A standard reorder formula is built on the concept of average demand. It looks at historical sales over some baseline period, calculates a daily or weekly average, and uses that average to set a reorder trigger and order quantity. For products with reasonably stable, year-round demand, this works well enough. For seasonal products -- those with a concentrated sales window, a ramp-up period, and a ramp-down tail -- the average is actively misleading.
The problem with averaging seasonal data is that the average smooths out the exact variation that matters most for inventory decisions. A product that sells 2 units per week for 46 weeks of the year and 40 units per week for 6 weeks in the summer has an annual average weekly velocity of about 6.5 units. If you set your reorder formula around that average, you'll be overcommitted for 46 weeks and catastrophically undersupplied for 6. The average tells you nothing useful about either phase.
What Makes Seasonal SKUs Different
Seasonal products in retail come in several forms, and they have meaningfully different forecasting requirements:
True seasonal products are only carried during a specific window -- holiday gift items, summer grilling products, back-to-school supplies. These products don't have a year-round baseline. Their entire demand lifecycle is the season itself, and the forecasting challenge is estimating total seasonal sell-through and setting an initial order quantity that captures the peak without leaving you with unsaleable excess after the window closes.
Year-round products with seasonal peaks are always on your shelf but sell at significantly different rates depending on the time of year. Oat milk in a college neighborhood sells faster September through November and January through May when school is in session. Hot cocoa mix sells 3x its annual average in November and December. These products have a baseline that exists all year, but the peak velocity can be dramatically higher and requires a temporarily elevated reorder trigger.
Event-driven products have demand concentrated around specific dates rather than seasons -- products tied to Valentine's Day, the Super Bowl, local festivals. The sales window may be only a week or two, and the demand spike may be both larger and shorter than a typical seasonal peak.
A reorder formula that averages across these different patterns produces a trigger that's wrong for most of the year. The seasonal adjustment needs to be built into the formula itself, not bolted on manually after the fact.
Why Manual Calendar Overrides Break Down
The typical workaround for seasonal forecasting in spreadsheet-based buying is the manual calendar override: the buyer adds a note to the spreadsheet in early November to "remember to order more hot cocoa mix," or they flag a few weeks in the buying calendar for summer products in May. This approach works when the buyer remembers to do it, when the timing of the seasonal shift is predictable, and when the magnitude of the shift is roughly what they expected.
It fails in several common ways. First, buyers forget. A manual process dependent on buyer memory will miss overrides, especially for products that are seasonal but not obviously so. Second, the timing and magnitude of seasonal demand shifts vary year to year. A calendar override set for "early November" may be a week late in a year where the holiday shopping pattern starts earlier. Third, calendar overrides are chain-wide -- they don't account for the fact that Location A in a student neighborhood may see a back-to-school demand shift in August while Location B in a retirement community sees no such shift at all.
The scale problem compounds quickly. A buyer managing 8 locations with 50 products that have meaningful seasonal variation needs to maintain and track roughly 400 seasonal adjustment decisions every year. That's not manageable as a calendar-override process without significant errors and omissions.
What Demand Sizing Does Differently for Seasonal Products
A demand sizing approach that handles seasonality well treats recent velocity trend as the primary signal, not the historical average. For a seasonal product entering its peak, sell-through velocity will be rising week over week -- the model detects that upward trend and adjusts the forecast accordingly, increasing the reorder trigger and order quantity in response to observed demand rather than a calendar flag.
This has two important properties. First, it adapts to the actual timing of the seasonal shift, which varies year to year, rather than assuming the season always starts on the same date. Second, it's location-specific: if Location A's seasonal velocity is rising while Location B's is flat, the model generates different orders for each location rather than applying a chain-wide seasonal multiplier that gets the magnitude wrong for both.
For true seasonal products with no year-round baseline, the initial order quantity has to be set manually -- there's no ongoing sell-through history to extrapolate from before the season starts. But once the season opens and real sell-through data starts accumulating, even a few weeks of data is enough to calibrate the in-season reorder trigger and manage through the peak more accurately than any pre-set formula.
End-of-Season Inventory Risk
The hardest part of seasonal SKU management isn't the ramp-up -- it's the wind-down. Most seasonal products are committed well in advance (because suppliers require early orders for seasonal items), which means you're making the bulk of your purchasing decisions based on forecast uncertainty. Getting the sell-in quantity wrong by 15-20% on a seasonal product isn't unusual, and the consequences are asymmetric: undersell means excess inventory that you have to mark down or liquidate, with margin impact. Oversell means missed revenue but often no lasting harm.
A demand sizing model that tracks in-season sell-through rate can flag early when a seasonal product is tracking below its initial forecast -- giving the buyer enough lead time to reduce future orders before the problem becomes an end-of-season inventory problem. The signal is in the weekly velocity relative to the initial forecast: if a product was projected to sell 30 units per week at peak and it's actually selling 18, and the season is 4 weeks in, the model should be reducing the remaining order quantity recommendation accordingly.
For perishable or highly seasonal categories, this kind of in-season course correction is the difference between a manageable markdown and a significant inventory write-off at season end.
