Prerequisites
The space is structured, and there is a surface to host the views. The actor's objects and activities are named, and the collection exists as a named thing with settled attributes and labels. That naming happens before this sequence, and no pattern here makes it.
The space is structured and a surface is waiting for views. Every collection needs the framing steps. Search, links between items, and a second view come later, when the collection calls for them.
Framing the collection
Build the item's ladder — Item view
Decide how an item looks at each level of detail, from a glyph or a reference through a summary to the full item, and what keeps it recognisable at every level.
Frame the collection as a named view — Data view
Choose the query (which items), the representation (which compact level of the ladder they show at, such as a row, a card, or a pin), and the arrangement. Set the defaults the actor meets before changing anything.
Decide the labelling layer — Tag · optional
Where items belong to several categories at once, or the words for them are still changing, let items carry labels instead of a single filed place. Skip this where each item has one settled category.
Decide what each view carries — Attribute visibility
Choose which attributes each view shows, including the labels, as a starting set, and let the actor change it.
Open the arrangement and the query
Three steps in no set order, and most designs take all three. They come after the attribute step because an attribute has to be visible before it can be sorted by, grouped along, or filtered on.
In any order:
Decide which edits pass through the view — Data view, its writable dimension
Decide which representations accept edits, what each edit changes in the underlying items, and a way to make every drag-based edit without dragging. Editing inside the view follows making changes.
Leaves: The actor meets the collection framed, and can see and change the framing. Framings pile up, and naming, sharing, and pruning them is decided in making sense at scale.
Adding search
Applies when
Narrowing is no longer enough: filtering leaves the actor at a dead end, their criteria are not ones the attributes can express, or they know the item's name rather than its properties. The information architecture is settled, and any labels are facets the search can narrow by.
Give intent an entry — Searching
Decide between keyword and natural-language queries, global or local scope, and how the actor's words are translated into the system's vocabulary.
Frame the result set
Treat the results, ordered by relevance rather than by the collection's own structure, as a named view, and take the framing steps again over them.
Keep the search reversible — Searching, its stability obligation
Keep the context the search was launched from, so returning finds the actor's scroll position and framing intact.
Cross-references in place
Applies when
The collection's items are documents with many links between them. A collection of records with few links skips these steps.
Let links be sampled in place — Link preview
Preview links on hover or focus, with increasing commitment: peek, open alongside, open fully. Use the item ladder's summary level as the preview.
Surface the unauthored connections — Dynamic hyperlinks
When the collection outgrows what anyone linked by hand, show inferred connections as ranked suggestions, never as authored structure, and give them previews, since nobody chose them.
Splitting and coupling
Applies when
One view no longer serves both scanning many items and checking one, or several views have each earned a place by showing a different aspect.
Split the view — Overview and detail
Show the collection as two linked views: an overview with every item at the same compact level, and a detail view of the current item in full. Use the named view as the overview and the item ladder for both sides.
Decide the content dimension per view — Attribute visibility
Decide which attributes appear in which view. The overview's set decides what can be compared, sorted, and filtered, so let the actor change it.
Decide layout and composition
Place the views side by side, in place, in a popover, or on a new page, and decide how many there are and how they nest.
Couple the views — Coordinated views
Once there are more than two views, carry the actor's selection or focus between them, so relationships across views can be seen. Couple views only where it clearly helps; two views linked by a current item do not need it.
Connections
Hands off to Making changes
At Framing the collection, via Selection
a framed collection makes picking items cheap, and picking is where that sequence starts
Calls Making changes
At Framing the collection, step 6, via Editing in place
editing inside the view follows that sequence's in-place arm
Hands off to Calibrating reliance
At Cross-references in place, step 1, via Citation
the link preview is also how a citation shows its source
Calls Making changes
At Splitting and coupling, step 4, via Selection
what coupling carries between views is sometimes a selection, which that sequence defines