Can Someone Follow Your Instructions Without Guessing?

The procedure has been approved. The terminology is consistent, the screenshots are current, and every sentence has survived copyediting. Then a real user reaches step four, turns the page over, and asks, “Which button do I press now?” That pause is not a reading problem to blame on the user. It is evidence about the instructions.

In August 2026, the U.S. Food and Drug Administration updated its human-factors and usability-engineering guidance for medical devices. The guidance is written for a regulated, high-consequence setting, but one of its most useful ideas travels well: instructions, labels, packaging, training, controls, and feedback all meet the user as one system. A writer cannot assume that polished words will repair a confusing task.

That does not make every help page a medical-device project, and it does not turn an informal content check into regulatory validation. It does give editors, medical writers, course designers, and content leads a disciplined question to borrow: can the intended person perceive the information, interpret it correctly, decide what to do, and complete the action under realistic conditions?

Clarity is a performance question

Editors are trained to spot ambiguity, unsupported claims, inconsistent terms, and missing steps. Those checks matter, but they happen while the reviewer is looking directly at the content. The reader may be holding a tool, working on a phone, wearing gloves, listening for an alarm, or trying to remember a step learned last month. Context changes what “clear” means.

FDA’s updated guidance describes the chain explicitly: a user perceives information, interprets it, makes a decision, takes an action, and then responds to feedback. It also asks teams to account for users’ sensory and cognitive abilities, literacy, language skills, experience, training, and emotional state, along with conditions such as noise, low light, clutter, and distraction. The point is not to write for an imaginary “average user.” It is to design for the range of people and settings the product or process actually promises to support.

Plain language helps people reach the meaning faster. Descriptive headings and front-loaded paragraphs help them find the right section. The CDC Clear Writing Hub adds a further step by offering small-scale and paraphrase tests that check whether readers interpret a message as intended. None of those methods, by itself, proves that the person can carry out a multi-step task. For that, the content has to leave the document and meet a user.

Choose the tasks that deserve observation

Start with consequences rather than page count. In medical-device work, FDA defines a “critical task” as one that could cause serious harm if it is done incorrectly or not done at all. Other teams can use a wider version of the same question: where would a missed step, wrong choice, or delayed action matter most?

• Safety: Could the error injure someone, compromise care, or hide a warning?
• Compliance: Could it create an incorrect filing, disclosure, consent, or record?
• Money or access: Could the user lose a benefit, payment, deadline, or ability to continue?
• Recovery: If the person makes the wrong choice, will the system show what happened and help them recover?

Pick three to five tasks with the highest consequence or the weakest recovery path. “Read the onboarding guide” is not a task. “Choose the correct tax form for a contractor,” “set the home monitor before the first reading,” or “find what to do after a missed dose” is. A concrete task creates an observable finish line.

A five-part test for instructions

Fix the system before adding another warning

One of the sharpest points in FDA’s guidance is that changing the design is usually more effective than revising labeling or adding training. Instructions may not be available at the moment of need, and training depends on memory. That hierarchy is useful beyond devices. If five learners click the wrong control, a bold sentence telling them not to click it may be weaker than changing the control, its label, or the path that leads there.

Imagine a home test that requires the user to prepare a timer before collecting a sample. The instruction appears halfway down the page, after collection has begun. Users keep discovering it too late. A wording-only fix might capitalize “IMPORTANT.” A system fix moves preparation before collection, places the timer beside the materials list, numbers the actions in the order they occur, and adds feedback that confirms the timer has started. The text still matters, but now the whole sequence supports the action.

This distinction also keeps training teams from making a course carry every operational weakness. If the real workflow uses inconsistent field names, inaccessible documents, or screens that do not match the lesson, repair those sources of friction. The same principle appears in our review of online training that is harder than it needs to be: useful difficulty belongs to the skill, not to deciphering the interface.

A 30-minute preflight for an existing instruction

A small formative check will not satisfy a regulated validation requirement, but it can expose preventable problems early. Choose one important instruction and one representative user who did not write it.

1. Five minutes: define the user, the task, the setting, the successful outcome, and the consequence of error.
2. Five minutes: assemble the exact files, device or interface, and supporting materials the user will have.
3. Ten minutes: give the scenario, observe silently, and record actions, pauses, wrong turns, prompts, and recovery.
4. Five minutes: ask the user to explain what they thought the instruction meant at the point of difficulty.
5. Five minutes: choose one system or content change, assign an owner, and schedule a retest with someone new.

Keep a short evidence record: the content version, participant characteristics relevant to the task, test setting, scenario, outcome, prompts, observed problems, likely causes, revisions, unresolved risks, and reviewer approval. Health Canada’s current application guidance for near-patient diagnostic devices illustrates the higher-consequence version of this idea: representative users are evaluated under similar conditions of use while following the supplied instructions without assistance, and the clarity and legibility of those instructions are assessed.

For ordinary help content, the record can be lightweight. For medical, safety, legal, or regulated instructions, it belongs inside the organization’s controlled development and approval process. FDA’s separate May 2026 guidance on human-factors information in medical-device submissions explains what manufacturers may need to document for the agency; it should not be replaced by a generic editorial checklist.

Sources and further reading

• FDA: Applying Human Factors and Usability Engineering to Medical Devices, updated August 3, 2026
• FDA: Human Factors and Medical Devices
• FDA: Content of Human Factors Information in Medical Device Marketing Submissions, May 29, 2026
• Health Canada: Guidance on Class III and Class IV in vitro diagnostic device application content
• CDC: NCEH/ATSDR Clear Writing Hub and small-scale testing tools
• UK Office for National Statistics: How people read online
• GOV.UK: Test documentation with users completing realistic scenarios

This article provides general editorial and instructional-design information, not medical, legal, regulatory, safety, or quality-system advice. Requirements depend on the product, use, jurisdiction, organization, and risk. Follow applicable regulations, standards, approved procedures, and qualified human-factors review.