Custom Reporting Variables for Historical Bot Dashboard

What it does

Lets you segment KPIs in the Omnichannel Historical Bot dashboard using custom variables defined in Copilot Studio — things like line of business, department, campaign type, or region. Also adds exit-pattern analysis at the question-node level so you can see where customers drop out of bot conversations and how that affects deflection and escalation rates.

Key facts

GA: February 9, 2026
Variables are defined in Copilot Studio and configured in Copilot Service admin center > Insights > Omnichannel Historical dashboard
Requires enabling the "Add advanced historical analytics for bot" checkbox in admin center
Segmentation works across KPIs: deflection rate, escalation rate, containment
Variables are captured during voice or chat interactions, prior to escalation or containment
No engineering effort required to add new dimensions — configured through the admin UI
Question-node exit analysis shows where customers abandon or are routed out at specific prompts

When to use / skip

Use this when the client has multiple lines of business or campaigns running through a shared bot and needs to slice bot performance by business segment rather than just by channel or workstream. Also useful for any client doing systematic bot optimisation — exit-pattern data at the question level is what you actually need to identify broken flows. Skip it if the bot is simple, low-traffic, or the client doesn't have the bandwidth to act on granular bot analytics.

Configuration decisions

Which business dimensions to expose as reporting variables (line of business, region, campaign, etc.)
Where in the conversation flow to set variable values — before branching points or at escalation
Whether to use the built-in dashboard customisation or export to Power BI for richer reporting
Who owns the Copilot Studio bot vs. who owns the analytics configuration — they may be different teams

Gotchas

Variables must be set in the Copilot Studio conversation flow before the escalation or containment point — if the variable is set after escalation it won't be captured in the bot dashboard
The "Add advanced historical analytics for bot" checkbox must be enabled in admin center — it's off by default and easy to miss during setup
Historical data only: these variables apply going forward from when they're configured, there's no retroactive population of historical records
If the client renames or removes a variable in Copilot Studio, the dashboard will have gaps; treat variable names as a data contract

Consultant notes

This is the feature that finally lets you answer "how is the bot performing for our enterprise clients vs. SMB?" — clients with segmented customer bases will find this genuinely useful.
The setup dependency chain is easy to get wrong: variable defined in Copilot Studio → bot published → checkbox enabled in admin center → dashboard customised. If the client does these out of order, variables won't appear in the reporting UI and the debugging trail is non-obvious.
Question-node exit analysis is the more technically interesting capability here. Pull the client into a review of their top exit nodes early in the project — it often reveals conversation design issues that should be fixed before go-live, not after.
Data volume matters: if the bot handles high traffic across many branches, the exit-pattern view can be noisy. Set expectations that this requires interpretation, not just dashboarding.

Worth revisiting after next release wave to check if real-time bot analytics gains similar segmentation support