Status: Public Preview — behaviour may change.
What it does
The Self Service tab in Real-time streaming analytics shows supervisors how self-service agents are performing right now: how many conversations they took, how many they contained, how many went on to a representative, and how long each outcome took. It sits alongside the Wallboard, Queues, Representatives, Conversations and Assisted service views.
Key facts
- Production-ready preview, under supplemental terms of use. Microsoft labels the article prerelease, so expect the metrics to move.
- Filters are Period, Channel, Business Unit and Queue. Live metrics ignore Period; every non-live metric respects it.
- Percentage metrics are calculated only from conversations completed in the last 15 minutes. A conversation that reached its outcome in that window but completed outside it isn't counted.
- Conversation volume split shows total conversations (inbound and outbound), conversations offered (to self-service agents versus straight to representatives) and active conversations (with self-service versus with representatives).
- Self-service agent performance covers engaged conversations (active and completed), containment rate broken into Contained, Escalated, Fallback and Customer abandonments, top 10 breakdowns by last language and by final intent, and average duration, time to escalate and time to contain.
- A per-agent table lists engaged, active, escalated and contained counts, escalation and containment rates, and the three average timings, with a search box for finding one agent.
- See details on any tile opens the underlying metric for a closer look.
When to use / skip
Use it once you have self-service agents taking real volume and a supervisor who'll look at it. The language and intent breakdowns are the useful bit: they show where the bot is losing people without anyone building a report. Skip it if you're still piloting with a handful of conversations a day, or if your supervisors are staying on the historical analytics.
Configuration decisions
- Which supervisors get the Self Service tab, and whether they read it alongside Queues or on its own.
- Whether containment is judged on the 15-minute live figure or the period figure. They answer different questions.
- Which channels and business units are filtered by default for each supervisor group.
- How you'll treat the Fallback category when agreeing targets with the bot team.
Gotchas
- Completed includes failures. Completed conversations cover resolved, escalated and abandoned ones, plus conversations that end because the bot failed. A big completed number isn't a big containment number.
- The 15-minute window moves the percentages. On a quiet queue a single conversation swings the rate wildly, so don't read much into it outside busy periods.
- Live and period metrics differ by design. Two tiles on the same screen can reflect different time ranges, and supervisors will assume they don't.
- It's not for people decisions. Microsoft states the per-agent view isn't for employment decisions, and customers are responsible for the legal side of monitoring and notification.
Consultant notes
- Agree the definition of containment with the business before anyone sees this dashboard. Fallback and abandonment sit in the same breakdown and people argue about which counts as a failure.
- The top-10 intent view is a cheap way to find content gaps, but it only helps if the bot team has a route to act on what supervisors see.
- Keep expectations modest: it's a live operational view, not a trend report. Take weekly containment conversations to the historical analytics.
- It's a preview, so keep it out of any SLA or contractual reporting until it reaches GA.
Worth revisiting when the Self Service view leaves preview or the metric definitions change.