Forecasting Remaining Inventory and Seasonal Trends
End-of-season clearance sales are never a surprise: they are a sign of a missing or unreliable forecast of remaining inventory, one that was prepared too late.
The combined cost of stockouts and excess inventory reached approximately 1,730 billion dollars worldwide in 2025 (IHL Group).
Projecting this surplus several weeks in advance is exactly what BOOPER’s AI-powered sales forecasting does: it enables proactive and measured markdowns, rather than rushed, last-minute discounts.
End-of-season inventory clearance comes as no surprise: it’s the visible result of a forecast that failed to anticipate, several months in advance, how much inventory would remain to be sold. Forecasting remaining inventory turns a decision made in a hurry into a well-prepared one.
Clearance sales begin months before the markdowns
When a category manager discovers at the end of the season that 30% of their stock remains unsold, the problem didn't begin that week: it started at the time of purchase or production, based on a sales forecast that proved overly optimistic. The destocking isn't the cause of the problem; it's a symptom of a lack of or insufficiently reliable forecast of remaining stock.
Forecasting remaining inventory involves, even before the start of the sales season, answering this question: If sales follow the expected trajectory, how much will remain to be sold by the deadline?
The Overall Cost of Poor Planning
The scale of the problem across the industry has been documented. According to IHL Group, an analytics firm specializing in retail inventory imbalances, the combined cost of stockouts and overstock reached approximately $1,730 billion globally in 2025, representing nearly 6.5% of global retail sales.
estimated annual global losses—resulting from both stockouts and excess inventory—amount to approximately 6.5% of global retail sales (IHL Group).
Excess inventory (half of this problem) is directly linked to an inadequate forecast of remaining inventory: without a reliable estimate of what will remain to be sold, end-of-season markdowns become deeper and occur later than they should have.
Seasonality: The Least Well-Accounted-For Variable
Seasonality is not simply a matter of “it sells better in the summer.” Three distinct factors must be taken into account; otherwise, any forecast of remaining inventory will be inaccurate:
- Seasonality of the category: the expected sales pattern over the course of the year, specific to each product family.
- One-time events—such as sales, holidays, or unusual weather—that temporarily shift demand without changing the underlying trend.
- A trend shift: a lasting change in purchasing behavior, as distinct from a simple, normal seasonal variation.
Confusing these three dimensions is the most common mistake: treating a deviation from the trend as a mere seasonal variation delays the decision to write off inventory, until the remaining inventory becomes too large to be sold under favorable conditions.
How to Build a Reliable Forecast of Remaining Inventory
Start from the history, category by category
Develop a typical seasonal sales curve for each product family, covering several years if historical data allows.
Project the remaining sales pace
Based on sales already made at the start of the season, project the likely sales trend through the deadline.
Estimate the likely remaining inventory
Compare available inventory with the projected sales trend to estimate, several weeks in advance, how much inventory will remain to be sold.
Initiate markdowns proactively, not reactively
Use this forecast to activate the planned markdown schedule; see our article on the markdown schedule.
What This Means in Practice
A reliable forecast of remaining inventory does not eliminate the need for markdowns; it simply changes the timing of the decision. Instead of discovering a problem at the end of the season and correcting it in a rush with the deepest possible markdowns, the retailer anticipates the shortfall several weeks in advance and can apply a more measured markdown, spread out over time, with a significantly less severe impact on margins.
Before the next end-of-season clearance sale
- Do I have a forecast for remaining inventory before the peak of the sales season?
- Can I distinguish between normal seasonality, a one-time event, and a trend deviation?
- Does this forecast trigger any action, or is it just a single number?
- Did I compare my residual forecast from last season with the actual results to adjust the method?
FAQ
The questions we are most frequently asked before getting started.
This is an estimate, made before the end of the sales season, of the amount of inventory that will remain unsold by the deadline if sales continue on their projected trajectory. It answers a specific question, asked as early as possible: if the current pace continues, how many items will remain on the shelves or in stock on the day the season ends?
This forecast is developed in three steps: we start with the typical seasonal sales curve for each product family, based on historical data; we project the remaining sales pace based on sales already made at the beginning of the season; and then we compare the available inventory to this projected trajectory to estimate the likely remaining inventory several weeks in advance.
It is this foresight that makes all the difference: a category manager who discovers 30% of inventory unsold at the end of the season does not have an end-of-season problem; he has a forecasting problem that stems from purchasing or production. The combined cost of stockouts and overstocking reached approximately $1,730 billion worldwide in 2025, according to the IHL Group—nearly 6.5% of global retail sales. Overstocking, which accounts for half of this total, is directly linked to this forecasting error.
For a retailer, having this forecast at hand transforms a last-minute end-of-season markdown into a planned and measured price reduction, with a significantly less severe impact on margins than a belated price adjustment—and that is the focus of our reference guide on sales forecasting.
Because it combines three distinct phenomena—normal seasonal patterns, one-time events, and long-term trend shifts—that are often confused in forecasts made by hand or using overly simplistic tools.
The seasonality of a category reflects the expected sales pattern over the course of the year, which is specific to each product family. One-time events (sales, holidays, unusual weather) temporarily shift demand without altering the underlying trend. A trend deviation, on the other hand, is a lasting change in purchasing behavior that must be clearly distinguished from a simple, normal seasonal variation; otherwise, the resulting forecast for remaining inventory will be skewed.
Confusing these three dimensions is the most common mistake observed in the field: treating a deviation from the trend as a mere seasonal variation delays the decision to set aside the stock, until the remaining stock becomes too large to be sold under favorable conditions.
For a retailer, failing to distinguish between these three signals is like managing markdowns with a flawed tool: the right decision—whether to adjust the price, speed up sales, or, conversely, leave things as they are—depends entirely on knowing which of the three phenomena is actually at play, as detailed in our method for incorporating exogenous events into sales forecasts.
From the start of the sales season, using the first few weeks of actual sales to project the trend through to the deadline, waiting until the end of the season to assess the gap is like discovering the problem at a point when it’s no longer possible to address it except under urgent circumstances.
In practical terms, the method starts withsales history broken down by category—compiled over several years whenever possible—to determine the typical sales curve for each product family. Once the season begins, actual sales from the first few weeks are used to project the remaining sales pace, which is then compared to available inventory to estimate, several weeks in advance, how much inventory will remain to be sold.
This forecast is only valuable if it triggers concrete action: it must inform the planned markdown schedule, not remain an isolated figure that is consulted too late. It is this shift—from discovering the need for markdowns at the end of the season to anticipating them from the very start—that distinguishes a reactive markdown from a proactive one.
For a retailer, starting the forecasting process on the very first day of sales provides several additional weeks of flexibility to adjust prices gradually, rather than having to make all adjustments at once at the end of the period—provided there is sufficient sales history to make accurate forecasts.
No, but it allows for a more measured price reduction that is spread out over time, rather than a sudden, drastic markdown decided in a rush at the end of the season.
The difference does not lie in the principle of markdowns (remaining inventory must always be sold off) but in the timing of the decision. Without forecasting, the retailer discovers the extent of the problem at the end of the season and must apply the maximum discount to quickly sell off what remains. With a reliable forecast established several weeks in advance, however, the retailer can spread out the markdowns, gradually adjusting prices as the gap between actual sales and the expected trajectory becomes clear.
This phased approach directly affects the margin: a one-time, late markdown significantly erodes profitability across the entire remaining inventory, whereas a gradual, early markdown allows for the sale of part of the inventory at intermediate price points before applying the steepest markdown to the final balance.
For a retailer, the goal is therefore not to eliminate end-of-season markdowns, but to turn them into a planned decision rather than a last-minute reaction—resulting in a significantly less severe impact on margins, as detailed in our article on markdowns and inventory clearance without sacrificing margins.
Sources: IHL Group, study on global retail inventory distortion, ihlservices.com · Booper product data (GENIUS Predict module, sales forecasts, and seasonal scenarios).
The volume sold during a promotion does not indicate whether it created value: part of it comes from similar products (cannibalization), and another part from purchases that were simply brought forward.
Each promotional scenario must be costed out prior to launch using the same metrics: base sales, actual incremental sales, cannibalization, carryover, halo effect, net margin for the category, and cost per unit actually gained.
Margin-volume arbitrage can then be explained: a stated objective, visible forecasting factors, constraints adhered to, and a documented validation process.
A national food retailer with more than 1,700 stores and several million price points per year: With Booper, its pricing teams simulate the impact of each decision on margins, competitiveness, and price perception before implementing it.
Key takeaway: Pricing, promotions, and markdowns are three factors that constantly influence one another, but are still managed using separate tools at most retailers.
This fragmentation creates inconsistencies that are invisible in the short term (a muddled pricing image, margins eroded by promotions that aren’t properly coordinated with markdowns) but costly in the long term. Gartner has, in fact, formalized this convergence as a distinct market category: unified optimization of pricing, promotions, and markdowns.