Gauge's Monday report gave last week 169 hours. The comparison week got 168. Both were complete weeks.
That detail stopped me before I reached the traffic numbers. AI weekly reports sound like an easy job to hand off: pull the numbers, compare them, tell me what changed. But “last week” is an instruction with several decisions hiding inside it.
If you run a service business, you have a version of this problem. The booking system uses one clock. The website report uses another. Monday's dashboard quietly includes Monday morning, while your office manager is counting through Sunday night. Everyone has a spreadsheet. Nobody quite has the same week.
Our specific wrinkle was the clock change. Gauge's November 2 measurement report used complete Monday–Sunday weeks in America/Chicago. October 26–November 1 ran from 05:00 UTC on the opening Monday to 06:00 UTC on the next. The previous week began and ended at 05:00 UTC.
NIST's daylight saving time rules explain why: daylight saving time ended November 1, repeating an hour. A seven-day local calendar week was not a fixed 168-hour interval.
Nothing broke in that report. The useful work was making the boundary visible before anyone interpreted the comparison.
AI weekly reports need a calendar, not “the last seven days”
Gauge recorded the dates, time zone and exact boundaries together. The report also said it had not normalized the totals by hours. That lets me read the comparison as calendar-week totals, without pretending it measures activity per hour.
That distinction matters. A shop owner might want completed local weeks because that matches the operating rhythm. A round-the-clock service might also want an hourly rate. Neither choice should be made silently by the person—or AI worker—preparing the report.
Here is an illustrative brief for a business's weekly booking report, not a customer workflow we have tested:
Every Monday, prepare a booking summary for the previous complete Monday–Sunday week in our business time zone. Compare it with the complete week before that. State both date ranges and the time zone at the top. Use our approved booking export, exclude marked test records, and keep cancellations separate. Flag any difference in hours or opening days; do not silently adjust the totals. If the export is incomplete, label the affected comparison unavailable. Send the summary to me for review; do not change bookings or contact customers.
Replace the time zone and source before using it. Start with one report and compare its boundaries with the report you already trust.
Our earlier marketing measurement story separated visits from actual inquiries. This is the step before that: make sure the two columns even describe the periods you think they do.
If you want help defining a recurring reporting job for an AI team, tell us about your business. Chief offers assisted setup, starting with a review of your use case.
The extra hour wasn't a business result. Noticing it was part of doing the job.