Peak season isn't a surprise. You know Q4 is coming. You know your supplier's lead times extend by three to six weeks during their own production peak. You know that an inventory position that would be adequate in July will be dangerously thin by the second week of November. The knowledge is there — the difficulty is translating it into specific, defensible order quantities per SKU, placed at the right time, calibrated for what you actually expect to sell rather than what you sold last year plus a gut percentage.
Seasonal demand forecasting is the structured version of this translation. It's not a single algorithm — it's a collection of practices that together give you a more reliable inventory position heading into a peak period than either "order what you ordered last year" or "add 20% buffer to safety stock and hope."
What makes seasonal demand different from baseline demand
Seasonal demand has two distinct characteristics that complicate standard forecasting approaches.
First, there's a level shift — demand spikes to a multiple of baseline. For a gift-oriented consumer product, that multiple might be 4×–6× in the six weeks around December holidays. For a back-to-school stationery brand, it might be a 3× spike in late July through August. The direction and timing of the spike are predictable; the exact magnitude is not, and the magnitude varies by SKU in ways that don't always move proportionally with category.
Second, the forecasting window is compressed. In baseline season, a forecast error that leads to a reorder two weeks late has a cost. During peak season, a reorder that's two weeks late can mean stocking out mid-Black Friday weekend — because the demand rate is high enough that two weeks' delay at peak velocity has a far larger unit impact than the same delay at baseline velocity.
Standard exponential smoothing models (including Holt-Winters triple exponential smoothing, which handles seasonality) can capture seasonal patterns — but they do so by learning from historical seasonal patterns. If your pattern changes, if a new product launches, or if a promotional event creates a demand event on top of the seasonal pattern, the model's seasonal component may not update fast enough to give you the right signal before you need to place your orders.
The lead time math during peak
Most supply chain planning frameworks recommend placing your peak season purchase orders well in advance of the selling window. The standard guidance is to work backward from your target in-stock date to your order date using your supplier's lead time — plus a buffer for lead time variability during their own busy period.
Here's the arithmetic that trips up most brands: peak season doesn't just compress your selling window. It also extends your supplier's lead time. A factory in Vietnam or Bangladesh with a 45-day baseline lead time often runs 65–75 days during Q3–Q4 when global apparel orders are peaking. If you're planning orders using the baseline lead time, you'll place orders too late for on-time delivery.
The correct calculation: your target in-stock date for peak selling minus your peak-season lead time (not baseline) minus your safety buffer for lead time variability. For a brand targeting November 1 in-stock, with a 70-day peak-season lead time, orders need to be placed by mid-August at the latest. For brands that don't account for the lead time extension, mid-September feels like plenty of time — and it isn't.
Compounding this: your MOQ (minimum order quantity) per SKU may force you into larger orders than your demand forecast suggests. If your supplier requires a minimum of 500 units per colorway and your forecast says 340 for a slow colorway, you're either leaving potential stockouts on the table for the fast colorways (by consolidating into fewer options) or accepting some overstock risk on the slow ones. This is a real trade-off that no algorithm resolves automatically — but having a demand forecast per colorway gives you the numbers to make the trade-off deliberately rather than arbitrarily.
Using prior peak data correctly
For brands with at least one full prior peak season of data, the prior year's actuals are the starting point for seasonal forecasting. But "use last year's numbers" is not a seasonal forecast — it's an extrapolation with no model.
A proper seasonal forecast uses prior peak data in combination with current-year trailing demand to estimate the seasonal multiplier. The seasonal index for a given week or month is: this period's historical sales divided by the full-year average weekly or monthly sales. That index, applied to your current-year baseline demand, gives a season-adjusted demand estimate that accounts for both the historical seasonal pattern and any year-over-year demand trend.
The limitation of prior-year data is also worth naming directly. If you had a supply constraint last year — a stockout during peak — your actual sales data for that period understates true demand. Your 2023 November sales may show 800 units sold when true demand was 1,100; you just ran out at 800. Using 800 as your baseline for 2024 peak forecasting will reproduce the same stockout. A good seasonal forecast system allows you to annotate or adjust for known constraint periods in the historical data.
The pre-peak inventory position review
Regardless of your forecasting method, a pre-peak inventory position review is a critical planning step. This is a structured review, typically six to eight weeks before the peak selling period opens, that asks: for each SKU in your peak-critical set, what is the current on-hand quantity, what is the expected inbound (purchase orders placed and in transit), and what does the demand forecast project we'll sell from now through peak?
The gap between expected available inventory (on-hand + inbound) and forecasted demand is your exposure. SKUs with positive gaps are in good shape. SKUs with negative gaps — where you're forecasting more demand than you have inventory to cover — are candidates for emergency reorders, if the lead time window still allows, or for demand management (deprioritizing ad spend, adjusting promotion timing) if it doesn't.
A mid-market home goods brand managing around 1,200 active SKUs did this review exercise for Q4 2024 in mid-September. They identified 34 SKUs where the demand forecast for October–December exceeded their available inventory plus all confirmed inbound orders. Of those 34, 22 still had time to place supplemental orders with domestic or nearshore suppliers (3–4 week lead times) to close the gap. The remaining 12 were committed to overseas factories with 10+ week lead times; for those, they deprioritized paid media spend to avoid accelerating demand on SKUs they couldn't fulfill. Without the structured pre-peak review, none of those decisions would have been made proactively — they would have become stockout events discovered in real time during November.
Safety stock calibration for peak season
Your baseline safety stock formula is not right for peak season. Safety stock is designed to cover demand uncertainty during the lead time window. At peak season, both parameters in that formula change: demand uncertainty is higher (demand spikes are harder to forecast precisely than baseline demand), and the lead time window may be longer if you're working with stretched supplier capacity.
A practical approach is to calculate a separate peak-season safety stock for your top-30 SKUs by revenue contribution, using peak-season demand variability (the standard deviation of weekly demand during the comparable peak period from prior years) rather than full-year demand variability. This typically results in a meaningfully higher safety stock target for high-velocity seasonal SKUs — which is correct, because the cost of a stockout during peak is much higher than during baseline.
We're not saying you should carry your peak-season safety stock year-round — that would permanently inflate carrying costs for the benefit of a 6–8 week window. We're saying peak-season safety stock should be sized for peak-season conditions, built into your Q3 order plan, and deliberately drawn down post-peak to return to baseline levels rather than leaving excess inventory to age on the shelf.
Post-peak markdown planning
Seasonal forecasting isn't only about avoiding stockouts. The overstock side of the ledger matters equally. Carrying unsold inventory from one peak season into the next ties up working capital and usually requires markdowns that destroy margin — particularly for fashion or seasonal goods where year-old colorways have reduced appeal.
A demand forecast going into peak season should come with a planned sell-through scenario: by what date do you expect to hit zero, or near-zero, on each peak SKU? If your forecast suggests you'll still have 20% of your starting inventory remaining six weeks after peak, that's a signal to plan markdowns early rather than waiting until the end of the season. Proactive markdown earlier in the tail sells at a higher price point and clears faster than a distressed end-of-season clearance.
Getting seasonal demand forecasting right is one of the highest-value inventory decisions growing brands make. Stockorlo incorporates seasonal pattern detection into SKU-level forecasting and surfaces pre-peak coverage gaps automatically. See how the forecasting engine handles seasonality, or explore our seasonal planning resources further. For brands preparing for their first structured peak season plan, reach out to the team for a catalog assessment.