.

Warning Effectiveness Checklist

Marc Green


Over the years, attorneys have often asked me about the criteria used to evaluate warning effectiveness. Below, I have compiled a checklist of requirements. I provide only a brief explanation, although each item merits a full discussion on its own.

The list has four categories, 1) those that are relatively unimportant although they often arise in litigation, 2) those that are necessary, but not by themselves sufficient (N-NS), 3) those that are important and 4) those that are not part of the warning itself but rather part of the development process.

Factor Importance Comment
Signal word Low Users do not readily distinguish among terms such as "warning, "danger" and "caution."
Color Low Users do not readily distinguish among colors (red, orange, yellow) as to warning relevance (although highly saturated colors may enhance urgency.)
Standard format Low There is little compelling evidence that specific formats, such as ANSI or ISO, have any special effectiveness.
Conspicuity N-NS The warning should be located where the user is likely to encounter it, especially on the first usage.
Legibility N-NS The warning text should be readable in the given environment (lighting, etc.) by the expected users (elderly, etc.)
Intelligibility N-NS The warning should use language that is understandable to the entire user population (children, nonexperts, etc.)
Explicitness N-NS The warning should be perfectly clear about the hazard, the prohibited behavior and the potential harm.
Causation High Humans reason causally, so if the causal connection between behavior and harm is not obvious, the warning should state it.
Taxation level High People do not buy products to read a warning. The time and effort required to read and to understand the warning adds a mental tax to product use. This tax should be at a minimum - the warning should be highly legible, clear and brief.
General appearance High People judge a message's importance partly by its appearance, so a warning of a major hazard should be large, bold, prominent, and neatly formatted. Conversely, an old, small, faded, messily scrawled warning suggests that the information is unimportant.
Overwarning High Too many warnings is as bad as too few. Warning against everything is the same as warning against nothing.
Dilution High Warnings about trivial hazards weaken the effects of warning on major hazards. Extra warnings also increase length and add to the tax.
Source credibility High The warning should come from a credible source and not appear as just another "cover your ass" warning. Warning appearance also plays a role here.
Enforcement High If a warning is not enforced, it will be taken less seriously. Especially when the user views other people ignoring the warning. (See next.)
Modeling High Users base their behavior largely on social norms and are less likely to comply with a warning when they see other people ignore it with impunity.
Commitment High The warning should appear before the user has made a large time and effort commitment to the product. The closer to reaching the goal, the less likely the user will stop short of the goal.
Competing affordances High Affordances are properties of objects that suggest their use. The product form should not suggest uses that are contrary to the warning.
Perceived risk High A warning is less effective when the user perceives the product as safe, so the warning should be especially strong on products and environments that seem innocuous.
Familiarity High Familiarity is perhaps the major enemy of warning effectiveness. Warnings on highly familiar products require especially strong warning measures.
Product transfer effects High If the user perceives a new product as belonging to a familiar product class, then familiarity effects transfer from that class. If the new product has unique hazards, special care must be taken to warn of the new hazards.
Contingency High The precise connection between behavior and consequence is critical. Warnings of negative consequences that are severe, immediate and certain are more effective than those that are weak, delayed and uncertain.
Cost of compliance High Compliance should not block an important goal without offering an alternative path to the goal or an alternative goal.
Cost of noncompliance High If the warning concerns a low-probability hazard, then the user may face little real cost of noncompliance. It may be useful to increase the cost of noncompliance through enforcement, fines, etc.
Consistency with mental model Very High Users filter information through their mental model of the world and how it works. Warnings that contradict entrenched parts of this mental model will be ineffective.
Validation testing Development A warning is not a warning until it has been proven effective in performing the function. Warnings on products that are novel and/or have significant risk should be tested.
Professional development Development Warnings should designed by human factors professionals and not by lawyers, by marketing personnel, by graphic designers or by untrained engineers.
Warning alternatives Development Warnings should not be used when more reliable safety methods such as redesign and guarding, are feasible. Safety responsibility should not be downloaded on to the user unless it is unavoidable.

The list is not meant to be exhaustive and there may be specific situations where the various items are more or less important or may not apply at all. There are also many other factors that affect warning effectiveness, such the user variables of age, gender, expertise, expectations, experience and concurrent tasks. In general, the creation of effective warnings requires "user centered design," where the focus is on the behavior of actual users rather than on some theoretical, ideal and highly convenient assumption of how users ought to behave.