Sales Nowcast

Adjusting a weekly traffic index when the holiday calendar moves

A desk calendar showing a full year's months, next to a cup of coffee and a small succulent plant.

Every retail analyst has had this moment: the weekly traffic number looks terrible, you start drafting the "footfall is rolling over" note, and then someone on the desk points out that Easter fell in March last year and April this year. The traffic didn't roll over. The calendar moved.

Fiscal retail calendars are built around a 52/53-week year, and most of the holidays that drive traffic don't sit on fixed calendar dates. That mismatch is the single biggest source of false signal in any week-over-week traffic series, and it's worth having a real process for it instead of eyeballing it each quarter.

Why the calendar keeps lying to you

Three things move independently of the retailer's fiscal week structure:

Easter floats between late March and late April depending on the lunar calendar, a swing of up to five weeks. A chain that indexes heavily on spring apparel or seasonal goods can show a traffic spike or trough that's purely a function of which fiscal week absorbed the pre-Easter shopping days.

Thanksgiving always falls on the fourth Thursday of November, but "fourth Thursday" can land anywhere from the 22nd to the 28th. That changes how many shopping days sit between Thanksgiving and month-end, and it changes which fiscal week catches Black Friday and Cyber Monday. A chain reporting a weak Thanksgiving week one year and a strong one the next may just be reporting a different number of pre-holiday shopping days.

Then there's the 53rd week. Retailers add an extra week roughly every five or six years to keep the fiscal calendar aligned with the actual calendar, and that one-time insertion pushes every subsequent fiscal week out of sync with the prior year by seven days until the next reset. If you're comping fiscal week 14 this year against the same fiscal week last year without checking where you are in that cycle, you're not comparing the same shopping days at all.

Building the adjustment

The fix isn't complicated once you set it up, but it does require daily granularity. Weekly-bucketed data forces you into the retailer's fiscal week boundaries, which is exactly the problem. Daily data lets you re-bucket.

Start by anchoring to the holiday itself rather than the fiscal week number. Instead of asking "how does fiscal week 47 compare to last year's fiscal week 47," ask "how do the fourteen days before Thanksgiving compare to the same fourteen days before Thanksgiving last year." That ties the comparison window to the holiday itself, the actual driver of the traffic pattern, instead of a fiscal week label that moves every year.

For Easter, the cleaner approach is a days-to-Easter index. Pull daily traffic for the eight weeks surrounding Easter in both years, line them up by distance from the holiday instead of by date, and compare. You'll usually find the "slowdown" or "acceleration" the fiscal-week comp implied mostly disappears once the days are lined up correctly.

For the 53rd-week years, the simplest workaround is to compare against the fiscal calendar two years back instead of one, since that's the last point where the week numbering lined up the same way, or to build a calendar-day crosswalk that maps each fiscal week to its actual date range and re-bucket both years on that.

None of this is exotic math. It's bookkeeping. But it's bookkeeping that most app-based panel data doesn't let you do, because those feeds are usually delivered pre-aggregated into weekly or monthly buckets by the provider, with no way to pull the underlying daily series and redo the cut yourself.

That's the practical reason to want a daily feed in the first place, not a weekly one. A daily vehicle-count series for the chain's lots, seasonally adjusted but still delivered at daily grain, is what lets you build your own holiday-adjusted comp instead of inheriting someone else's week buckets. That's the shape of feed Sales Nowcast is built around.

If a shifting Easter or a 53rd-week reset is about to muddy your comp, it's worth checking whether a daily, chain-level traffic series would settle the question before the print does.

Get started

← Back to the blog