brauniq.
HUEN
Back to the blog
Webshop

AI inventory planning: prepare your webshop for the Christmas peak

Webshop stock planning before Christmas: product-level forecasting, reorder point, safety stock, what still fits this year.

BBrauniq team· AI and automation 6 min read
AI inventory planning: prepare your webshop for the Christmas peak

Decide your Christmas peak stock as early as you can this autumn: Black Friday 2026 falls on 27 November, and anything ordered later can arrive too late given supplier lead times. This year, a simple product-level forecast from last season and this year's trend is already worth a lot. A database-backed model is next year's job.

Why November is the critical month for webshops

According to flash reports from the Hungarian Central Statistical Office (KSH), mail-order and online retail made up 8.9% of total retail turnover in October 2025, 11% in November and 9.3% in December. A year earlier the same figures were 8.2%, 9.8% and 8.9%.

So in both years the online channel weighed most in November of the three months, even though total retail turnover is higher in December.

One fixed point of the November campaign period is Black Friday, the Friday after US Thanksgiving. According to the US federal holiday calendar, Thanksgiving falls on 26 November in 2026, so Black Friday is 27 November.

What demand forecasting can do at SMB scale

Demand forecasting gives you one number per product per week: expected sales. The input is your order history, promotion calendar and the season. The AI part means the machine does, for hundreds of items at once, what a buyer can only do in their head for a few dozen key products.

Promotions need separate handling. The forecasting textbook by Hyndman and Athanasopoulos handles special events with a dedicated variable, and uses Easter to show how to treat a holiday that moves each year. Black Friday also lands on a different date every year, so align last year's data to the event, not the calendar date.

Reorder point and safety stock

Two decisions follow from the forecast. The reorder point is the stock level at which you have to order: the expected sales during the supplier lead time plus the safety stock. Safety stock is the buffer for when demand is higher than expected or the supplier is late.

The documentation of Microsoft Business Central, an enterprise resource planning (ERP) system, describes the same logic. The system typically suggests an order when projected available inventory reaches or drops below the reorder point. If it falls below the safety stock, it suggests a replenishment order to refill it.

A worked example: when to order

A made-up example: you expect 40 units a week of a product in the peak weeks, and the supplier delivers in two weeks, so the lead time needs 80 units. Add a 30-unit buffer and the reorder point is 110 units. If your stock is below that, order today, not next week.

What a stockout costs, and what excess stock costs

A stockout hits at the most expensive moment: during the campaign, when you have already paid for the ad click and the customer sees "out of stock" and goes to a competitor. The damage can outlast the lost order, because a disappointed customer may well start their next search elsewhere.

Overstock is quieter, but it is money too: seasonal goods left on the shelf in January tie up capital and warehouse space, and often end in a markdown. A good forecast balances the two risks product by product, rather than driving stockouts to zero at any price.

Two speeds: what still fits this year

This year the goal is a fast, transparent model you can run within a few days. The starting point is the seasonal naive method, where by the textbook definition the forecast equals the value from the same period last year; we then adjust it for this year's trend. It is a statistical baseline you can calculate in a spreadsheet.

  1. Last season by product: weekly, November to mid-December, aligned to Black Friday week.
  2. This year's trend: compare this year's August-September with last year's, and scale last season by that ratio.
  3. Manual adjustments: new products, items being phased out, new promotions.
  4. Order list: subtract current stock and placed orders, then schedule by lead time.

What this year's model gives you, and what to ask suppliers

This is not perfect, but it is a better starting point than "we order the same as last year", because it works by product and with this year's trend. Its biggest benefit is that for your main products you find out in time where stock will be tight.

Do not take the lead time from the catalogue: ask the supplier whether they will hold it in the November weeks, and if they cannot guarantee it, size the safety stock for the slower delivery. Ask about the minimum order quantity too, because it can override the calculated quantity.

In January: measure the result

After the season, compare the forecast with actual sales product by product, and note where something sold out early. That sheet becomes the first test of next year's model. It shows which product groups were fine with last year's pattern and which need more inputs.

Next year: a database and automatic purchase suggestions

After the season it is worth building a proper foundation. Your webshop orders, stock movements, promotion calendar and supplier lead times go into one database, and the model refreshes the forecast for every product weekly. From that, purchase suggestions are generated automatically, which the buyer approves or edits.

This often needs development on the webshop side too, for example a reliable export of stock data and supplier information. If you have already automated what happens after the order, as we described in our article on webshop automation, the forecast sits on the same data flow.

The prerequisite: clean order data

The seasonal naive method needs at least one full season from last year to compare against. Separating season from growth takes at least two seasons, since only a two-year comparison shows your growth. With less history, it stays closer to an estimate.

Cleaning the data

Cleanliness matters more than volume. Filter out cancelled and returned orders, treat items with a new SKU (item code) as one product, and mark weeks when the product was out of stock. Zero sales then means missing stock, not missing demand, and a model unaware of this under-plans.

Want to see how reliable a forecast your own order data supports this season? See our predictive analytics service, or tell us which system you export your sales from.

Sources

Further sources

#inventory planning#demand forecasting#Black Friday#predictive analytics#webshop

FAQ

How much data do I need for webshop demand forecasting?

The seasonal naive method needs at least one full season from last year, because that is what it compares against. If you also want to separate the season from growth, you need at least two seasons of clean order data, because only comparing two years shows how much you grew.

What are the reorder point and safety stock?

The reorder point is the stock level at which you have to order: the expected sales during the supplier lead time plus the safety stock. Safety stock is a buffer for the case when demand is higher than expected or delivery is late.

When is Black Friday in 2026?

On Friday, 27 November 2026, because Thanksgiving falls on 26 November. Since the date moves from year to year, last year's sales data has to be aligned to the event, not to the calendar day.

Is it still worth planning Christmas stock with a forecast this year?

Yes, if your suppliers' lead times still fit. The simple model, meaning last season by product scaled by this year's trend, can be built in a spreadsheet from your existing order export, and it shows the products where stock will be tight in time.

Related service

Data analysis and predictive AI

Brauniq builds predictive analytics from a company's own data: it forecasts demand, churn and stock, detects anomalies and shows the results on an interactive dashboard.

Service detailsFree assessment

Also related: Webshop development

Related articles