Calendar audit: reading last month before booking the next

Next month is already half booked, and the invitations arriving now are for the half that is left. A calendar audit is the habit of reading the month that just ended before agreeing to the shape of the one that has not started. It takes an hour, and the output is not a feeling about being busy. It is a small number of decisions that can be turned into settings.

Most audits fail in the same way. They produce a chart, the chart confirms that there are a lot of meetings, and nothing changes because the chart does not say which meeting to stop attending. The version worth running answers three narrower questions: where the booked hours went, which of them were never really a choice, and what work happened that never appeared on the calendar at all.

What the calendar can and cannot tell you

A calendar is a record of what was booked, not of what happened. That distinction decides how much weight any audit finding can carry.

Booked and attended are different. An event that was accepted and then skipped still sits in the record at full length. A meeting that ran twenty minutes over ended at a time the calendar does not know about. Work that filled two hours between meetings, because nobody booked it, is invisible.

This is not a reason to skip the audit. It is a reason to treat the calendar as evidence about commitments rather than evidence about effort. Commitments are what the next month is made of, and commitments are the thing an audit can actually change.

The useful frame is to read the month twice. The first pass counts what was booked and by whom. The second pass looks for the gaps: the blocks where nothing was scheduled and something clearly happened anyway. The second pass is where the surprises are, and it cannot be automated, because the evidence is missing by definition.

The number the calendar already computes

Anyone on a work or school Google account has a built-in answer waiting in the left panel. Time Insights breaks the visible date range into categories and is reached by clicking More insights; if the option is absent, an administrator has turned it off for the organization. It can only be viewed on a computer, and it reads the primary calendar, which means anyone managing someone else's calendar needs the permission to manage sharing on that calendar to see their numbers.

The categories are worth knowing before reading the chart. Focus time is planned focus time with no conflicting meeting. The 1:1 category is meetings with exactly one other guest, and the next one up covers meetings with at least two other guests. Need to respond collects the meetings that have not been accepted or declined, with Maybe answers excluded. Remaining time is the time inside working hours with no meeting and no out-of-office entry, and it only appears at all if working hours have been set.

The averages are calculated over a window that changes with the view, which catches people out. In Day, Week, Schedule or 5 days view, the average comes from the previous three weeks. In Month view, it comes from the previous three months. Reading one number in one view and a different number in another view and concluding that the calendar is inconsistent is a misreading of the same data over two different windows.

Overlapping bookings are resolved rather than double counted. Google states that time in meetings counts once for any period, so responding Yes to two one-hour meetings at the same time contributes one hour. Where a 1:1 overlaps a group meeting, the 1:1 is the one counted. Several overlapping 1:1s are all counted.

What Time Insights quietly leaves out

The exclusions matter more than the categories, because they are where an audit built on this one panel goes wrong.

Counted toward time in meetings	Not counted
Google Calendar meetings answered Yes	Anything not answered Yes, including Maybe
Events with at least one other guest	Events where nobody else was invited
Events marked Busy	Events marked Free
Events shorter than eight hours	Events of eight hours or longer

Read that column on the right as a list of things the built-in number treats as empty time. Solo work blocks do not count, which is the correct behavior for a meeting metric and the wrong behavior if the question is where the week went. A recurring meeting that was never formally accepted does not count even if it was attended every week. A day-long workshop marked as a single block does not count at all, because it crosses the eight hour boundary.

The per-person view has its own rules. The five people with the most shared meeting time appear by default and up to ten contacts can be pinned, but shared time is only counted for meetings with fifteen or fewer other people, and it is not counted when the other person declined or when the Yes was missing on either side. A weekly all-hands with forty people contributes nothing to anyone's shared time, which is either a sensible exclusion or a large blind spot depending on what the audit is for.

None of this makes the panel useless. It makes it a meeting meter rather than a time meter, and it means the categories it reports need to be read next to the calendar grid rather than instead of it.

Counting from an export when there is no panel

Personal Google accounts do not have Time Insights, and any audit that needs to combine a work calendar with a personal one has to leave the panel behind anyway. The alternative is to count from the underlying files, which every calendar can produce.

In Google Calendar, Settings has an Import and Export section, and exporting produces a .zip file containing one .ics file for each calendar. The files have to be taken out of the archive before anything can be done with them. Apple's Calendar on a Mac exports the same way, one .ics file per calendar through File and then Export, with a second option that writes every calendar into a single .icbu archive. The archive is for backup rather than analysis, and Apple's own warning is worth respecting: importing an .icbu file replaces all current calendar information and data.

Inside the file, the format is the one defined by RFC 5545. Each event sits between a BEGIN:VEVENT line and an END:VEVENT line, with its start, end, summary and organizer as separate fields. That is enough structure to answer the questions a chart will not: total hours by title pattern, how many events were recurring instances rather than one-offs, how many carried no other attendee, and which organizer's name appears most often.

Counting by hand is slower than reading a panel and answers more. The one habit that saves the most time later is to export before making changes, so that next month's audit has a fixed reference rather than a calendar that has already been rearranged.

The gaps are not meetings

The second pass, the one looking for what never got booked, usually surfaces the same three things.

The first is travel. A meeting across town is not thirty minutes, and if the calendar says thirty minutes then the surrounding hour is lying. Apple's Calendar can attach travel time to an event and compute a departure alert from the previous event's location, and its documented limits are instructive: no departure notification is sent for a journey longer than three hours, and changing the location of an event does not automatically update the travel time that was already set. Both limits point the same way, which is that travel has to be visible as a block rather than implied. The reasoning behind treating movement as part of the event is set out in Travel time.

The second is preparation. Meetings that require reading something first tend to show a cost in the hour before them that appears nowhere. If the audit finds a recurring meeting that reliably eats the preceding half hour, that half hour belongs on the calendar under its own name.

The third is the recovery time after a run of calls, which is real and never booked. It is the item most often dismissed as not a real commitment, and it is also the reason a day with six meetings produces less than a day with three.

Turn findings into settings

An audit that ends in intentions decays within two weeks. The findings that survive are the ones moved into a setting that keeps working without attention.

Working hours are the first of them, on a work or school account. Setting them makes the Remaining time category meaningful and warns other people when they try to book outside them. They can also be split, which matters more than it sounds: a day can carry two separate blocks, so a genuine gap in the middle of the afternoon stops being bookable rather than being defended by hand each week.

Focus time is the counterpart, and it is a booking rather than an intention. The same panel that reports the categories offers a control to schedule focus time, and a focus block with no conflicting meeting is what the Focus time category is counting. An audit finding that says there were four hours of uninterrupted work in a month is only actionable if those hours can be booked in the same way a meeting is booked, which is the point of putting them on the calendar rather than on a list.

Out of office does something stronger. Marking a period as out of office automatically declines the invitations that arrive for it, with a decline message that can be edited. Anyone whose audit shows meetings landing on travel days or on a fixed writing day has a setting waiting.

Booking pages carry two limits that a recurring audit finding maps onto directly. Buffer time inserts a gap between consecutive bookings, and a daily booking cap limits how many appointments a single day can take. The booking window controls how far ahead and how late someone can take a slot. Three numbers, decided once, replace a recurring argument with yourself about whether to accept an eleventh meeting.

The last setting is where the calendar is read from. An audit that requires opening a browser tab and clicking into a panel gets run once. Keeping the month, the day and the list in one window on the desktop is what makes the second pass cheap enough to repeat, and Features covers which views are worth switching between when reading a month backwards rather than forwards.

What to change first

Pick the single recurring meeting with the worst ratio of hours to outcomes and change its length, its frequency or its guest list this week, before booking anything new. Then set working hours and one buffer number so that next month's audit has fewer findings to make. Caltimate keeps the month, week and list views a keystroke apart, which is what makes reading last month a habit rather than a project.