Only a few icons reach cross-product consensus

What it is

A handful of graphics point at the same action in almost any product: a magnifying glass is search, an × is close, a trash can is delete, a triangle is play. That cross-product consensus is why an unlabeled icon button sometimes works. Most functions never enter this tiny set. Share, filter, more, favorite, and settings use different drawings across products, and the same drawing points at different actions. “Search can be icon-only” does not generalize to “filter can too.”

Why it happens

Consensus is statistical learning, not graphic transparency. After hundreds of apps have paired a magnifying glass with search, the mapping is nearly a word. Share still forks: a square with an arrow, three dots, a paper plane, a chain—each product teaching a different word, so the mapping never settles. A new metaphor has not yet had enough shared exposures, so identification is a guess. Worse is false consensus: the hamburger is widely used, yet a substantial fraction of people still do not know what it holds. “Everyone ships it” is not “everyone understands it.” The set also rewrites slowly: the floppy disk was once the public word for save; once the object vanished, new users lost the hop. Shipping a non-member function as an icon-only button asks people to read a dialect where there is no shared vocabulary.

Studying it

Compare products: collect icons for one function across several widely used apps, count how many distinct drawings exist, and test whether people can name the action without having seen that product. A matching task works too: give the action name and ask which candidate graphic would most likely trigger it.

Independent variables: whether the function sits in the high-consensus set, how scattered the market’s drawings are, breadth of the person’s product experience. Dependent variables: cross-product identification agreement, confusions with neighbors (filter vs sort, share vs export).

Testing only veterans of one product mistakes in-product convention for cross-product consensus. The sample has to include people who rarely use that class of app.

Where it stops holding

Vertical domains grow domain consensus: waveform and solo buttons work for audio engineers and fail for everyone else. Treating domain consensus as public consensus fails. Platform system glyphs (iOS share, Android overflow) are near-consensus inside that platform and scatter across the other; a cross-platform product cannot trust only one side. Emoji and icon fonts replace glyphs with OS versions; yesterday’s consensus drawing may not look like itself tomorrow. Consensus is not a license: a trash can still needs protection against accidental activation.

Applying it

• Keep a whitelist of icons allowed to appear unlabeled; by default only search, close, play/pause, forward/back, delete, and similarly well-tested items.
• Pair share, filter, more, settings, and favorite with words, or drop the words only after the product has exposed that function repeatedly.
• Do not assume a system glyph still reads as the same action on another platform; switch to a text button if needed.
• Verify with people who do not use your product, icons only. If a function outside the whitelist is named stably, then discuss dropping the label. If it is not, it is not in the consensus set.

Related

• Within the group: E1.04.1 Recognition of unlabeled icon buttons is far worse than expected · E1.04.3 Icon buttons need an accessible name
• Adjacent: F6.01 Icon ambiguity · F6.05 Culture-bound visual metaphor · E1.13 Icon-plus-text buttons
• Search terms: icon convention · universal icons · cross-product consensus