Saved views are quiet promises about which cases belong in a queue. When the category list changes, an old view can hide new work or surface cases nobody owns. A short audit keeps the team from trusting a filter that no longer means what its name says. This audit is written for the support lead who owns the queue design and the agent who relies on the views during a shift.
What a saved view actually does
A saved view is a stored filter with a name. The name is what agents read, but the filter is what the tool executes. When the two disagree, the agent sees a queue that looks normal and is actually wrong. The disagreement is easy to miss because the view continues to load without an error.
Views often carry assumptions about category names that were true when the view was built. A view called Returns might filter on a category called Returns and Refunds. If the category is later split into Returns and Refunds, the old filter may catch only one half. The view still opens, but it no longer shows the work its name promises.
Views also carry assumptions about ownership and status. A view that shows unassigned cases may rely on a field that the new workflow no longer uses. A view that shows overdue cases may use a timestamp that the migration changed. The filter logic has to be read, not inferred from the label.
List every view and its owner
Export the list of saved views from the tool and put it in a spreadsheet. Include the view name, the filter definition, the owner, the team that uses it, and the last review date. Do not accept a view with no owner, because an unowned view is a decision nobody is maintaining. If the tool does not expose the filter definition, ask the administrator for an export.
Group the views by purpose. Some views support daily triage, some support quality review, and some exist for a report that runs once a month. The purpose determines how much drift the team can tolerate. A daily triage view needs a tighter review than a historical report that nobody uses for live work.
Mark the views that are used by more than one team. Shared views create shared assumptions, and a change made for one team can break another. The owner of a shared view should be the person who can represent both teams, or the view should be split. A shared view with no clear owner is a common source of hidden work.
Read the filter logic against the new categories
Compare each filter with the current category list. Look for old names, deleted names, and names that now mean something different. A filter that references a deleted category may return nothing, which looks like a quiet queue rather than a broken view. A filter that references a renamed category may return the wrong work.
Check the operators as well as the values. A view that uses is not may behave differently after a category is split, and a view that uses contains may catch unintended matches. The operator is part of the meaning, so the audit should record it. Small operator changes can create large differences in a live queue.
Look for filters that combine category with status or ownership. These combinations are where stale logic hides, because the category may be correct while the status field has changed. Read the whole filter as a sentence and ask whether that sentence still describes the work the view is meant to show. If the sentence is hard to read, the view is hard to trust.
Test with cases that should and should not appear
Pick five cases that should appear in the view and five that should not. Open the view and confirm the result for each case. This test is more reliable than reading the filter because it shows how the tool interprets the logic in practice. Record the case references and the expected result so the test can be repeated.
Include edge cases such as reopened cases, transferred cases, and cases with a missing category. Those are the cases most likely to fall through a stale filter. If a case has no category, decide whether the view should show it or hide it, and make that decision explicit. A view that silently drops uncategorized work hides the very cases that need attention.
Run the test with an agent who uses the view every day. The agent will know which cases are missing before the test begins, because the absence is felt in the workflow. Ask the agent to name a case they expected to see and could not find. That example is often the fastest route to the broken filter.
Retire or rebuild
Give each view one decision. Keep means the filter still matches the name and the work. Rebuild means the view is useful but the logic needs to change. Retire means the view should be deleted because the work has moved or the purpose is gone. Write the decision and the owner beside the view so the choice is visible.
When rebuilding, change the name if the meaning changes. A view called Returns that now includes refunds should say so, or it should be split. A name that stays the same while the logic changes creates a new kind of drift. The name is the part agents read, so it has to match the work.
When retiring, check whether any schedule or report depends on the view. A view that looks unused may be feeding a weekly export that nobody has traced back. Ask the administrator to confirm the dependencies before deletion. A retired view that breaks a report is a worse outcome than a view that stayed for another month.
Brief the team and set the next review
Tell the team which views changed and what the new names mean. A short message with a before-and-after list is more useful than a long policy document. Put the message in the shift handoff so the next shift sees it before the queue opens. Agents who understand the change are less likely to fall back on an old view.
Set the next review when the category list changes again. The review should be triggered by the change, not by a calendar date that arrives after the damage. Add the review to the category change checklist so it is not forgotten during a busy release. The queue naming standard and this audit should be run together.
Close with a dated record of the views reviewed, the cases tested, and the decisions made. Store the record with the queue documentation so the next reviewer starts from evidence. A view audit that leaves no record becomes a one-time cleanup instead of a maintained practice. The record should also name the next review trigger so the audit does not wait for another complaint.
