Four Chairs, Ninety Minutes: The Adversarial Review That Saves a Release
Photo by Photo by ThisisEngineering on Unsplash on Unsplash
Every organization has an approval process, and almost none of them are designed to find problems. They are designed to distribute permission. Each reviewer confirms the document is not their fault, the file moves to the next desk, and the launch date arrives with a release nobody has actually attacked.
Reporters attack it. Not maliciously — professionally, and quickly. They read a headline, skim the first hundred words, look for a number, look for a quote, and decide. A review that mirrors that behavior catches weaknesses while they are still cheap to fix.
Why the Standard Pass Misses
Approval passes reward agreement. Red teams reward friction. The difference is structural: in a normal review, the safest contribution is a minor edit and a sign-off. In an adversarial review, the contribution that earns credit is the question nobody wanted asked.
The second difference is time. An open-ended review expands to fill the calendar. A ninety-minute session with a stated agenda produces decisions, because the clock forces prioritization.
The Setup
Book ninety minutes, invite four people who are not the release's primary author, and appoint one facilitator who does not defend the draft. One person keeps a shared score sheet. No side documents, no slide deck, no reopening of strategy.
The only materials in the room: the release, the headline, the quote sheet, the visual assets, the one-page data source list, and the FAQ that will go to spokespeople.
Seat One: The Skeptical Editor — 20 Minutes
This person's job is to say no out loud. Four questions, in order.
What changed this week? If the answer predates the current month, the release is a newsletter item, not news.
Who else could say this? If a competitor could issue the same paragraph on the same day, the news hook is a category statement, not an announcement.
So what, for a reader who does not own the product? The answer belongs in the second sentence of the release, not the fifth paragraph.
What is the strongest version of the counterargument? If no one in the room can state the skeptical view in one sentence, the draft is unarmed.
Seat Two: The Busy Reporter — 15 Minutes
The rule for this seat is brutal and useful: read only the headline, the first hundred words, and the pull quote. Then try to write a lede in one sentence.
If the attempt requires reading further, the structure is wrong. If the number that matters is missing from the opening, the desk will not hunt for it. If the quote restates the headline in different words, it is decoration rather than evidence.
Score this seat honestly. Most drafts fail the first pass; that is the point of running it early.
Seat Three: Legal and Compliance — 20 Minutes
This is not a full legal review and should not be billed as one. It is a targeted scan for the four claim types that generate follow-up questions: superlatives without a source, forward-looking statements without a qualifier, endorsement language that implies a relationship the company does not have, and data whose provenance the company cannot produce on request.
Give this seat one instruction: flag, do not rewrite. Every rewrite proposed in the room adds a day to the cycle.
Seat Four: The Customer Advocate — 15 Minutes
Ask whether a real customer would recognize themselves in the release. Does the quoted customer sound like a person, or like a marketing deck with a name attached? Would the customer's procurement team be comfortable reading their own quote in public? Does the claim match the product they actually bought, at the tier they bought it at?
This seat catches the errors that survive every other review, because they are errors of lived experience rather than of language.
The Score Sheet
Five rows, five scores, one reason per score. Use a one-to-five scale and require the reason to be written down.
- Hook: is the change clear and dated?
- Proof: is there a number a reporter can quote and a source they can verify?
- Quotes: do they sound human and add information?
- Headline: does it survive being read aloud on a phone screen?
- Assets: are images cleared, captioned, and correctly sized?
Anything scoring below three gets a named owner and a deadline. Not a discussion — a name and a time.
The Fix Window
Allow forty-eight hours for fixes, then run a second read that touches only the failed rows. Resist the urge to reopen the whole document, which is how a ninety-minute session becomes a two-week cycle. If a fix requires new data, cut the claim rather than delay the release; a smaller accurate announcement outperforms a larger unverifiable one.
What It Does to the Contract
The red team is also a commercial instrument, and agencies should treat it as one. Insert a standing adversarial review slot in the statement of work, with a defined length and a named facilitator. Specify who holds approval authority and how quickly that person must respond. Add an escalation path for claims that legal will not clear on the scheduled date. Set a rule that review comments arriving after the fix window roll to the next release.
Clients benefit from the clarity. Agencies benefit from a defensible boundary between scope and extra rounds. The review stops being an open-ended courtesy and becomes a scheduled, priced, ninety-minute commitment.
The Measurable Return
Track three numbers across releases: days from draft to distribution, number of rewrites after the release is shared externally, and number of clarifying emails from reporters in the first seventy-two hours. All three should fall. When they do, the case for the next session makes itself.
The broader argument is simpler. A release that has survived four hostile readings arrives at the desk already answering the questions it will be asked. That is what an approval process should produce, and rarely does.