessions produced a complete reading — Friday 08h30, Friday 10h45 and Saturday 08h30 — and two of those were Friday’s, running 18-seat sessions rather than 45.
So 247 is the floor of unmet demand. It is not a measure of it. Of those 247 queue entries, 11 were called into a seat and 236 were still waiting when the show closed. Any capacity conversation for 2027 that treats 236 as the size of the audience left outside will size the activation too small.
The fix costs nothing. The cap is a single configuration value. Lift it, keep a short callable queue for genuine refills, and word the screen so a guest knows they are joining a register rather than a promise. Next year’s number would be an honest count. It would still be a floor — a guest who sees a long queue and walks away is never recorded — but a far tighter one.
Friday’s board stopped moving at 09h46, with eight hours of the day still to run.
Friday ran 18 seats a session against 45 on the other two days, and it ran without a timed release. The first booking landed at 07:49:42 and the last at 09:45:53. In one hour and fifty-six minutes, with no countdown to concentrate it, the day had booked 104 of its 108 seats, 96.3%, and nothing was booked after that. From 09h46 on, every Audi prospect who walked up could be offered a queue, not a drive.
The cleanest reading Friday produced is 10h45: eighteen seats booked, eleven more in the queue, and that queue never filled — so 29 people for 18 seats is a complete count rather than a truncated one. Friday’s other four queues all hit the cap, so those are floors. Friday also returned the best show rate of the three days, 96.2%, on four no-shows.
Our read: Friday should grow, and that direction is settled. The size is not. 104 bookings plus 74 queue entries is 178 recorded requests, which would fill 66% of a 270-seat Friday — but the cap censors that figure upward, and queue entries count requests rather than distinct people, which pulls it back down. Neither can be resolved from this year’s data.
We would plan around roughly 30 seats a session and shape the day rather than flatten it. Friday 08h30 finished a seat short with nobody in its queue; Friday 18h00 recorded 25 drives against 18 seats and a full queue behind them. Growth belongs at the back of the day. We would not present a full match to Saturday and Sunday as ruled out, because this year cannot rule it out either way.
833 people passed through the booking record. 541 of them opted in.
That is 64.9%, and it is the contactable list Audi keeps from Kyalami.
It is not the audience figure, and the two should never be swapped. The audience is the 667 who drove. The 541 is drawn from the whole record, so it includes people who joined a queue or cancelled and never got into a car.
Every record in the system — all 640 bookings, all 56 door guests, every name on every queue — carries a full name, surname, mobile number and a signed indemnity. Every one of the 667 drives was signed for before the car moved, and there was nothing left to collect after the show closed.
That record is what makes the list worth having. Consent sits alongside a verified attendance, at a named activation, on a specific date. It is what turns 541 names into an asset rather than a file. We recommend one clean handover: the 541 consented individuals, stamped Audi Driving Experience, Kyalami, 28–30 August 2026. The other 292 gave no marketing consent, so none of their details travel — excluded from the file, not sent across flagged.
The door more than covered 52 no-shows without an extra car, an extra minute of track time or a rand of media. It is also the only thing that can close the eight empty seats: seven of the eight sat in sessions where a booking was cancelled after the session had already begun, which no booking window can re-fill. Put a named standby queue on the run sheet, give one person the job from the first session on Friday, and agree in advance how many extra guests a session may carry.
Fifteen of eighteen queues filled to a limit the system imposed. Lifting it is the change that makes every other capacity decision measurable instead of argued — and it is the only way to learn what Kyalami actually wanted.
12h30 emptied 135 seats inside a minute on both days, and 07h45 took 266 of its 270. The staged release is doing real work: it puts inventory in front of an audience that is already on site, twice a day, instead of once before the gates open. There is nothing in this data asking for it to change.
Sunday’s no-show rate was 12.0% against Friday’s 3.8%. Nothing here supports changing the booking mechanism, which delivered 98.8% of seats. It supports keeping the door team and the standby queue at full strength on the day the schedule usually winds them down.
| Session | Released | Booked | Drove | Door | Queued | Drives |
|---|---|---|---|---|---|---|
| Friday 28 August | ||||||
| 08:30 | — | 17/18 | 17 | — | — | 17 |
| 10:45 | — | Full | 18 | — | 11 | 18 |
| 12:30 | — | 17/18 | 17 | — | 15 filled | 17 |
| 14:45 | — | Full | 16 | +1 | 16 filled | 18 |
| 16:10 | — | Full | 16 | — | 16 filled | 17 |
| 18:00 | — | 16/18 | 16 | +3 | 16 filled | 25 |
| Saturday 29 August | ||||||
| 08:30 | 07:45 | Full | 45 | — | 2 | 45 |
| 10:45 | 07:45 | 44/45 | 39 | — | 15 filled | 39 |
| 12:30 | 07:45 | Full | 41 | +4 | 16 filled | 45 |
| 14:45 | 12:30 | Full | 44 | +4 | 15 filled | 48 |
| 16:10 | 12:30 | Full | 44 | +3 | 15 filled | 47 |
| 18:00 | 12:30 | Full | 40 | +14 | 17 filled | 54 |
| Sunday 30 August | ||||||
| 08:30 | 07:45 | 42/45 | 36 | +3 | 15 filled | 40 |
| 10:45 | 07:45 | Full | 39 | +7 | 17 filled | 48 |
| 12:30 | 07:45 | Full | 39 | +7 | 15 filled | 48 |
| 14:45 | 12:30 | Full | 41 | +3 | 15 filled | 46 |
| 16:10 | 12:30 | Full | 42 | — | 16 filled | 45 |
| 18:00 | 12:30 | Full | 38 | +7 | 15 filled | 50 |
Six hundred and forty-eight seats, every one of them bookable only for the day it ran. Six hundred and forty taken, and 667 drives delivered to 628 people. Fifteen queues that filled to a ceiling the system put there rather than one the audience did.
Audi did not simply book out at Kyalami. It ran over its own capacity, thirty-eight people came back to drive a second time, and it was still turning people away — at the queue and at the door — on the last afternoon of the show.
That is the case for 2027, and it is made of timestamps rather than adjectives.