Last Year's Week Isn't This Year's Week

The calendar doesn't repeat, so your comparison periods don't either

The most common year-over-year comparison is by date: November 1–30 this year against November 1–30 last year. It looks fair because the labels match. It quietly assumes that both Novembers contain the same mix of days: the same number of Saturdays, the same holiday weekend, the same paydays. They usually don't.

A year is 52 weeks plus one day, and plus two days after a leap day. So every date falls one weekday later than it did the year before, and two later after February 29. Over a few years that drift produces four kinds of calendar artifact. A holiday moves to a different week or month. A month gains or loses a weekend day. A peak event crosses a month boundary. And a 52-week reporting year eventually has to absorb a 53rd week.

On a date-aligned report each one looks like demand, and a calendar lift read as growth becomes a bigger purchase order.

Four calendar artifacts that look like demand

Black Friday and Cyber Monday. Thanksgiving is the fourth Thursday of November, so Black Friday always lands between November 23 and 29. In 2025 it fell on November 28 and Cyber Monday spilled into December 1. In 2026 Black Friday is November 27 and Cyber Monday is November 30, so the whole weekend sits inside November. Compare November 2026 with November 2025 by date and this November gets an extra peak day it didn't have last year.

Easter. Easter can fall anywhere from March 22 to April 25. It was April 20 in 2025, is April 5 in 2026, and will be March 28 in 2027. In 2027 the run-up to Easter moves into March. A March-vs-March comparison will show a surge and the April comparison a slump, though neither reflects a change in what customers wanted.

The fifth Saturday. A month has four or five of each weekday, depending on where it starts. November 2025 had five Saturdays and November 2026 has four. October goes the other way: four in 2025, five in 2026.

The 53rd week. Retailers that report on 52-week years drop a leftover day each year. Every five to six years those leftover days add up to a full extra week. In that year, a total that isn't adjusted contains a week more of selling than the year it is compared against.

Dates are computed from calendar rules; the 53rd-week cadence is from the NRF 4-5-4 calendar.

Align the periods before you read the trend

Compare by week, not by date. Go back 364 days instead of one calendar year. That's exactly 52 weeks, so Monday lines up with Monday and every weekend lines up with a weekend. It also tends to keep events together. Friday, November 27, 2026 minus 364 days is Friday, November 28, 2025, which is the previous Black Friday.

Compare event to event. For a sale weekend or a holiday, line up the event weeks themselves, whatever dates they fell on. Treat “the week of Black Friday” as the unit, not “the last week of November.”

Widen the window. If an event can move between two adjacent months, compare both months together. November plus December takes in Cyber Monday wherever it lands. March plus April takes in Easter. Per-month detail is lost, but the total is reliable.

Restate the long year. In a 53-week year, set the extra week aside or compare a shifted 52 weeks against the prior year. The National Retail Federation restates the 53-week year in its own 4-5-4 calendar for this reason.

Calendar artifact or a genuine move?

Once you've seen a difference, ask one question before anything else: does the calendar explain it? Rerun the comparison aligned, with matching weekdays, matching events and a window wide enough to hold both. If the difference disappears, the calendar caused it.

Artifacts behave in a recognizable way. They push units across the edges of a window, so one period looks better and its neighbor looks worse by about the same amount. Over a wider span they cancel out. A real shift doesn't cancel. It stays when the weekdays are matched, and it stays when November and December are read as one block.

Unlike trend-or-noise checks by variant, store or week, this only asks whether your chosen dates created the change; settle that first.

Doing the aligned comparison in Assisty

The AI Report Builder lets you ask for sales reports “by product, variant, category, vendor, location, customer, or over time, from the last 7 days to year-over-year.” You ask in plain English, and Atlas queries your synced store data and answers with a live table. Assisty doesn't align the calendar for you. Ask for sales over time and refine the time range by talking until both periods cover the same weeks.

Say “Show the last 90 days,” and Atlas changes only what you asked for and keeps the rest of the report intact. When the comparison is right, save it as a Custom Report.

If the aligned comparison still shows a real move, test it in the forecast rather than letting one raw number set it. Ask “What if sales trend up 20%?” and the forecast and reorder needs update in front of you. Ask why demand is changing and Atlas explains the drivers.