Departing from convention must be significantly better to be worthwhile

What it is

Departing from an interaction convention people know brings relearning, misoperation, hesitation, support, and migration cost. A new practice should therefore deliver a significantly better outcome in efficiency, accuracy, safety, accessibility, capability, or long-term value before assuming those costs. “Better” should refer to real tasks and users, not merely novelty in a demonstration or visual distinction.

Why it happens

Strong convention creates quick action–outcome expectation. On departure, old expectation creates early negative transfer, requiring attention to learn new symbols, positions, or flow; teams also assume explanation, documentation, compatibility, and recovery cost. If the change produces only a minor local benefit, transition loss may exceed it for a long time. If it eliminates serious errors, enables critical capability, or substantially improves efficiency, its value may offset learning cost.

Studying it

Define the old-convention baseline and the specific problem a new option solves, then compare efficiency, error, comprehension, accessibility, and longer-term retention across newcomers, experienced users, and target settings. Account for affected task frequency and population size rather than a single test's average time. Use staged rollout, reversible settings, or controlled comparison to observe real transfer, support requests, and unintended consequences.

Where it stops holding

Not every convention merits preservation: outdated, exclusionary, unsafe, or mismatched patterns should be challenged even if short-term data is imperfect. But “we need innovation” is not evidence; deviation without a clear benefit hypothesis often transfers cost to users. In a genuinely novel task, existing convention may be weak, allowing freer exploration—but it still needs learnable cues and recovery.

Applying it

• Before departure, state the old pattern's problem, the new approach's measurable benefit, affected groups, and migration and rollback plan.
• Retain the anchors users rely on most, introduce the new rule gradually, and provide clear feedforward and help on first and edge cases.
• Decide rollout from long-term segmented evidence; if benefit is unstable or helps only a few, revise, offer choice, or return to convention.

Related

• Same group: B2.11.1 Conventions come from broad exposure rather than inherent correctness · B2.11.3 Conventions evolve with platforms and need periodic review
• Nearby: B2.10.3 When consistency conflicts with optimization, evaluate transfer cost · B2.04 Mapping
• Search terms: convention deviation · innovation threshold · learning cost