Reading the chart
Scheduled time is only part of the working week.
The demo adds each selected event duration and uses that sum as the denominator for category percentages. It does not measure attendance, unscheduled work or whether a meeting was useful. Overlapping events can also count the same clock time more than once.
Suppose a fictional 40-hour week contains 14 hours of meeting entries and six hours of other calendar entries. Meetings are 70% of the 20 recorded hours. That does not mean meetings occupied 70% of the working week, or that the unrecorded hours were available for focused work.
What to inspect
Three limits behind the summary
Category: the event title supplies keywords. A vague title can produce a category that does not match the work.
Duration: start and end times describe a scheduled interval, not verified attendance or effort.
Coverage: missing events, overlaps, breaks and working hours need explicit treatment before interpreting capacity.
Implementation choices
What the source actually does
These details are visible in the artifact implementation. They explain how to read the demo without treating its labels as findings about a team.
Categories
The first keyword match wins
The code lowercases the event title and checks category keyword lists in a fixed order. Meetings are checked before admin, follow-ups and deep work. If no keyword matches, the event falls back to Meetings & calls.
For example, CRM review matches the meeting keyword review before the admin keyword CRM. CRM update instead reaches the admin category. The labels are deterministic rules, not AI judgments.
Time
Keep the denominator visible
The overview reports the sum of selected event durations, meeting time and admin time. Category shares refer to those recorded durations. A focus-labelled appointment is evidence that a block was scheduled, not that uninterrupted work occurred.
A production analysis would need a declared work schedule and rules for overlaps, all-day events, cancellations and incomplete coverage. It should not equate an empty calendar with spare capacity.
Period comparison
Compare like-for-like windows
The demo can compare two date ranges within its fixed sample. A shorter period can show fewer meeting hours simply because it includes fewer days. Check the number and kind of working days before interpreting a difference.
There is no universal minimum of one month that proves a pattern. Select enough representative periods for the question, explain unusual weeks and avoid extrapolating a single sample into a long-term trend.
Recommendations
Read the saving as an assumption
The recommendation code multiplies admin duration by 0.8 and follow-up duration by 0.7 for suggested savings. Those are fixed illustration factors, not results measured after an automation was installed.
It also counts adjacent events with gaps of ten minutes or less as back-to-back. Its display wording says no gap, but the implementation includes short gaps and overlaps. Inspect the schedule before assuming a sequence of calls had no recovery time.
Illustrative title check
Similar titles can land in different categories.
These titles demonstrate the current keyword rules. They are illustrative inputs, not customer appointments or a measured classification-accuracy test.
| Example title | Demo category | Why |
|---|---|---|
| CRM review | Meetings & calls | review matches before CRM |
| CRM update | Admin & data entry | CRM matches the admin list |
| Research | Deep work | research is a deep-work keyword |
| Unlabelled session | Meetings & calls | No match; the default category applies |
The operational implication
A calendar category cannot tell you whether a meeting earned its time.
Use the view to select an event for discussion. Ask what decision the meeting supports, what work would remain if it changed and who depends on the result. A chart alone cannot establish that a meeting should be cancelled or that admin can safely be automated.
This build is that method in practice: a clickable diagnostic that changes the brief.
How to test a workflow with a live artifact ->