Problem
One panel in a grid of twenty has something interesting in it. At a sixth of the screen the viewer can see that it moved and cannot see when, by how much, or which series. They do not want to navigate away, because the surrounding panels are the context.
Solution
Expand in place. The panel takes the full page, keeps its query and its time range, and closes back to exactly where it was.
Shneiderman names details-on-demand as one of the seven tasks and the ordering matters: it comes after overview, zoom and filter, which is to say the viewer has already decided this is the interesting thing. The interaction’s job is to be cheap in both directions, because it will be used dozens of times in a session and most expansions end in “no, not that one”.
Reversibility is the whole design. If closing is as cheap as opening, viewers expand freely and the pattern does its job. If closing loses the scroll position, resets the grid, or drops the time range, they learn not to expand, and the detail may as well not exist.
The three common forms trade differently.
Full-screen panel gives the most room and the strongest context switch, which is right for a chart with many series and wrong when the neighbouring panels are the reason you are looking.
Side panel keeps the page visible and works well when the detail is a record rather than a chart—a log line, a host, an issue.
Inline expansion pushes the grid down, which preserves the most context and disturbs the layout most, and is the option that most often makes the surrounding page unusable while open.
The subtle trap is state. An expanded panel is a different view of the same query, and if expanding changes the interval, the aggregation or the series limit, the viewer sees something that disagrees with the small version they were just looking at. That disagreement is rarely explained and quietly undermines trust in both.
Use when
Panels are necessarily small, the page’s arrangement is the context, and viewers need occasional close reading rather than navigation.
Don’t use when
The detail belongs to a different subject. That is drill-down, and dressing a navigation as an expansion makes back behave unexpectedly.
Trade-offs
Expansion state rarely survives a link, so what you found is not what a colleague opens. It is nearly always mouse-driven, so touch and keyboard get a worse version or none. Modal expansions trap focus and are frequently the least accessible thing on the page. And the pattern lets teams justify panels too small to read on the grounds that they can be expanded, which quietly degrades the default view everybody actually uses.
Checklist
- Is closing exactly as cheap as opening?
- Does closing restore scroll position and grid state?
- Does the expanded view use the same query, interval and time range as the small one?
- If it uses a finer interval, does anything explain the disagreement?
- Full-screen, side panel or inline, and does the choice suit the content?
- Does the expansion go into the URL so it can be shared?
- Is it reachable by keyboard, and does focus return correctly on close?
- What is the touch equivalent?
- Is any panel too small to read in its default state, relying on this to be usable?
- Does expanding trigger a fresh query, and how long does it take?
Compare
Grafana offers full-screen panel view from a keyboard shortcut or the panel menu, keeps the dashboard’s time range, and puts the view in the URL so an expanded panel is linkable, which is the detail most implementations miss. Sentry favours the side-panel form for events and traces, keeping the issue list visible, which suits a queue where the next item matters as much as the current one. Datadog uses full-screen with the surrounding scope preserved, so an expanded graph stays filtered to whatever the page was showing. Honeycomb barely needs the pattern, because the result of a query already occupies the page and there is no grid of small panels to escape from.
Related
Drill-down is the navigate-away sibling and the entry that covers moving to a different subject. Hover detail is the cheapest form of the same task. Collapsible row applies the idea at section scale. Panel grid is what has to hold still while a panel expands. Two-pane list and detail is the layout that makes this permanent rather than on demand.