The obvious answer is to reduce the price. But deciding when to reduce it, how deeply to reduce it, and which products deserve a discount is much more complicated.
A markdown is not simply a red sticker placed on an item. It is a business decision involving inventory, customer demand, seasonality, competition, product age, store capacity and expected future sales. A poorly timed discount can destroy margin on an item that would have sold at a higher price. A delayed discount can leave a retailer with obsolete inventory that eventually has to be liquidated at an even greater loss.
This is why modern retailers are increasingly treating markdowns as an analytical problem rather than a last-minute clearance exercise.
How Did Retail Markdowns Begin?
Markdowns have a much longer history than today's online sale events might suggest. Research on retail pricing traces the use of markdowns, sometimes described as price skimming, back to the 1920s. Their importance expanded considerably from the late twentieth century as department stores and apparel retailers increasingly relied on promotional pricing to move merchandise.
The basic logic was straightforward.
A retailer might launch a new product at a relatively high price because some customers were willing to pay more for immediate access. As the selling season progressed, the retailer could lower the price to attract customers with greater price sensitivity.
Fashion made this approach particularly important. A winter jacket has limited commercial value after winter. A festive collection may lose relevance after the festival. A particular colour or design may become unpopular long before the physical product becomes unusable.
Therefore, the objective was not necessarily to sell every item at full price. The objective was to capture the highest practical value from the inventory over its entire selling life.
That principle remains relevant today, but the technology used to make the decision has changed dramatically.
Why Traditional Markdown Decisions Often Fail
Historically, markdown decisions were heavily dependent on fixed calendars and managerial experience.
For example, a retailer might decide:
10% off after four weeks
20% off after eight weeks
30% off near the end of the season
50% off for final clearance
The problem is that products do not behave identically.
One SKU may sell 80% of its inventory within two weeks. Another may sell slowly but consistently. A third may suddenly become popular after a social-media trend. Applying the same markdown schedule to all three products can produce unnecessary margin loss.
There is another problem: price alone does not determine demand.
Weather, location, stock availability, competitor prices, customer demographics, online traffic, product reviews and seasonality can all influence purchasing behaviour.
Modern markdown optimization therefore asks a more useful question:
What price is most likely to produce the best financial outcome given the inventory and demand situation?
That shift—from discounting products to optimizing decisions—is at the heart of modern retail analytics.
What Markdown Optimization Actually Measures
A sophisticated markdown system can evaluate several variables simultaneously.
1. Inventory position
How many units remain?
An item with 20 units left requires a different strategy from an item with 2,000 units sitting across multiple warehouses.
2. Sales velocity
How quickly is the product selling at its current price?
A slow-selling product may require intervention earlier than one that is naturally approaching sell-through.
3. Price elasticity
Price elasticity estimates how demand may respond when the price changes.
If reducing a product from ₹2,000 to ₹1,800 is likely to create a meaningful increase in demand, the reduction may be justified. If demand barely changes, the retailer may simply be sacrificing margin.
4. Time remaining
Seasonal products become more difficult to sell as the relevant season approaches its end.
A retailer therefore needs to consider not just today's demand but the number of selling opportunities remaining.
5. Location
The same product can perform differently in different locations.
A raincoat may sell rapidly in one region while remaining stagnant in another. A retailer that applies a nationwide markdown could unnecessarily discount products in stores where demand remains healthy.
A Real-World Example: Macy's
One frequently cited example of large-scale markdown analytics comes from Macy's.
With tens of millions of items across hundreds of stores, the retailer faced an enormous pricing-analysis problem. Historical sales information was used to support pricing decisions, but the process became increasingly difficult to manage manually.
After implementing SAS Markdown Optimization, Macy's reported that it could complete its analysis 22 times faster than before. The case demonstrates an important lesson: at large retail scale, the challenge is not merely knowing what happened historically. It is converting large amounts of historical information into pricing decisions quickly enough to influence current inventory.
The significance of the example extends beyond Macy's.
For a retailer managing thousands or millions of SKU-location combinations, even a small improvement in the quality or speed of pricing decisions can have a substantial financial impact.
Case Study: ASOS and Machine Learning
The next stage of markdown optimization is moving beyond conventional rules and statistical models toward machine learning.
Research describing systems deployed at ASOS.com provides an interesting example. The researchers developed two markdown-management approaches designed for online fashion retail. One system was designed as a practical starting point where limited demand information was available, while the other incorporated price elasticity into a broader optimization framework.
In controlled online testing, the researchers reported profitability improvements relative to manual pricing strategies, with the two systems producing improvements of 86% and 79% respectively in the reported experiments.
The broader lesson is important.
Machine learning does not necessarily mean giving an algorithm complete control over prices. It can instead help retailers estimate what might happen under different price scenarios and provide decision-makers with better alternatives.
For example:
₹2,499 → expected sales: 120 units
₹2,299 → expected sales: 145 units
₹2,099 → expected sales: 190 units
The retailer can then compare expected revenue, margin and inventory clearance rather than selecting a discount arbitrarily.
Case Study: Fresh Retail and Perishable Inventory
Markdown optimization becomes even more interesting when products have extremely short lifespans.
Consider fresh food.
A supermarket cannot keep strawberries, prepared meals or other perishables indefinitely. Waiting too long can result in waste. Discounting too early, however, can unnecessarily reduce revenue.
Research on e-commerce fresh retail developed a multi-period approach that combines demand prediction, price elasticity and optimization. The framework was also reported as being deployed in a fresh-retail environment associated with Freshippo.
This illustrates a crucial distinction between ordinary retail and perishable retail.
For fashion, the cost of waiting may be leftover inventory.
For fresh food, the cost of waiting may be inventory that cannot be sold at all.
The optimal markdown therefore depends heavily on the product's remaining economic life.
The Rise of AI in Markdown Decisions
Retail markdown optimization is now entering another phase.
Artificial intelligence can combine signals that traditional spreadsheets struggle to process at scale.
A modern system could consider:
historical transactions
current inventory
competitor prices
website searches
product views
conversion rates
weather
seasonality
regional demand
customer behaviour
promotional history
product attributes
The objective is not simply to predict sales.
The more useful objective is to estimate what could happen under different pricing decisions.
This distinction is becoming increasingly important in e-commerce, where retailers can potentially test pricing strategies across digital channels much faster than traditional stores.
Markdown Optimization Is Not the Same as Bigger Discounts
One of the biggest misconceptions about markdown optimization is that it means finding ways to discount products more aggressively.
In reality, the opposite can be true.
A good markdown strategy may recommend not discounting an item.
Suppose a product is selling steadily and inventory is limited. A large discount could create unnecessary demand and eliminate profitable sales.
Another product may have substantial inventory but almost no customer interest. A modest markdown may not be sufficient to change behaviour, making a deeper intervention economically rational.
The objective is therefore not:
“How much can we discount?”
It is:
“What pricing action creates the best balance between demand, margin and inventory?”
A Practical Markdown Optimization Framework
Retailers looking to build a modern markdown process can start with five stages.
Stage 1: Establish product-level visibility
Track sales, inventory, age, price and location for every important SKU.
Stage 2: Identify performance patterns**
**
Separate fast-moving, stable, slow-moving and declining products.
Stage 3: Forecast future demand
Estimate expected sales under the current price and under possible markdown levels.
Stage 4: Evaluate financial outcomes
Compare expected revenue, gross margin, inventory remaining and potential clearance costs.
Stage 5: Monitor and learn
After implementing a markdown, measure what actually happened. The result becomes new information for future pricing decisions.
This creates a continuous feedback loop rather than a one-time discounting exercise.
The 2026 Perspective: Markdown Optimization Becomes More Contextual
The next generation of retail markdown systems will likely be less dependent on rigid discount calendars.
Retailers increasingly operate across stores, websites, marketplaces and mobile applications. Inventory can move between channels, while customers can compare prices almost instantly.
That means a markdown decision may need to consider the entire retail ecosystem, not just a single store.
A product might be slow in one location but selling quickly online. Instead of discounting it immediately, the retailer could transfer inventory. Another item may have weak demand in physical stores but perform well after targeted digital promotion.
This creates a broader principle:
The best markdown is not always a lower price. Sometimes it is a better inventory decision.
Final Thoughts
Markdowns began as a practical way for retailers to extract value from merchandise that might otherwise remain unsold. Over time, the process evolved from handwritten price changes and calendar-based clearance events into sophisticated analytical systems.
Today, retailers can combine inventory data, demand forecasting, price elasticity, experimentation and artificial intelligence to make more informed decisions.
The winners will not necessarily be the retailers offering the deepest discounts.
They will be the retailers that understand which products need a price change, when that change should happen, how customers are likely to respond, and what alternative action might create greater value.
Markdown optimization, in that sense, is no longer simply a clearance strategy.
It is becoming an important part of modern retail decision-making—connecting pricing, inventory, analytics and customer behaviour into one continuous system.
This article was originally published on Perceptive Analytics.
At Perceptive Analytics our mission is "to enable businesses to unlock value in data." For over 20 years, we've partnered with more than 100 clients — from Fortune 500 companies to mid-sized firms — to solve complex data analytics challenges. Our services include Enterprise AI Consulting and Power BI Consultant, turning data into strategic insight. We would love to talk to you. Do reach out to us.
Top comments (0)