Last Year’s Peak Is Already in Your POS
Last year’s summer peak is already in your POS. You sold 340 units of your top performing sandal between June 8 and August 31. That is a daily velocity of 3.7 units. This year’s order starts with that number, not a guess.
That sounds obvious, but many seasonal buys still begin with a conversation around growth targets, supplier minimums, or someone’s memory of a busy Saturday. The useful starting point is more concrete. What did each product sell, in each location, during the same weeks last year? How quickly did stock move before the peak? When did demand slow down? Which products sold together?
Seasonal planning is not about reproducing last year perfectly. Weather changes. Prices change. Store coverage changes. The purpose of historical data is to give you a defensible base, then make the assumptions around that base visible. If you expect demand to rise by 10 percent, you should be able to show where that number came from and how it affects the order.
Build the Base Forecast from Actual Velocity
Start at the smallest useful level: SKU, location, and week. Monthly totals hide the shape of seasonal demand. A product that sold 120 units in June may have sold 20 in the first week and 35 in each of the following three weeks. Those are different replenishment problems.
For each seasonal product, pull at least the following from last year:
- Units sold by week and location.
- Opening stock, receipts, and closing stock.
- Days or weeks when the product was out of stock.
- Promotional periods and price changes.
- Supplier lead time and minimum order quantity.
Then calculate observed sales velocity. If a store sold 84 units over six fully stocked weeks, its average weekly velocity was 14 units. If it was out of stock for one of those weeks, do not divide by seven weeks. That would make demand look weaker than it was. Either exclude the stockout week or estimate the missed sales using nearby weeks and document the adjustment.
Do this separately for every store or channel before adding the results together. A location near the beach may sell twice as many sun hats as a city center shop, while the city shop may sell more travel bags. A single network average can cause you to send the right total quantity to the wrong places.
Once you have a clean weekly base, apply the business changes expected this year. A store opening, a larger selling area, a planned price reduction, or a new online channel can all change the starting point. Keep the adjustment separate from the historical rate. For example:
- Last year’s comparable weekly velocity: 14 units.
- Expected store traffic increase: 8 percent.
- Adjusted base velocity: 15.1 units per week.
Do not quietly mix an assumption into the historical number. When the season is over, you will want to know whether the forecast was wrong because the data was poor or because the assumption did not happen.
Apply a Seasonal Index to the Shape of Demand
The annual average is useful as a reference, but it is not a seasonal forecast. If a product sells 520 units across 52 weeks, its annual average is 10 units per week. That does not mean you should plan for 10 units during every summer week.
Calculate the seasonal index by comparing demand in a specific period with the annual average for the same product. If the four peak weeks averaged 14 units per week and the annual average was 10, the seasonal index is 1.4. Your peak forecast is therefore:
10 units x 1.4 seasonal index = 14 units per week.
You can use a separate index for each week or group of weeks. A simple curve might look like this:
- Four weeks before peak: 0.9.
- Three weeks before peak: 1.1.
- Two weeks before peak: 1.3.
- Peak week: 1.4.
- Following week: 1.2.
- Two weeks after peak: 0.8.
This shape matters because inventory arrives before demand does. If your supplier takes three weeks to deliver, the order placed in early May must cover the projected demand through the first part of June, plus the stock you need while waiting for the next receipt.
Use comparable weeks where possible. Easter, school holidays, public events, and weather can move the calendar. Comparing a holiday week with an ordinary week can produce a misleading index. If the selling dates do not line up, compare the commercial context rather than the calendar date alone.
Seasonal Buffer Stock Needs Its Own Adjustment
Standard safety stock formulas are often based on average demand and average variability. That is reasonable for a stable product. It is less useful when demand is climbing quickly and a stockout during the peak means lost sales that cannot be recovered later.
A common starting point is:
Safety stock = service factor x demand variability x the square root of lead time.
The exact service factor depends on the availability target you choose. The important point is that demand variability should reflect the season you are planning for, not the quiet months that make the average look comfortable.
Suppose a beach towel sold between 8 and 12 units per week during the spring, but between 18 and 31 units per week during the summer peak. Using spring variation to calculate summer safety stock will leave you exposed. Use the peak period’s deviation, or apply a seasonal variability multiplier based on prior seasons.
You also need to account for the consequences of a stockout. A slow moving accessory can wait for the next delivery. A core product with a short selling window cannot. For a peak item, a practical buffer can include:
- Safety stock based on peak demand variation.
- Coverage for supplier lead time.
- Extra units for known delivery delays or receiving constraints.
- A controlled allowance for forecast error.
Be careful with the word buffer. It should be a calculated quantity, not a general feeling that the order seems too small. If you add 15 percent because the team feels nervous, record that as a 15 percent forecast risk adjustment and review whether it was justified after the season.
Scenario: A Four Location Beach Accessories Retailer
Consider a retailer with four locations preparing for summer 2026 using its summer 2025 sales. The business sells beach accessories, travel goods, and outdoor leisure products. Three product families account for most seasonal demand: sandals, sun hats, and insulated bottles.
In 2025, the comparable summer period produced these weekly patterns across all four locations:
- Sandals rose from 42 units four weeks before peak to 68 units at peak, then fell to 36 units two weeks later.
- Sun hats rose from 28 units to 47 units at peak, with the strongest demand in the two weeks around the local school holidays.
- Insulated bottles rose from 35 units to 52 units at peak, but online demand remained high for three weeks after store sales started to decline.
The retailer’s supplier lead time is three weeks for sandals and two weeks for hats and bottles. The stores also have different profiles. The two coastal locations generated 62 percent of sandal sales, while the two inland stores sold more bottles. Planning only at total company level would miss that difference.
For sandals, the retailer calculates a peak index of 1.42 against the annual average. It expects comparable store traffic to increase by 6 percent in 2026. The peak forecast becomes:
Annual average of 48 units per week x 1.42 x 1.06 = 72 units per week.
The retailer then applies a peak safety buffer of 12 units, based on the variation observed in the 2025 peak and the three week lead time. Its planned peak position is 84 units per week across the network. That does not mean ordering 84 units every week. It means receipts and opening inventory should support the expected 72 units while leaving room for a hotter week.
Allocation follows the historical location mix, then gets reviewed against current stock and local demand. The coastal stores receive more sandals, while inland stores receive a larger share of bottles. The online channel is planned separately because it continued selling bottles after store demand softened.
This kind of plan is not complicated, but it is more useful than a single seasonal growth percentage. It shows the demand curve, the location mix, the lead time, and the reason for the buffer.
Know When Last Year Is a Poor Baseline
Historical demand is valuable only when the product and selling conditions are comparable. Treat last year’s data cautiously when:
- The product launched after last year’s peak, so there is no relevant history.
- The category is being discontinued or replaced by a materially different product.
- The item was a trend product with a likely one year life.
- Distribution, pricing, or store coverage changed substantially.
- Last year included long stockouts that suppressed recorded sales.
In these cases, use analogues, supplier information, early season signals, and a lower initial commitment where possible. A new style of sandal may borrow the shape of an older style, but it should not inherit its history without adjustment.
There is another risk in the opposite direction: over indexing on last year’s peak. A viral post, an unusual heatwave, a nearby event, or a competitor’s stockout can create a one time demand spike. If you copy that week into this year’s forecast, the model turns an exception into a plan.
Review the cause of every unusual peak before accepting it. Mark weeks affected by weather, promotion, publicity, or supply shortages. If the event is unlikely to repeat, reduce its weight or use the surrounding weeks to estimate a normal peak. The goal is not to make the history look tidy. It is to separate repeatable demand from an event you cannot reasonably plan around.
Finally, make the forecast reviewable. Keep the historical velocity, seasonal index, growth assumption, buffer, and allocation logic visible in the same planning record. A buyer should be able to explain the order without rebuilding the calculation from five spreadsheets.
When the season starts, watch the first two peak weeks against the model. If demand is running hotter, place a focused reorder on the fastest selling locations and protect the remaining supplier capacity for core sizes and colors. If demand is colder, pause the next receipt, redirect stock to the locations still selling, and reduce the reorder rather than defending the original forecast. Mid season decisions should respond to actual velocity and remaining lead time, not pride in the first order.
