Activity attention and troubleshooting
Read the operation's fields and status before deciding whether it needs another attempt or a different owner.
Evidence first
Clear ownership
Verified next step
Activity does not currently have a separate notification inbox
The current page has an Attention panel with up to six recent queued, running, partial, failed, or cancelled operations. It is derived from the returned dashboard data. It does not send, acknowledge, dismiss, or retain notifications as a separate system.
Use Attention for triage and Timeline for the returned set
There are no Relevant and All notification tabs in the current Activity surface. Attention is a short triage list; the timeline is the fuller returned operation set. Both inherit source, viewer, time-window, filter, and cap boundaries.
Investigate partial and failed operations in their source
A missing row has several possible causes
- Wrong workspace, 30-day rather than 90-day window, or a selected calendar day.
- Category, status, initiator, or cost filter applied on the server.
- Free-text search hiding a returned row while aggregates remain unchanged.
- Source read, usage read, contributor profile, or 100-row timeline cap.
- Viewer-scoped Action Center work or hidden organization billing data.
- A product action that is not among the normalized Activity sources.
Preserve the support code exactly
Copy the code together with workspace, operation, scope, start time, current status, and visible counters. Do not paraphrase it. The code is designed to support diagnosis without exposing raw internal errors.
Keep the support code with a reproducible troubleshooting record so the next person can inspect the same state.
Separate hidden, partial, and unmatched cost
No money controls means the viewer lacks billing access. An at-least spend label means the usage read was capped. A cost-only row means a settled charge was retained but could not be linked confidently to one normalized operation. None of these states is the same as zero cost.
Treat the yearly calendar as a day index
The calendar uses maintained daily rollups when complete and a bounded live fallback while backfilling. It counts selected source operations by day and deliberately deduplicates some comparison and device-group records. Use the 30-day or 90-day dashboard and source product for row-level diagnosis.
Escalate with a reproducible record
- Workspace, browser time zone, selected window, filters, and search text.
- Generated date, visible count, capped state, and relevant Coverage rows.
- Operation title, category, scope, timestamps, status, counters, and initiator.
- Support code and billing state when visible, without sharing confidential tokens.
- Expected output and current state in the owning product.
Example: A keyword refresh that returned fewer rows
A keyword data refresh appears in Attention as partial. The strategist expands the row, records operation, scope, time, counters and the support code exactly as shown.
From the row link she opens the Keyword Database, finds which rows returned no data and retries only those, instead of repeating the whole paid request.
Frequently asked questions
Does Activity send notifications?
Not as a separate inbox. The Attention panel shows up to six recent queued, running, partial, failed or cancelled operations from the returned data.
Why is an operation missing from Activity?
Check the workspace, the 30 or 90 day window, active filters, free text search, source caps and your permissions. Any of them can hide a row that exists.
What should I send to support?
The support code exactly as shown, plus workspace, operation, scope, start time, status and counters. The code allows diagnosis without exposing internal errors.