Review incident connections
Summary: You will use the Incident Connections review queue to browse proposed connections as a batch, compare each pair, and connect it into a case or dismiss it.
The Incident Connections review queue loads your organization's pending AI-proposed connections in one go (highest score first, up to a fixed ceiling of 2,000 pairs) and lets you filter, sort, and browse them as a batch. It is the batch counterpart to Event Link Analysis, which works from a single anchor incident.
Who can use it
Platform Administrators, Administrators, Supervisors, Managers, ORC Analysts, and Investigators can open the queue. All except Platform Administrators can act on a pair (Platform Administrators view only).
Choose a view
Three views over the same filtered set:
- Gallery (default): side-by-side photo cards per pair.
- Photo list: one row per pair with suspect and vehicle photos inline.
- Compact list: dense text rows.
All three show suspect and vehicle photos only: scene and CCTV attachment imagery is deliberately not shown here. Any photo can be zoomed in place (scroll, pinch, or double-click); a plain scroll over a photo still scrolls the page.
Filter and sort
A quick-filter bar offers one-tap chips (Strong, loss at or above $1k, incident type, three or more suspects, and so on). Each chip is tri-state: tap to show only matching pairs, tap again to hide them, tap again to clear. Which chips appear is customizable and remembered between visits in your browser.
A filter pane adds fuller facets, each showing how many pairs would match: territory (Division › Region › District), state, connection reason, license plate and vehicle make, loss range, suspect attributes, number of suspects, the ORC flag, incident type, product category, and score band.
Because every row is a pair of incidents, filters combine at pair level: a pair shows if either incident matches a "show only" filter, and a "hide" filter on either incident removes the pair. A Reset control clears all filters and the search term (but leaves the view mode alone). Your view, filters, search, and sort are remembered per browser.
Score bands
Rows and cards carry a colored band pill: Strong (80 and above), Medium (50-79), Weak (below 50). The page header shows a live count of pending pairs currently scored Strong: a running total of what is pending, not a "new since your last visit" figure. (These bands differ from the labels used elsewhere; they are not interchangeable.)
Act on a pair
- Review: opens the pair's network graph, anchored on the first incident with the second as a connected neighbor, so you see the fuller picture before deciding.
- Compare: a side-by-side dialog with both incidents' photos, an attribute comparison (incident type, loss, product, method, suspect count, with matches highlighted), and the score breakdown. From the same dialog you can connect the two incidents into a case or dismiss the pair.
- Dismiss (removes the pair from the queue, available as a row action or from inside Compare. There is no undo), no user action returns a dismissed pair to the queue.
Editing the underlying incidents later re-checks the score automatically, but your confirm or dismiss decision is preserved: the platform never silently re-opens or re-closes a call you made.
Edge states
- Empty queue: nothing pending; the Strong count reads zero.
- Filters match nothing: no rows; clear with Reset.
- More than 2,000 pending pairs: the queue loads the 2,000 highest-scoring and warns that the list was cut at that number, so filtering may be incomplete. There is no paging beyond the ceiling; the rest surface only as the queue is worked down.