Read Apple Messages channel metrics
Open Apple Messages > Metrics to inspect channel activity. The dashboard separates Outbound, Inbound, and Conversations because they measure different events.
You need channel metrics access in the workspace. Use Messages to investigate a specific record and business-outcome measurement for confirmed orders or bookings.
Select the period and scope
Choose the date range at the top of the report. Keep the same period when comparing tabs and check the displayed time-zone context.
On Outbound, the breakdowns include business, message kind, intent, group, category, and tag. Select a breakdown to inspect that scope. Outbound statistics support one dimension filter at a time. Return to the unfiltered report before comparing a different dimension.
Read the attribution note above the cards. Outbound outcomes are assigned to the time a message was accepted; a later failure can update an earlier day. Inbound and conversation statistics use the time of their events. Do not add these tab totals together as though they describe one population.
Read outbound activity and failures
- Accepted: Messages admitted for sending in the reporting scope, excluding processing rejections. Use this count as the denominator for send outcomes.
- Sent: Messages acknowledged by Apple's gateway. Compare this count with accepted messages and inspect failures; it is not a delivered count.
- Send-failure rate: Failed sends divided by accepted messages. It is No data when there is no denominator. Inspect error-code and message-kind breakdowns when it rises.
- Rejections: Messages refused before a send attempt, such as a billing rejection. Investigate these separately from provider send failures.
The dashboard labels the send-failure rate Risk at 5% or above. That display threshold is an investigation cue; your service may need attention at a lower rate. There is no Apple device-delivery or read-rate card.
Compare against the previous period, then open message history for the affected business, type, or error. A summary can show the scale of a problem; the message record gives its specific reason.
Interpret latency and first response
Processing latency measures from message acceptance to Apple's acknowledgement. It does not measure device delivery. The table reports p50, p95, and p99: respectively, half, 95%, and 99% of measured processing times are at or below those values.
First response measures from the oldest unanswered incoming message to acceptance of the first outgoing reply that answers it. Measurements are attributed to the reply. They are reported for the whole selected period, without a trend line.
Unanswered conversations have no completed first-response measurement. A missing value is not zero, and a good percentile can coexist with customers still waiting. Review the conversation queue alongside this measure.
Read inbound and conversation reports
Inbound shows received message activity and breakdowns such as business and intent. Use it to understand when customers contact you and which entry points generate requests. Compare like periods before changing staffing.
Conversations reports distinct conversations with lifecycle activity, plus Started, Reopened, and Closed event counts. A conversation can contribute more than one lifecycle event, so the event lines can exceed the distinct-conversation count. Closing does not establish that the customer's request was resolved.
These pages do not provide a per-agent productivity report, a customer-satisfaction score, or confirmed booking and payment totals. Use the systems that own those observations.
Explain unexpected totals
Match the report's date range, workspace, and outbound filter before comparing it with a message list. Recent processing can change an earlier outbound total; event-time reports can place the same exchange in a different period. Large distinct counts can be estimates.
Summary percentiles are calculated over the whole period. Do not average daily p95 values to recreate a period p95. If a discrepancy remains, collect the scope, time zone, range, and example message IDs using the support checklist.