How Critical Checks Work | What Belongs on a Checklist

How critical checks work begins with a refusal to put everything on the checklist.

A checklist is strongest when it protects a small number of conditions that are important enough to verify, easy enough to miss and actionable enough that a failed check can still change what happens next.

This is the first pillar of reliable checklist design: selection.

It belongs to the larger How Checklists Fail series. The other two pillars explain timing and verification.

The Selection Problem

Every activity contains more facts, actions and preferences than a checklist can carry.

A mathematics solution may involve reading, notation, arithmetic, algebra, diagram interpretation, units, estimation and presentation.

An essay may involve argument, evidence, paragraphing, citations, grammar, formatting and submission requirements.

A laboratory procedure may involve identity, calibration, concentration, sequence, protective equipment, temperature, timing and disposal.

If all of these are turned into checklist items at every stage, the list stops acting as a filter.

It becomes another manual.

The central design question is therefore not, “What can we remind the user about?”

It is, “Which omissions or wrong states deserve a deliberate stop?”

A Critical Check Protects a Consequence

The easiest way to identify a critical check is to begin from consequence.

Ask what happens if the condition is wrong and the process continues.

Some mistakes are cheap and reversible.

Others propagate.

A missing comma can often be repaired late.

Submitting the wrong file can invalidate an otherwise excellent assignment.

A small sign error at the beginning of a long mathematics solution can contaminate every line that follows.

A wrong patient, wrong site or wrong procedure in surgery has consequences so serious that explicit confirmation is justified even when the information appears obvious.

The check earns its place because the consequence earns attention.

This does not mean only life-and-death conditions belong on checklists.

It means every item should have a reason stronger than “this seems like a good thing to remember.”

The Four Filters for Checklist Inclusion

A practical way to choose checklist items is to pass each candidate through four filters.

1. Consequence

What is the cost if this is wrong or omitted?

The cost may be safety, marks, time, money, rework, credibility, lost data, missed opportunity or an irreversible transition.

2. Forgettability

Is the item vulnerable to omission even by a competent person?

Routine, interruption, time pressure, divided attention and familiarity can all make obvious steps disappear from awareness.

3. Recoverability

Will the check happen while repair is still possible?

If the answer arrives after the point of no return, the item may belong in an audit rather than an operational checklist.

4. Checkability

Can the user obtain meaningful evidence?

“Be careful” is not checkable.

“Confirm that the final answer is in metres, not centimetres” is.

The strongest candidate items pass all four filters.

What Usually Does Not Belong on the Checklist

Removing weak items is as important as adding strong ones.

Several kinds of content often belong elsewhere.

General knowledge

If the user must understand a concept, teach the concept.

Do not compress an explanation into a vague checklist line and pretend the knowledge problem has been solved.

Long procedures

If the user needs a sequence of many actions, use a procedure or worked method.

The checklist may sit beside it and protect the most failure-sensitive points.

Preferences with no meaningful consequence

A team may prefer a particular folder name, font or layout.

These can be standardised if useful, but they should not crowd out genuinely critical checks.

Items that cannot change action

A check is weak if a failed result produces no decision.

Operational checks should normally lead to continue, correct, escalate, stop, repeat, replace or seek help.

Items whose only purpose is to make the list look complete

A checklist is not improved by symmetry.

It is improved by risk coverage.

The Memory Test: Would a Competent Person Still Miss It?

A checklist is particularly useful for failures of prospective memory: remembering to do the right thing at the right time.

Competence does not eliminate this problem.

In fact, experienced people can become vulnerable in a different way.

Familiarity makes routine actions feel automatic.

That reduces the subjective need to stop and confirm.

The better selection question is not “Do experts know this?”

It is “Can an expert under realistic conditions still omit this?”

A known step can still deserve a check if omission is plausible and consequential.

This is one reason aviation and surgery use checklists despite relying on highly trained professionals.

The Propagation Test: Can One Small Error Infect Everything Downstream?

Some errors are local.

Others propagate.

Propagation makes a check valuable.

Imagine a student copies x + 3 = 11 as x − 3 = 11.

Every later algebraic step can be internally consistent and still produce the wrong answer.

A quick transcription check near the beginning can prevent minutes of flawless work on the wrong representation.

In project work, using the wrong source file can produce hours of correct editing on an obsolete version.

In data work, selecting the wrong date range can make every later chart technically correct but substantively wrong.

A critical check often sits upstream because upstream errors multiply.

The Interface Test: Does Responsibility Cross a Boundary?

Checks become especially valuable at interfaces.

One person hands work to another.

One system exports data to another.

A student moves from reading a question to forming an equation.

A writer moves from research to drafting.

A team moves from planning to execution.

Boundaries create translation risk.

The sender and receiver may hold different assumptions.

An interface check asks whether the essential state survived the transition.

Examples include:

Is this the latest file version?
Does the receiving person know what remains unresolved?
Did the numerical value retain the correct unit?
Did the written problem become the correct equation?
Did the cited source actually support the claim now attached to it?

These checks protect meaning as it moves.

The Irreversibility Test: Is This the Last Cheap Chance to Catch the Error?

A critical check often appears just before a commitment.

Before submission.

Before payment.

Before publication.

Before deletion.

Before starting a dangerous procedure.

Before leaving an examination question.

The closer a process gets to an irreversible state, the more valuable a targeted confirmation can become.

This does not mean checking everything at the end.

Some errors become too expensive to repair if discovered late.

The design task is to identify the last cheap repair point for each important failure.

That question leads directly into How Checklist Timing Works.

The Distinct-Failure Test

Two checklist items should not exist merely because they use different words.

Each item should open a different failure channel.

Consider these two lines:

Check answer.
Make sure answer is correct.

They are duplicates.

Now consider:

Requirement: Did I answer the quantity the question asked for?
Calculation: Can I verify the numerical result independently?
Representation: Are units, signs, symbols and rounding correct?

These are different.

A student can pass one and fail another.

That makes the checklist diagnostically useful.

Distinct items reduce redundancy and improve attention.

The Action Test: What Happens When the Check Fails?

A checklist is not only a detection device.

It is a routing device.

A failed check should send the user somewhere.

If identity is uncertain, stop and resolve identity.

If a calculation fails a reverse check, return to the first divergence.

If a source does not support a claim, replace the source or rewrite the claim.

If the latest file version is uncertain, reconcile versions before continuing.

If equipment is not ready, do not proceed as though the check were a ceremonial step.

The action test is simple:

For every item, finish the sentence “If no, then…”

If there is no meaningful completion, reconsider whether the item belongs.

Critical Does Not Mean Dramatic

Criticality is contextual.

In an exam, a unit check may protect one mark.

In engineering, a unit error can produce a catastrophic mismatch.

The item itself is not universally critical.

The consequence inside the system determines its weight.

This prevents checklist design from becoming a collection of generic best practices.

A good checklist belongs to a particular task, user, environment and consequence structure.

WHO’s implementation guidance explicitly encourages local adaptation of the Surgical Safety Checklist while protecting its essential safety purposes.

The larger lesson is that checklist design should preserve the function of the check, not blindly copy the surface form.

Three Layers: Critical, Conditional and Reference

One way to keep a checklist compact is to separate three kinds of information.

Critical checks

Always-visible conditions that deserve deliberate confirmation.

Conditional checks

Items that appear only when a particular state exists.

For example, “If the question requires an exact value, do not convert to a decimal.”

Reference information

Details that are useful but do not require active confirmation every time.

Examples include formulas, definitions, formatting examples or troubleshooting notes.

Mixing all three layers into one flat list creates clutter.

Separating them protects attention.

Checklist Bloat: How a Useful Tool Becomes Noise

Checklist bloat usually happens gradually.

A mistake occurs.

Someone adds a new item.

Another concern appears.

Another item is added.

Nobody removes anything.

Eventually the list becomes a historical archive of every anxiety the organisation has ever experienced.

This is understandable but weak.

Each new item imposes a recurring attention cost.

The right response to a failure is not automatically “add a checklist item.”

First ask whether the failure came from:

missing knowledge;
poor task design;
unclear responsibility;
bad interface design;
insufficient training;
unavailable information;
unreliable equipment;
weak incentives;
timing;
or genuine omission.

Only some of these are best repaired by a checklist.

A Removal Rule

Every checklist should have a way to remove items.

Removal does not mean lowering standards.

It means protecting signal.

An item may be removed or relocated when:

the underlying process has been redesigned so the failure can no longer occur;
the item is reliably enforced by the system itself;
the item duplicates another stronger check;
the item has become reference information rather than an active decision;
the item cannot be acted on at the checklist point;
the item is routinely skipped because it does not fit the real workflow; or
evidence shows it does not address the failure it was intended to prevent.

A checklist should be curated.

It should not merely accumulate.

Example: Building an Exam Checklist from Failure Data

Suppose a mathematics student repeatedly loses marks.

Do not begin with a generic internet checklist.

Begin with the student’s errors.

Across six papers, the tutor notices:

three answers use the wrong unit;
two equations were copied with a changed sign;
four questions were left partly unanswered because the student missed a second instruction;
one numerical result was wildly unreasonable but never challenged;
and several questions were correct with no need for additional checking.

The resulting checklist should reflect the real failure pattern.

A compact version could be:

Requirement: Have I answered every part of the question?
Transfer: Did every sign, exponent and bracket survive correctly?
Unit: Is the final quantity in the requested unit?
Reasonableness: Is the size and sign of the answer plausible?

That is a better checklist than twenty generic reminders.

It is built from evidence.

Example: Building a Writing Checklist from the Rubric

For an essay, the checklist can be built backward from what must be true at submission.

Suppose the task rewards claim quality, evidence, explanation and organisation.

The checklist should not repeat the whole rubric word for word.

Instead, identify the failure points that are easy to overlook during final review.

Central claim: Can I state the main answer in one sentence?
Evidence link: Does each major claim have evidence that actually supports it?
Reasoning link: Have I explained why the evidence matters?
Paragraph role: Does each paragraph do a distinct job?
Submission state: Is the final file complete, correctly named and actually uploaded?

The rubric defines quality.

The checklist protects execution.

They are related tools, not identical tools.

Example: Building a Laboratory Checklist from Hazards

A laboratory checklist should be built from hazard and task analysis, not aesthetics.

The designer asks where error can cause harm, invalidate the experiment or contaminate later stages.

Candidate critical checks might include identity, concentration, protective equipment, calibration, sequence and disposal.

But the exact selection depends on the procedure.

The United Kingdom Health and Safety Executive’s human-factors guidance makes the broader point that procedures and job aids should be informed by risk assessment, task analysis, user involvement and the consequences of failure.

The checklist is part of risk control.

It is not the entire risk-control system.