> ## Content Index
> Fetch the complete content index at: https://www.formatforward.com/llms.txt
> Use this file to discover other available public pages before exploring further.

# Premortem Case Study: The Box
- URL: https://www.formatforward.com/premortem-case-study-the-box/
- Published: 2026-09-06T23:49:31.000Z
- Updated: 2026-09-06T23:49:46.000Z
- Description: Five days before The Box aired, a Premortem flagged the exact failure mode that showed up on screen. Here's the report, next to what actually happened.
- Author: Jonathan Glazier
- Tags: Format Lab, premortem Case Study, format forward, ITV, TV Formats

### The Premortem Called It. Here's The Receipt.

A Premortem is a stress test run on a format before it's commissioned, or before it airs. Instead of waiting to find out what an audience thinks after transmission, we travel forward to the point where the format has already failed, work backwards to find out why, and bring the fix back with us. Hence the, **"find the failure before failure finds you log line."** It scores a format against eight dimensions, from Shared Script to Monetisation, runs it past a standing panel of audience avatars, and comes back with a verdict: Develop, Park, or Pass, along with the specific conditions attached to that verdict. You can download the report and read it in full; it's the same one referenced throughout this case study. 

[The Box Demo Format Forward Forensics — Premortem ReportCompanion download for this case study, the full premortem rportThe Box Demo Format Forward Forensics — Premortem Report.pdf940 KBdownload-circle](https://www.formatforward.com/content/files/2026/09/The-Box-Demo-Format-Forward-Forensics-----Premortem-Report.pdf "Download")

On 31 August 2026, five days before The Box aired on ITV, I ran it through my Premortem analysis. Not the full brief, a demo run, built from a one-page format overview and press releases, the kind of quick stress test you'd do on a project you'd come across rather than one you'd been formally commissioned to assess. The verdict was Develop, with a 77/100 format score and one condition flagged in plain language: fix the scoring, or the show will feel rigged.

Nobody fixed the scoring. Here's what the report said next to what actually happened on screen.

## The verdict wasn't a pass. It was a conditional.

The headline result was **Develop**, not Pass, and Gate 1, the Shared Script test, came back **Conditional**, not clear. The scorecard gave Shared Script a 3 out of 5, with a note that read: *"Unknown situations are lived, but the title hides the performance stakes."*

That's the whole story in one line, written before a single episode aired. The lived experience, facing the unknown, unprepared, is real and universal. What the title and format don't do is tell a viewer, from the outset, that the thing they're actually watching is a fair, trackable contest. The gap between those two things is exactly where the show fell down on transmission night.

My take adds to that. "The Chase" is a great title, we all share that experience, from playing kiss chase at school to the adult nightmare of being chased but never caught. The title doesn't explain that it's a quiz, a Q&A where your score is chased down by a professional quizzer, but it tells us everything we need to feel. One episode in, we get the entire deal.

The Box's broadcasters have had to run promos just to explain what the box actually is. That's not a difficult job, we get it fast, celebs in a box get to do big challenges, but the show still failed because the first challenge never delivered on that promise. Contestants searched for tokens and chose where to place them, one to ten. The audience could have been let in on what was being chosen as it happened, so we'd feel the mistake when a celeb picked nine or ten, maybe even shout at the screen, don't pick ten. Instead nobody knew until the end of the game. No jeopardy, in a long game with no time limit.

## The specific failure was named, not just implied

Under Gameplay, the report says: *"A Duel chosen through editorial judgement rather than visible scoring becomes a gameplay failure."* Under Legibility: *"A first-time viewer must know who is safe, who is in danger and why from a freeze-frame."*

What actually aired was an opening game, an egg and spoon race, with no ticking clock and no visible stakes until the very end, when it turned out the game had been deciding safety and elimination the whole time without saying so. The finale, a psychological rock-paper-scissors variant, needed a voiceover to explain why Graeme Souness had lost. Both are the exact failure mode described above: outcomes decided somewhere the audience can't see, surfaced after the fact rather than tracked in real time. The hidden box device was famously used on 8 out of 10 Cats does Count Down at the carrot in a box game, they did it better. At the end we needed to be told if Danny (Aarons) produces X then Graham you are going home. They didn't and the moment was lost.

## The engine problem was flagged before the tasks were even public

The Format Engine section scored 4/5 but carried this specific warning: *"physical, mental, logical and comic tasks can feel incomparable. The engine needs one visible performance currency or a clear category system so each box changes the same game."*

Press reviews of the actual broadcast drew unflattering comparisons to The Traitors and Taskmaster on precisely this point, convoluted rules that made it hard to track where the competition stood. The Competition section of the report had already named this risk by comparing The Box to its nearest neighbours: *"Its neighbours own clearer measures: survive, deliver, deduce or answer. The UK version needs an equally repeatable verb for how players advance."*

## Lineker was scoped correctly, in both directions

The report flagged "talent as decoration" as triggered, and said explicitly: *"Do not make his booking the core commissioning case."* It also credited Lineker as strong launch value, format-appropriate and trustworthy.

That's exactly how the actual reception split. Reviews were warm about Lineker personally, steady, likeable, well-cast for the tone, and separately critical of the format underneath him. The report predicted that split before Lineker had presented a single episode.

## The fix was numbered, and it's the fix that was needed

Of the eight numbered fixes in the report, the first two are:

1. *Create one visible performance currency. Convert every task into the same simple points, danger bands or ranking movement.*
2. *Show the danger table after every box. The sofa must see who is safe, exposed and heading towards The Duel.*

That is, almost verbatim, the fix the show needed and didn't get.

## Where the report didn't reach

I want to be straight about the gap, because a case study that only claims hits isn't useful to anyone deciding whether to trust the process. Watching the actual broadcast, there's a problem the report doesn't name: the box transport itself reads as staged rather than real, the celebrities are visibly manoeuvred into position rather than genuinely sealed and moved, and a savvy modern audience clocks that fast. The Premortem's trust language is entirely about scoring fairness and Duel legitimacy, editorial judgement replacing a visible measure, not about a core physical conceit being caught out as fabricated. That's a different, and arguably more corrosive, kind of trust failure, and it's not one this run tested for.

That's a real finding for the product itself. The Shared Script and Format Engine tests are built to catch legibility and fairness gaps, and they did, precisely. They're not currently built to ask "will a modern audience believe this device is real," and on the evidence of this case, they should be.

If the celebs wee hooded and loaded into the box, show it, don't pretend it isn't happening, Yes i am used to the smoke and mirrors of TV, but so are the audience of today and we must never take that for granted. 

The premortem isn't an AI gimmick. It uses my own database of thousands of formats from all over th world. many have my own notes, but they all have the impact thay made, the gameplay they used and how many series were produced. It uses a sophisticated set of audience behaviours and has every broadcaster platform and streamer with their profiles. 

And possibly most importantly is the reports have my own personal observations and interpretations using 40 years of experience. 

## What this actually proves

Not that a Premortem guarantees a hit. It doesn't, and it isn't supposed to. What it proves is narrower and more useful: a two-hour stress test, run on a single-page pitch document five days before transmission, correctly identified the specific mechanism of failure, named the exact fix, and scoped the presenter risk accurately in both directions. It also missed something a full audience-reality pass, run against footage rather than a paper pitch, would likely have caught.

A Develop verdict was never a green light. It was a list of conditions. What happened here is what happens when a conditional pass gets treated as a pass.

---

*This is what a Premortem run looks like before the fact, not with hindsight built in after transmission. If you're taking a format into a broadcaster and want the same stress test run properly, on a full brief rather than a one-pager,* [*find out more at Format Forward*](https://formatforward.com/?ref=formatforward.com)*. For the full breakdown of what actually went wrong on screen, read* [*"The Box, ITV: Where Did It All Go So Wrong?"*](https://www.formatforward.com/the-box-itv-what-went-wrong/)

*Theres more information on the* [*work with me*](https://www.formatforward.com/tv-format-consultant/) *page.*

*or*

[Book a free zoom](https://scheduler.zoom.us/jonathan-glazier?ref=formatforward.com)

I am dyslexic. AI helps me make this readable. The thoughts and opinions are my own. There are no auto generated articles and I read and research with rigour.

[Jonathan Glazier](https://jonathanglazier.media/?ref=formatforward.com) Format creator and executive producer writer [altmedia](https://altmediajg.substack.com/) on substack