A dashboard reports that the AI handled every call, but the office still has unresolved enquiries. “Handled” has combined answered calls, captured messages, failed transfers, and completed bookings into one flattering number.
A HubSpot reporting view should distinguish stages of the workflow so the business can see what still needs action.
Choose the reporting grain first
Decide whether each row represents a call, appointment, enquiry, or customer. One caller can produce several interactions and one appointment can be changed repeatedly.
Do not count every CRM property update as a new result. Keep stable record and operation references so repeated deliveries do not inflate totals.
| Measure | Useful definition |
|---|---|
| Calls answered | Eligible incoming calls that reached the answering flow |
| Requests saved | Verified intake records created |
| Bookings confirmed | Scheduling outcomes meeting the approved definition |
| Handoffs accepted | Verified human acceptance, not only dial attempts |
| Follow-up outstanding | Owned work not yet resolved |
Use the account's supported reporting capabilities and data model rather than promising that every field automatically appears in every dashboard tier.
Show the denominator and exclusions
For a fictional pilot, 12 valid bookings from 20 booking-eligible calls means 60% of that defined set booked. It does not mean 60% of all callers became customers or that revenue grew by that amount.
Separate spam, tests, general questions, duplicate events, and requests that were never eligible for automatic booking. Record the time period and which timestamp each measure uses.
The cost-per-booking guide explains how a changing denominator can distort a cost calculation too.
Put unresolved work beside the outcome totals
Show pending CRM delivery, unassigned callbacks, transfer failures, and uncertain writes with a way for staff to open the underlying record.
A report that only shows completed cases can hide the queue that determines whether the pilot is actually useful. Report queue age and ownership where the workflow supports them.
Reconcile a small fixture before trusting the dashboard
Create fictional cases with one booking, one duplicate delivery, one pending request, one accepted handoff, and one failed callback. Write the expected counts before executing the flow.
Compare individual records with the aggregation. A visually polished chart can still group by the wrong date or count changed states more than once.
The outcome-property article covers the data definitions feeding this view.
Dashboard questions
Does an answered call prove a customer outcome?
No. Track intake, action, and follow-up separately.
Should all measures share one denominator?
Only when they describe the same eligible population and purpose.
What should a client review?
Review definitions, fixture counts, exclusions, and unresolved work during the pilot assessment.