#Principle
Information becomes easier to process when related pieces are organized into meaningful units that can be understood as one structure.
#Why it matters
People rarely process a complex interface as an unstructured list of individual details. They build larger units: an address, a plan, a filter group, a table row, a billing section. Good chunking reduces the number of independent items that must be held in mind and exposes the structure of the task. Arbitrary grouping creates the opposite effect because users must first decode why items were grouped together.
#Deeper design interpretation
Chunking reduces the number of independent relationships users must manage at once. A well-formed chunk gives several details a shared meaning: an address becomes one concept, a billing section becomes one task, a row becomes one record. The interface helps users reason in units closer to the domain instead of individual controls.
Effective chunks are semantic rather than merely visual. Two columns do not automatically form two meaningful groups. Likewise, splitting a form into many cards can increase navigation and scanning cost if the sections do not correspond to recognizable sub-tasks.
Layr should ask whether the grouping helps users answer a question, make a decision, or complete a coherent piece of work. The best chunk often has a useful name that a user would naturally use.
#What the agent should reason about
Inspect headings, labels, data relationships, sequence, and repeated task boundaries. Determine whether the grouping reduces the amount of context the user must carry forward. When groups are arbitrary, the agent should recommend reorganization around user concepts rather than simply adding more containers.
Chunk boundaries should also survive responsive layout changes. A desktop grid may create obvious groups that disappear when stacked on mobile.
#Common failure patterns
- Long settings pages mirror implementation modules instead of user concepts.
- Related comparison values are separated across distant panels.
- Mobile stacking removes the visual boundary between sections.
- A section heading describes only half of the controls beneath it.
- Every field becomes its own “chunk,” producing fragmentation without reducing mental work.
#Audit questions
- Can each group be named in the user's language?
- Does the chunk correspond to one decision, object, or sub-task?
- Are relationships easier to understand inside the chunk than outside it?
- Does responsive reflow preserve the same conceptual structure?
#Example agent finding
Medium — Notification settings are grouped by backend channel instead of user intent. Email, push, and SMS controls are separated into three cards, forcing users to reconcile the same alert type across distant sections. Group by notification purpose first and expose channel choice within each group so each chunk corresponds to one user decision.
#Agent directive
Group information around meaningful concepts or sub-tasks. Prefer chunks that match the user's mental model and can be named clearly.
#Apply this when
- Long forms, settings, onboarding, documentation, dense dashboards, tables, command palettes, menus, and multi-step flows.
#Observable signals
- A long flat list contains items with obvious conceptual subgroups.
- Users must compare distant values that belong to the same decision.
- A section contains so many independent controls that its heading no longer describes one coherent idea.
- Groups exist only because of layout columns rather than semantic relationships.
- A mobile layout destroys grouping that was clear on desktop.
#Recommended actions
- Create sections around concepts users can recognize and name.
- Keep a chunk internally coherent and avoid mixing unrelated decisions.
- Use headings, spacing, and containers proportionally to the strength of the grouping.
- Preserve chunk boundaries when responsive layouts reflow.
#Do not over-apply
More sections are not automatically better. Excessive fragmentation forces users to navigate artificial boundaries. A chunk should reduce mental organization work.
#Example
Weak
A settings page lists 28 toggles in one continuous column sorted by implementation order.
Stronger
Controls are grouped by notifications, privacy, billing, and workspace behavior, with each section containing related decisions.
#Evaluation notes
- Default severity: Medium for complex tasks; low for short content.
- Typical confidence: High when semantic groups are inferable from labels and data.
- Primary evidence: Headings, DOM sections, field semantics, spacing, task sequence.
- Related rules: Cognitive Load, Working Memory, Law of Proximity, Law of Common Region