ATTENTION, MANAGERS WHO HAVE PURCHASED A LAMINATOR IN THE NAME OF SAFETY!
PEOPLE WHO BELIEVE A CLIPBOARD BECOMES WISE ONCE IT ACQUIRES A PLASTIC SKIN! PEOPLE WHO HAVE SEEN A COMPLETION RATE AND ASSUMED A DECISION HAPPENED SOMEWHERE NEARBY!
Are you tired of not having a laminated alibi for the next serious failure?
Have you watched a team sign every box, nod gravely at the cover sheet, and then discover that the one question which actually mattered had quietly gone on holiday?
Then congratulations.
You qualify for the Five Number One Rules of a Checklist That Survives Tuesday — five rules, five Number Ones, and one very limited warranty.
“Wait,” you’re shouting at the screen, “they can’t all be Number One!”
That’s exactly what Big Numbering wants, while it sells “make it short” as a substitute for “make it happen.”
Checklist failures are specialists.
A list with all the right items can arrive after the decision.
A perfectly timed prompt can be run in the wrong mode.
A brilliant list can be locally edited into something nobody has tested.
A reported completion rate can conceal a completely empty ritual.
And a mandate can put a document everywhere while putting its questions nowhere.
None of them covers for another. A short list can still be mistimed. A well-timed list can still hold the wrong items. And a high completion rate can’t prove that any single item ever changed anybody’s behaviour.
So here they are.
Five rules.
Five Number Ones.
No substitutions. No “completed” stickers. No laminated force field around a process nobody has actually changed.
RULE #1: START WITH THE FAILURE YOU CANNOT AFFORD
Introducing EVERYTHING-EVERYONE-COULD-EVER-DO™, the checklist so comprehensive it requires a snack break, a second shift, and the eventual collapse of the original operation.
Don’t draft from a wish list of respectable behaviours.
Start from specific things that go wrong and matter when they do.
A checklist isn’t a training manual in tiny boxes. It’s a prompt for the failures competent people still miss under pressure, interruption, routine, or divided responsibility.
The WHO Surgical Safety Checklist has 19 items, and the number isn’t a style choice. It was built from documented recurring categories of harm: wrong-site surgery, unrecognised blood loss, retained instruments, airway failure, surgical-site infection. It is emphatically not a restatement of surgery.
That’s the whole point. A list trying to encode everything a capable operator knows becomes too long for the pauses real work actually allows.
Call an entry a killer item when its absence leaves a serious failure mode unchallenged. “Confirm the thing that prevents the bad thing” beats “be mindful of quality” — which is what a checklist says when it wants credit for specificity without the risk of meaning anything.
Then trial the wording before defending it. Early WHO reviewers found that apparently plain instructions — confirm patient identity, confirm the marked site — still raised practical questions. When does the team ask? Who answers? What counts as an adequate answer?
An item isn’t clear because a committee read it silently. It’s clear when people in the workflow run it the same way.
Put a prompt on the list only when its absence permits a failure the work cannot afford.
If every good habit is a checklist item, then no bad omission has to be.
RULE #1: CHOOSE READ-DO OR DO-CONFIRM ON PURPOSE
NOW AVAILABLE: CHECKLIST BLENDER PRO™.
Pour in a sequence. Press “professional.” Enjoy a read-do list performed entirely from memory, right up until the omission it was designed to prevent.
Read-do means the operator reads an item, does it, reads the next. The list drives the work. It’s slower on purpose, because it can’t be outrun by memory — and after an interruption, the list itself records where things stopped.
Do-confirm means the operator works from memory and experience, then stops and confirms each item afterwards. Faster in familiar work. And the tradeoff is precise: it catches an omission after the action, rather than preventing the step from being skipped at all.
Two questions pick between them.
How practised is the operator?
How expensive is correction while the work is still underway?
A walkaround on an unfamiliar aircraft type runs read-do — the flow isn’t automatic and mid-sequence correction is cheap. A surgical pre-incision time-out runs do-confirm — the team has already done the underlying work and now needs a verbal, cross-checked confirmation.
The WHO list actually mixes modes by phase. Sign-in, before induction, sits closer to read-do because it gates whether anaesthesia proceeds at all. The time-out before incision is do-confirm; it isn’t there to dictate the order of prior work, only to make the team say out loud what’s already true.
This isn’t an academic distinction in a hard hat. Degani and Wiener’s 1993 cockpit field study recorded crews performing nominally read-do items from flow-memory, then reading the list afterwards to confirm them.
That’s do-confirm wearing a read-do nametag. And the omission risk walks straight back in.
A checklist mode is a promise about when the list catches the mistake.
Don’t buy a brake and then install it in the rear-view mirror.
RULE #1: KEEP ONLY KILLER ITEMS
From the makers of “JUST ONE MORE BOX” comes THE COMPLIANCE PARADE DELUXE — so many questions that the important one has to wave from a float.
This isn’t a length preference dressed up as a principle. It protects the relationship between a prompt and a failure.
An item requiring a fast binary confirmation behaves completely differently from one requiring discussion and judgment. Pretend otherwise and you get a list whose total completion looks sturdy while its hardest questions quietly rot.
Direct observation across ten Colorado hospitals watched 854 procedures, without announcing the observers in advance. Patient and surgical-site verification ran at 91–99% item-by-item compliance.
Discussion of expected case duration: 16%.
Discussion of anticipated blood loss: 19%.
Physicians actively participated in only 71% of runs at all.
Those are not interchangeable boxes in different fonts. The fast binary items stayed near the front of everyone’s attention. The discussion-heavy items — the ones requiring somebody to think in front of colleagues — were exactly what the system was least reliably getting.
A report that says “checklist compliance” without an item breakdown cannot tell steady partial use apart from a ritual that always catches the easy questions and abandons the hard ones.
So revise against what actually breaks: items skipped, misunderstood, or answered wrong at a meaningfully higher rate than the rest. Not against the feeling that an item reads awkwardly in a conference room. A difficult question might need a clearer owner, better wording, a different pause point — or removal, if it simply isn’t actionable in that setting.
No fixed length is right. The test is whether each surviving item earns its interruption.
A short checklist is not a small checklist. It is a list where every item has survived a fight for its place.
An aggregate score is where a dying item goes to look alive.
RULE #1: PUT THE CHECK AT THE POINT OF ACTION
BEHOLD REMEMBER-LATER 6000™ — the revolutionary add-on that asks busy people to leave their real work, visit a separate compliance shrine, and volunteer the truth at an unspecified future hour.
A good prompt isn’t merely nearby. It rides on a pause the work already has to take.
The WHO checklist fixes three: sign-in before anaesthesia, time-out before incision, sign-out before the patient leaves. Those work because the whole team is already stopped there — not because a checklist needed somewhere to live.
That’s the broader test, and it shows up well outside medicine.
Amazon didn’t bolt a ceremonial values meeting onto the end of its decisions. It changed the decision meetings themselves. Slide decks banned company-wide. Narrative memos of four to six pages, in complete sentences. Meetings that open with silent reading before anyone talks. The two-pizza rule keeps the room small enough that a shared memo becomes an argument rather than a presentation performed at thirty witnesses.
Netflix made the same move more expensive. Its 2009 culture memo states plainly that adequate performance doesn’t retain a role, and ties that standard to a generous severance package rather than a formal improvement plan. The consequence is funded, in advance, to actually occur.
That’s a rule with teeth, not a sentence in a lobby.
Borrow the architecture, not the specifics — a checklist shouldn’t turn every decision into Amazon or every missed item into an HR proceeding. Attach the prompt to a required event. Name who answers. Decide what happens when the answer fails, before the moment arrives.
A principle that needs optional future enthusiasm is a principle looking for a ride home.
A prompt changes work when it occupies a moment the work cannot skip.
The ritual is the delivery system. The sentence is just the label.
RULE #1: AUDIT THE ITEM, NOT THE COMPLETION THEATRE
AND NOW, BOX-CHECK-O-VISION™ — the device that converts “somebody marked it complete” into “every required action occurred,” while a small fan blows confetti across the evidence gap.
Observed compliance beats self-report, and costs far more to collect. That cost is exactly why a completed form feels so tempting.
It’s also why the form can’t be allowed to grade itself.
In 212 endovascular procedures in one hybrid operating room, teams reported completing 95.5% of pre-incision time-out items.
Independent video review observed 46.8%.
The gap was roughly 49 percentage points.
The form measured documentation. The video measured behaviour. Those had been the same number right up until somebody checked.
So audit with video, an independent observer, or an objective record — a scanned barcode, a timestamped log — wherever the item allows one. And keep the audit at item level, because a single aggregate percentage can’t distinguish “everything is moderately unreliable” from “three decision-heavy items have vanished while the easy ones carry the score.”
Even then, compliance still isn’t effectiveness.
The 2009 WHO trial covered 7,688 operations at eight hospitals across eight cities of varying resource levels. Major complications fell from 11.0% to 7.0%, in-hospital death from 1.5% to 0.8%. That design answers whether the programme changed outcomes — a different question from whether a form was present.
And a mandate doesn’t make that result automatic. Ontario compared 109,341 procedures before a province-wide checklist mandate against 106,370 after, across 101 hospitals. Mortality moved from 0.71% to 0.65%. Complications, 3.86% to 3.82%. Neither change was statistically significant.
Adoption is not execution. Execution is not outcome. Each one needs its own measurement, and the checklist will happily accept credit for all three.
A completed checklist is a claim. An item-level observation is evidence. An outcome study is a different question again.
The list doesn’t become real when the ink dries. It becomes real when a hard item changes a hard decision.
BUT WAIT, THERE’S MORE!
“What if it’s a software deployment?”
Same rules. Routine deployments can run do-confirm when experienced people verify the work happened. Infrequent high-risk changes — a database migration, a security-sensitive config change — run read-do, because finding the omission afterwards means a live outage.
“What if it’s a crane lift?”
Same rules. Pre-lift runs read-do, with operator and rigger, before the load moves. A lift in progress can’t safely pause mid-step.
“What if a local team needs to change the wording?”
Same rules. Adaptation is allowed and often necessary — but the adapted version is a new artifact. Test it where it will actually run, including at sites that don’t resemble the one it was written in.
“What if leadership already published the principle?”
Same rules. Enron published a 64-page Code of Ethics in July 2000. It had no traceable effect on any of the accounting decisions that followed. The document isn’t the mechanism. The changed decision is the mechanism’s receipt.
The details change.
The architecture doesn’t.
THE FIVE, WITHOUT THE LAMINATE
Start from a serious, specific failure and keep only the items that prevent or catch it.
Choose read-do when the list must drive the action, do-confirm when familiar work needs shared verification afterwards — and choose deliberately, not by drift.
Keep only killer items. Test the wording, and audit each item for decay rather than protecting a long list out of habit.
Put the prompt at a pause the work already takes, with a named owner and a real consequence.
Measure observed item-level compliance separately from documentation, and outcomes separately from compliance.
ACT NOW, BEFORE THE LAMINATOR COOLS
Take one existing checklist tonight. Beside every item, write the serious failure it prevents or catches.
If an item has no answer, it isn’t a checklist item. Put it back into training, or a procedure, or the general category of things people hope somebody remembers.
For what survives, write the mode: read-do or do-confirm. Name the pause point where work already stops. Name the role that answers.
Then write what happens when the answer is no.
That last line is the price of admitting this is a decision aid rather than office decor.
Run the draft with people doing the real work, and watch specifically for wording that leaves timing, ownership, or an adequate answer vague. If the list will travel, test it somewhere that doesn’t resemble the room where it was invented.
Then refuse the easy number. Don’t ask only whether the form was completed. Observe some runs. Pull an objective record where one exists. Report item by item.
And if outcomes are what matter, compare before and after across enough sites and volume that you aren’t crediting the list for something else entirely.
For the low, low price of making one real ritual answer one real question, the complete Checklist That Survives Tuesday system is yours.
No bonus clipboard. No ceremonial signature. No warranty against every failure.
Just a prompt that meets the work at the moment the work has to answer.
Operators are no longer standing by.
The operator is already at the point of action.