Dynamics 365 Customer Service Analytics: Reports, KPIs and Power BI

Abstract illustration of translucent teal glass bars of different heights rising from a reflective surface, with one bar glowing orange at the top

Last updated: November 18, 2026

By ZapAI Team

Dynamics 365 Customer Service analytics tells support leaders what happened, what’s happening now and why. Our Dynamics 365 Customer Service consultants set up reporting for mid-market US support teams, and this guide covers the built-in reports, the KPIs worth tracking, CSAT surveys, Power BI model customization, Fabric, licensing and a readiness checklist, checked against Microsoft Learn on November 18, 2026.

Dynamics 365 Customer Service analytics combines historical dashboards, real-time supervisor reports, knowledge, Copilot and AI agent reports. Start with the built-in reports, extend them with Power BI model customization, and move to Microsoft Fabric for cross-system analysis.

Here’s a hypothetical. A support director opens the Monday review with three numbers. CSAT is up. Backlog is down. SLA attainment looks great. Then a regional manager asks why escalations doubled, and nobody in the room can say whether the three numbers even describe the same cases.

That’s the real problem. Not missing data. Mismatched definitions. Our Dynamics 365 Customer Service guide covers plans, agents and fit at a high level, so this post stays narrow and goes deep on measurement, from report inventory to metric fine print.

Which Analytics Are Built Into Dynamics 365 Customer Service?

Seven report families ship in the product. You enable them in Copilot Service admin center under Insights, and supervisors open them from Copilot Service workspace.

Report familyWhat it coversFreshnessWho it’s for
Customer Service historical analyticsCase volume, escalations, resolve time, CSAT, representative performance and AI-discovered topicsRefreshed once every 24 hoursSupport managers
Case analytics with enriched dataSeven pages on volume, response and SLA, reopens, aging, agents and queuesDaily, with backlog at the latest capture onlyOperations leaders
Omnichannel historical analyticsConversations, queues, voice, voicemail, representatives and AI agentsDailyContact center managers
Omnichannel real-time analyticsLive volume, service levels, capacity, ongoing and backlog conversationsNear real timeFloor supervisors
Real-time analytics for record routingCase and record work items by queue and representativeNear real timeCase queue supervisors
Knowledge analyticsArticle usage and internal search terms, including searches with no resultsHistoricalKnowledge authors
Copilot analyticsHow representatives use Copilot and handle time with and without itHistoricalService managers and AI owners

Two facts shape everything downstream. Microsoft’s Customer Service dashboards page says historical data refreshes every 24 hours, you can’t trigger a refresh yourself, and data is kept for 24 months. And if an environment sits unused for 30 days, refresh pauses, with the feature deprovisioned after 60 days. Sandboxes go dark quietly that way.

October 30, 2026 brought the newest piece. Microsoft’s case analytics overview describes a Shared Analytics Platform model with common definitions across seven report pages, and it warns plainly that the new dashboard isn’t a direct replacement for the older historical report. Compare both before you swap the one your executives see.

Access is role-based. System Administrator and supervisor roles see the dashboards by default, and editing visuals needs the Analytics Report Author role.

Which Legacy Reports Were Retired, and What Replaced Them?

Quite a few. If a blog post, consultant deck or old runbook mentions any of these, treat it as dated.

Retired itemKey datesMicrosoft’s replacement
Intraday insights for OmnichannelDeprecated May 1, 2023. Unavailable to existing orgs from April 30, 2024Omnichannel real-time analytics
Power BI template reports for Customer Service Analytics and Omnichannel InsightsDeprecated November 6, 2023. Availability ended February 28, 2024Built-in historical analytics reports
Customer Service Insights standalone appDeprecated, no removal date listedEmbedded analytics in the Customer Service apps
Historical analytics for unified routingDeprecated June 30, 2025. Support ended August 31, 2025Omnichannel and record routing reports, plus conversation diagnostics
Legacy AI topic clustering for cases and conversationsDeprecated June 20, 2025. Removed after August 31, 2025Current AI-discovered topics
Forecasting for case and conversation volumeSupport ended October 30, 2026Forecast scenarios in workforce engagement management
Omnichannel agent and supervisor dashboard experiencesDeprecated December 1, 2025. Support ends December 1, 2026The inbox and Omnichannel real-time analytics

All dates come from Microsoft’s Customer Service deprecations list, updated November 2, 2026. The last row matters right now. Supervisors still working from the old Omnichannel dashboard have about two weeks of support left.

Routing diagnostics is gone too, removed March 31, 2026, and our unified routing guide covers its Application Insights replacement.

Which Customer Service KPIs Actually Matter?

Fewer than you think. Eight cover most support operations, but each one hides a definition choice that can wreck a quarterly review.

KPIPlain definitionWhere Dynamics 365 measures itCaveat
First contact resolutionShare of issues fixed on the first interaction, with no reopen or repeat contactNo single built-in measure we could find. Reopened rate is the closest case proxyDefine the repeat window yourself, such as 7 days, and write it down
Average handle timeTotal handle time divided by conversations handledOmnichannel real-time and historical reportsCounts the primary representative only by default, so consult time sits in a separate metric
CSATAverage satisfaction score from post-interaction surveysHistorical reports, if survey responses land in the expected tablesResponse-weighted, so a few angry repeat responders can swing it
SLA complianceShare of SLA timers metCase analytics Response and SLA pageExecution attainment and case breach rate aren’t mirror images
BacklogOpen cases right now, by age bandCase analytics Aging page, real-time backlog reportsThe case model keeps only the latest snapshot, so no trend line
Abandon rateShare of incoming conversations ended before a representative engagedOmnichannel real-time and historical reportsExcludes overflow by default, and short-abandon variants exist
Transfer rateShare of conversations moved between representatives, agents or queuesOmnichannel historical reportsAn AI agent handing off to a person counts as a transfer
DeflectionShare of AI agent conversations that never reach a personBot dashboardAbandoned bot chats count as deflected

Read that last caveat twice. Microsoft’s bot dashboard documentation says a conversation is deflected if the AI agent resolves it or if the user abandons it midway. A rising deflection rate can mean your agent got smarter. Or that customers gave up. Pair it with the resolved session rate and bot CSAT before anyone celebrates.

The conversation numbers have their own fine print. Per the conversation metrics reference, the default abandon rate excludes overflow conditions, while separate measures strip out short abandons below a threshold you set, which changes both abandon rate and service level depending on which version a report uses.

SLA math trips up even careful teams. The case metrics reference gives an example where three succeeded timers, one breached and one pending produce 60% attainment, and those five timers can belong to fewer than five cases. So 60% attainment doesn’t mean 40% of cases breached. Our sister post on Dynamics 365 Customer Service SLAs covers how the timers themselves get configured.

Benchmarks? We won’t give you any. Published contact center benchmarks rarely share your channel mix, case complexity or definitions, so your own trailing 90 days is the only baseline we trust.

Eight small translucent glass discs arranged in a loose grid, seven glowing teal and one glowing orange, representing a short list of support KPIs

How Do CSAT Surveys Work in Dynamics 365 in 2026?

Through two survey systems that don’t talk to each other. That’s the short version.

Customer Voice still exists. We found no retirement notice for it on Microsoft Learn as of November 18, 2026, the post-conversation survey setup page still uses it, and the Customer Service historical Summary report still describes its CSAT and survey sentiment as based on Customer Voice responses. Teams with years of Customer Voice data don’t need to panic.

Microsoft’s newer path is different. Its Copilot Studio survey guide describes survey agents created from Customer feedback in Copilot Service admin center, with CSAT, NPS, customer effort and blank templates, across messaging, voice and custom channels. A Power Automate flow can also email a survey when a case is resolved.

  • Responses land in the Customer feedback survey response table in Dataverse.
  • CSAT reaches the historical reports only if the score is stored as a number in the MCS_CSAT variable.
  • Surveys built in Customer Voice can’t be managed from the Customer feedback page.
  • Copying an environment copies survey agents that won’t work. Rebuild them in the target.

Voice teams already had a forced move. Microsoft’s Contact Center deprecations list says the old post-call survey toggle on the voice workstream Language tab was deprecated and removed on August 1, 2025, with Copilot Studio surveys as the replacement.

Our take? Pick one survey system per channel and stick with it. Mixed plumbing produces two CSAT numbers that never reconcile, and the case model’s Survey Coverage metric isn’t a response rate either, so check what it actually counts before quoting it.

What Do Real-Time Analytics Show Supervisors?

What’s on fire this hour. Microsoft’s real-time Omnichannel dashboard overview lists Summary, Voice, Agents, Ongoing conversation, Backlog conversation, Bot and Proactive Engagement reports, plus Intent group and User group reports that need Dynamics 365 Contact Center.

Supervisors can assign, transfer, monitor and force-close conversations right from the Ongoing conversation report. Useful. Two blind spots, though. The Omnichannel real-time reports leave out the entity record channel, so case queues live in a separate real-time report for record routing, and they skip conversations an AI agent fully resolved.

That record routing report needs the CSR Manager role and works best with capacity profiles, since Microsoft says unit-based or mixed capacity makes it less actionable. Phone and chat channel setup is covered in our Dynamics 365 Contact Center post.

One warning sits at the top of Microsoft’s page. The reports aren’t meant for decisions about pay, rewards or other employment matters. Coach with them. Don’t rank people for raises off them.

How Do Copilot, AI Agent and Knowledge Analytics Work?

Three reports answer three different questions.

Copilot analytics tracks representative use of Copilot. It shows daily active users, responses copied, the share of conversations using Copilot, handle time with and without it, and thumbs-up or thumbs-down ratings. The underlying interaction tables sit in Dataverse, so you can build your own adoption metrics on them.

The Bot dashboard covers AI agents talking to customers. Expect escalation, resolved, abandon and deflection rates, session outcomes by reason, and topic-level drill-downs. It needs conversation transcripts saved to Dataverse, a setting in Power Platform admin center that’s easy to skip. Bot CSAT shows blank unless your agents ask for feedback.

Knowledge analytics looks inward. Article insights and search term insights flag the top internal searches, including searches that returned nothing, which is the cheapest content gap list you’ll ever get. It covers representative searches only, not your customers’ portal searches. Our knowledge management guide shows what to do with that gap list.

Contact Center customers also get a Quality Assurance Agent dashboard in Omnichannel historical analytics. Customer Service-only orgs don’t.

When Should You Customize the Power BI Data Model?

When the built-in metric is close but not quite yours. Microsoft’s own examples include a 150-second service level the default model doesn’t ship, handle time that includes subject matter expert consult time, and resolve time by a custom region field.

Microsoft offers two layers before you write any pipeline. Visual customization lets an Analytics Report Author change charts, filters and pages without touching the metric logic. Data model customization, per Microsoft’s model customization guide, deploys the out-of-the-box semantic model into a service-managed workspace and editable report copies into a Fabric workspace you choose, so your authors can add DAX measures, bring in Dataverse or outside data through a composite model, and embed up to 40 custom reports back in Copilot Service workspace.

Setup isn’t a toggle. You need an Entra security group containing the Dynamics 365 Analytics service account, six Fabric tenant settings switched on, a paid Power BI service and up to 24 hours of provisioning. Then the limits.

  1. Knowledge analytics and unified routing analytics can’t be customized.
  2. The historical model refreshes daily, so custom reports aren’t any fresher than the built-ins.
  3. Report copies are created once and never updated by Microsoft.
  4. Supervisor actions like assign and force close don’t work on customized reports.
  5. The feature applies to Customer Service and Contact Center standalone, not Contact Center embedded.

Build on the managed model. Don’t copy it. Microsoft updates the base model with new metrics and fixes, and a local copy drifts the day you make it. If your team needs help with DAX and semantic models, our Power BI consulting team does this work, and Copilot in Power BI can draft measures once the model is clean.

A large teal glass cube with a smaller orange glass cube attached to one face, linked by thin glowing lines, representing a custom model built on a base model

When Does Dataverse Link to Fabric Make More Sense?

When the question crosses systems. Support data plus billing, product telemetry, field visits or a workforce management tool usually outgrows a composite model.

Link to Fabric creates a lakehouse, SQL endpoint and Power BI semantic model from your Dataverse tables, using shortcuts so the data stays governed by Dataverse. Microsoft’s Link to Fabric documentation adds three practical notes. It adds every non-system table with Track changes enabled, it consumes Dataverse database storage, and today one environment links to one Fabric workspace.

Fabric gives you raw tables, not ready-made KPIs. Microsoft’s analytics reporting journey page is blunt about it. A shortcut provides selected tables without the default relationships and measures, so you reimplement and validate every business calculation yourself, and shortcuts to Microsoft’s processed contact center tables need separate entitlement or onboarding.

That’s real engineering. Our posts on Dataverse link to Fabric and Fabric vs Power BI Premium cover sync options and capacity buying.

Built-In vs Power BI Model Customization vs Fabric

FactorBuilt-in reportsPower BI model customizationFabric
Starting pointMicrosoft reports and metricsMicrosoft semantic model plus editable report copiesDataverse tables or processed analytics tables
Microsoft-defined measuresIncludedIncluded and extendableNot included
Outside dataNoYes, through a composite modelYes, any source in OneLake
FreshnessDaily historical, near real time for live reportsDaily historical, live for the real-time modelDepends on your sync and pipelines
Skills neededAdmin and supervisorPower BI author with DAXData engineer plus Power BI author
Licensing beyond Customer ServiceUsually none for viewingPaid Power BI service and Fabric workspaceFabric capacity plus Dataverse storage
Best fitMost teams in year oneCustom KPIs on standard support dataCross-system analytics and data science

Microsoft’s own advice on that journey page is to start with the least complex option that meets your needs. We agree. Most teams we review jump to Fabric a year early.

Which Licenses Do You Need for Customer Service Analytics?

Customer Service Enterprise or Premium, plus Power BI licenses once you build your own reports. That’s our reading of the pricing pages on November 18, 2026, because Learn only says specific licensing requirements apply.

ProductUS list price per user per monthAnalytics role
Customer Service Professional$50Analytics and KPI reporting not included
Customer Service Enterprise$105Analytics and KPI reporting included
Customer Service Premium$195Analytics and KPI reporting included
Power BI Pro$14Typical license for report authors who publish
Power BI Premium Per User$24Option for authors needing Premium features

Plan prices come from Microsoft’s Customer Service pricing page, which notes the apps include an option to embed Power BI content but a separate Power BI license may be required. Power BI prices come from the Power BI pricing page. Fabric capacity is a separate purchase, and whether viewers of customized reports need their own Pro license depends on your capacity, so confirm it against the licensing guide before you budget. Our Dynamics 365 services overview shows how Customer Service sits next to Sales and Field Service.

Example Dashboard Starter Set by Role

This is an example, not a Microsoft template. It’s the set we’d sketch first for a 20 to 60 person support team.

RoleOpen everyReport to start fromFour tiles
Floor supervisorHourOmnichannel real-time Summary, record routing QueueWaiting work, abandon rate, available capacity, oldest waiting item
Support managerDayCase analytics Aging and Response and SLAStale cases, awaiting first response, SLA breach rate, reopened rate
Support directorWeekCase analytics Summary, Omnichannel historical SummaryVolume trend, CSAT with response count, escalation rate, handle time
Knowledge leadWeekKnowledge analyticsSearches with no results, articles linked to cases, low-engagement terms, top articles
AI agent ownerWeekBot dashboard, Copilot analyticsResolved session rate, deflection with abandon split, escalation reasons, Copilot daily users

Four tiles each. Not twelve. If a tile doesn’t trigger an action when it turns red, cut it.

What Mistakes Break Customer Service Reporting?

In our experience, these cause most of the arguments.

  • Comparing old historical analytics with the new case analytics without matching filters first.
  • Reading backlog trends from a model that only keeps the latest snapshot.
  • Calling a case stale because nobody touched it, when automated updates reset the clock.
  • Celebrating deflection without splitting out abandons.
  • Quoting CSAT from 11 responses as if it were a trend.
  • Customized case or user tables that leave the built-in reports blank, which Microsoft warns about.
  • Copying the managed semantic model instead of building on it.
  • Letting a sandbox sit idle past 60 days, then wondering where analytics went.

A note on the date filter. In case analytics, it selects cases by creation date, not by when they were resolved or breached, so last month’s numbers keep moving as old cases close. Annoying? Yes. Wrong? No. Just label it on the slide. Our sister post on Dynamics 365 case management covers the case fields these reports depend on.

A vertical stack of thin translucent glass plates glowing teal, with the top plate glowing orange, representing a step by step readiness checklist

What Belongs on an Analytics Readiness Checklist?

  1. Confirm Enterprise or Premium licenses for everyone who needs analytics.
  2. Write a one-page KPI dictionary with formulas, filters and owners.
  3. Standardize case fields. Origin, priority, category and resolve reason must be required and clean.
  4. Pick one survey system per channel and test that CSAT reaches the reports.
  5. Enable historical, real-time and knowledge analytics in Copilot Service admin center.
  6. Grant roles. CSR Manager, supervisor, Analytics Report Author or a custom read role.
  7. Save conversation transcripts to Dataverse if you run AI agents.
  8. Move supervisors off the retiring Omnichannel dashboard before December 1, 2026.
  9. Run old and new case analytics side by side for 30 days and reconcile.
  10. Only then decide on model customization or Fabric.

ZapAI Decision Rules for Customer Service Analytics

These are our practice opinions from projects, not Microsoft guidance.

SituationOur rule
First 90 days on Customer ServiceBuilt-in reports only, plus a written KPI dictionary
One or two missing measuresVisual customization first, model customization if a new measure is unavoidable
Custom fields or regions drive the business reviewPower BI model customization built on the managed model
Support data must join billing, telemetry or field dataDataverse link to Fabric with a curated semantic model
Fewer than about 25 survey responses a month per queueReport CSAT quarterly with the response count beside it
AI agent live on chat or voiceNever show deflection without resolved and abandoned sessions beside it
Leadership asks for twenty KPIsFive KPIs with owners first. Add more after two clean months

Bias disclosed. We sell Power BI and Fabric work, and we still tell most clients to wait.

Wrapping Up

Define before you display. Use the built-in historical and real-time reports first, check every KPI against Microsoft’s metric definitions, fix your survey plumbing, then extend with Power BI and finally Fabric when the questions demand it. Our post on how Power BI supports better decisions covers the habit side of this.

ZapAI is a Microsoft Dynamics 365, Power Platform, Fabric and AI consulting firm in Irving, Texas, working with support teams across Dallas-Fort Worth and nationwide. We set up Customer Service reporting, build Power BI models on the standard data and connect Dataverse to Fabric when it’s justified. Want a second opinion first? Ask for a service reporting review, and we’ll walk through your KPIs, reports and data quality with you.

Questions Support Leaders Ask About Customer Service Analytics

Is analytics included in Customer Service Professional?

Not per Microsoft’s pricing table, which lists analytics and KPI reporting for Enterprise and Premium only. Professional orgs can still build standard Dataverse charts and dashboards.

How fresh is the data in historical analytics?

Up to a day old. Historical reports refresh once every 24 hours, you can’t force a refresh, and each page shows a Last updated time so you know what you’re looking at.

Why is my CSAT tile blank?

Usually the survey data isn’t landing where the report reads it. Copilot Studio survey scores need to be stored as numbers in the MCS_CSAT variable, Customer Voice responses need to reach the survey response tables the report reads, and bot CSAT stays empty until your AI agents actually ask customers for feedback. Check one real response end to end before you rebuild anything.

Is Customer Voice being retired?

We found no retirement notice on Microsoft Learn as of November 18, 2026. Microsoft’s newer survey path is Copilot Studio survey agents, and the old voice post-call survey toggle was removed in August 2025.

Can I add my own fields to the built-in reports?

With data model customization, yes. You extend Microsoft’s semantic model with a composite model in your own workspace, which needs a paid Power BI service and some Fabric tenant settings.

Power BI model customization or Fabric, which comes first?

Model customization, almost always. It keeps Microsoft’s tested measures, while a Fabric shortcut hands you tables and expects you to rebuild every calculation.

Does deflection rate prove our AI agent works?

Not on its own. Microsoft counts abandoned bot conversations as deflected, so read it next to the resolved session rate, abandon rate and bot CSAT.

Leave a Reply

Your email address will not be published. Required fields are marked *

Book a free consultation