Five months of the year are below average, and April lies
Databoat read 416 months of unadjusted retail sales and derived a seasonal index. December runs 15% above an average month — but the more useful finding is that April has swung 21 points between years, which makes it the one month you cannot compare.
- Source
- US Census Bureau advance retail sales, not seasonally adjusted, via FRED
- Period
- Series 1992–2026; index computed on the 11 complete years 2015–2025
- Published
- August 14, 2026
- Built with
- Databoat
15.19
points above an average month is where December sits. Five months of the year sit below the line, and a plain month-on-month comparison will mislead you in every one of them
December
- February-13.04
- January-11.29
- April-3
- September-2.56
- March-0.29
- June1
- October1.34
- July1.93
- August3.36
- November3.57
- May3.78
- December15.19
Horizontal: month of the year, 1 to 12. Vertical: seasonal index, where 100 is an average month for that year.
Hover a point
The year is not a slope, it is a trough and a spike. February sits at 87 and December at 115 — a business comparing December to February is looking at a 32% gap that has nothing to do with performance.
Computed across 2015–2025, so the shape reflects a decade rather than one year.
Basis — what this measures
- Sources
- US Census Bureau advance retail sales, retail trade and food services, not seasonally adjusted, retrieved as the full published series via FRED (RSXFSN).
- Period
- 416 monthly observations from January 1992 to July 2026. The index is computed on the 11 complete calendar years from 2015 to 2025.
- Grain
- One row per calendar month. The index is each month expressed against the mean month of its own year, then averaged across the eleven years, so 100 is an average month.
- Excludes
- Incomplete years at both ends of the series, so 2026 does not contribute. Deliberately uses the unadjusted series — a seasonally adjusted one has already had this pattern removed, which is the pattern being measured.
Movement — why the number is what it is
Left mark: the weakest reading that month has produced in eleven years. Right mark: the strongest. The bar is the swing.
- April81.35→102.01
- March91.94→103.19
- February80.54→90.03
- December110.6→120.04
- October98.46→106.17
- September95.2→102.01
- July99.4→105.98
- May99.58→106.02
- January85.09→90.86
- June97.95→103.28
- November100.4→105.63
- August101.19→105.07
April swings 21 points between years — more than twice any other month. Easter moves between March and April, so the same trading performance can produce wildly different Aprils. August, by contrast, moves less than 4 points and is the most predictable month in the year.
The year-to-year swing in each month’s index, in points. Shorter is more reliable.
- April20.66
- March11.25
- February9.49
- December9.44
- October7.71
- September6.81
- July6.58
- May6.44
- January5.77
- June5.33
- November5.23
- August3.88
Reliability is not the same as size. December is both the biggest month and one of the steadiest, so it compares well year on year. April is average-sized and the least trustworthy figure in the calendar.
Seasonal index, where 100 is an average month.
- December115.19
- May103.78
- November103.57
- August103.36
- July101.93
- October101.34
- June101
- March99.71
- September97.44
- April97
- January88.71
- February86.96
Five of twelve months sit below the line. A business seeing five down months a year may simply be having a normal year.
Distance from an average month, in index points, across all twelve.
Ten of the twelve months sit within five points of average. Almost the whole effect lives in three months — December up, January and February down.
Ties — the numbers reconcile
- Exact
The twelve indices average to exactly 100
Σ idx = 1,200.00 - Holds, 12 of 12
Every index falls inside its own observed range
lo ≤ idx ≤ hi - Verified
Computed from the unadjusted series, not the adjusted one
FRED series RSXFSN, not RSXFS - Not a benchmark
This is an economy-wide pattern, not a transferable one
Aggregate US retail; a single business will differ, sometimes completely
Method
The full published series of US retail sales, unadjusted for season — 416 monthly observations back to 1992 — read and reduced to a twelve-row seasonal index.
Each month is expressed against the mean month of its own year, which removes growth and leaves seasonality. Those ratios are then averaged across the eleven complete years from 2015 to 2025. An index of 100 is an average month; 115 means that month typically runs 15% above it.
The unadjusted series is used deliberately. A seasonally adjusted series has already had this pattern stripped out by the statistician — it is the right series for spotting turning points and the wrong one for measuring the season itself.
Why this is the report that needs history
Every other analysis on this site can be run on one month of data. This one cannot. It needs years, and it gets better as they accumulate — which makes it the clearest illustration of why a recurring report is worth more than a one-off analysis.
It also answers the question a client actually asks. "Sales are down on last month" is usually not a business problem, it is January. The useful question is whether this January was a good January, and nothing on a standard profit and loss can tell you that.
Two findings worth carrying into any monthly pack:
Five months of the year are below average. A business that reacts to every down month will spend nearly half the year reacting to the calendar.
April is the month that lies. Easter moves between March and April, and the swing that produces — 21 index points across eleven years — is larger than the entire December effect. Any April-on-April comparison needs the Easter date beside it, and almost no report carries one.
What this doesn't tell you
This is aggregate US retail. It is not a benchmark for an individual business and should not be used as one. A wedding venue, an accountancy practice and a ski shop have violently different calendars, and two businesses on the same street can be out of phase with each other.
What transfers is the method, not the numbers. The same calculation on a client's own three years of history produces their index, and that one is worth having.
Why we published this
The seasonal indices reconcile to exactly 1,200.00 across twelve months, which is the arithmetic guarantee that the calculation is a redistribution rather than an invention. Every derived column states its formula, and the one thing this report cannot support — being used as a benchmark — is recorded above as a failed tie rather than left implied.
Early access
Your files, the same treatment.
This report was built the way Databoat builds a monthly client pack — read the files, work out what is in them, publish. Tell us which report you rebuild every month.