Clear Demand Presents Fundamentals of Retail Science: Episode V (Markdown in Retail)

Markdown is a hot button in retail optimization for several reasons. By “markdown” we mean end-of-product-life discounts to clear inventory. The term is used to indicate change in product valuation in retail accounting and is often used to indicate promotion discounts in soft-lines. In our retail-science discussion, however, we are talking about decision support to end-of-product-life markdowns.

 The challenge of markdown science is to sell the last unit of inventory on the last day of the event while following the rules of the retail business. As forecasts are approximate predictors of retail sales, actual sales are almost always higher or lower than forecast and markdown price schedules often have to be revised during an event.

How big a business is markdown? In seasonal apparel sales it may be half the business and even grocery stores carry seasonal or changing products that need inventory clearance.

We will begin by describing markdown in the product life cycle, formulate the markdown science problem next, then discuss business constraints and rules, and finally explain the need for find-grain, low-level analysis.


While most grocery products are on the shelves year in and year out, many products have a defined life cycle. There are several reasons products are put on clearance.

Some product lines are simply discontinued. This is common in consumer electronics where one technology is overwhelmed by something newer. Video cassette recorders (VCRs) gave way to digital video disks (DVDs) and there were real bargains for those who wanted a nice VCR. Cathode-ray-tube (CRT) computer monitors gave way to flat screens and digital superseded analogue video cameras.  Technology advances and renders the current offerings obsolete.  These are perfect candidates for a markdown strategy.

Some products are one-season products. Typical in apparel fashion, these products are intended for one season and one season only. The buyer has one chance to order about a year in advance, the merchandise comes at the beginning of the season, and whatever is left at the end is worthless. By next summer, for example, this summer’s swimsuits will be outdated fashion to shoppers looking for the latest styles.

Other products are yearly and seasonal. The retailer may sell the same snow shovels each winter and the same barbecues each summer, but no off-season stock is kept. At the end of the season the remaining inventory is marked down and sold.

However they begin their lives, all these products end their lives the same way. The retailer decides on a date when this product will be cleared off the shelves one way or another. It will be sold, perhaps for pennies on the dollar, or it will be disposed of. Sometimes a vendor offers a return value or there is some other salvage value and sometimes the inventory is just thrown out.



The common thread in all the markdown cases is that there is some fixed inventory at each product and store and a predetermined time window for selling it. Retail shelf space is generally dear enough that we’re not going to leave merchandise out indefinitely. We’ll pick a date and decide that the markdown starts now and ends then and it’s over after that.

The delicate balance of inventory and time makes markdown science particularly difficult. We’re aiming for a specific-inventory sales target over a period of time through the variation of consumer demand. The markdown discount is itself a promotion with some promotional lift, so we have to take that effect into account in engineering markdown discounts.

To take full advantage of retail science in markdown, we monitor our progress regularly throughout the markdown event. Every two weeks, or maybe every week, we check product sales against our forecasts to see if we’re ahead or behind sales projects. If we’re selling more units than we expected, then maybe we can reduce the markdown schedule, maybe we can delay planned-future price changes or set the future price points higher. On the other hand, if we’re selling fewer units than we expected, the maybe we can accelerate sales with deeper price cuts, cutting prices sooner, or doing both.

As an airplane pilot, I think of the perfect markdown like the perfect approach to landing: we want the airplane to make a smooth descent to touch down perfectly and gently at the beginning of the runway with minimum airspeed. Similarly, we want the markdown to sell all the inventory so the last unit sells on the last day with maximum revenue. To this end, both pilot and retailer may find themselves needing adjustments to their trajectories in the turbulence of varying winds and demand during their altitude and inventory descents.



Retail science can tell us how best to price a product at a store for a period of time to exhaust a given inventory. While that’s a hard-enough problem to work on, scientific markdown optimization is compounded by business rule requirements.

Business rules restrict the length and depth of markdown steps. These rules are a retailer’s way to control a markdown event. These following rules may be applied:

  • Initial markdown discount: The percent-off of the initial markdown is restricted to a range, for example 25 to 50 percent.
  • Markdown discount change: The change of discount from one markdown step to another is restricted to a range, for example 20 to 40 percent.
  • Markdown price points: Specific markdown levels, either a list of percent-off values (15%, 20%, 25%, … , 85%) or a list of ending-number rules (ending in .49 or .99).
  • Number of markdown steps: A markdown could have many markdown steps (subject to the other business rules) or it could have none if regular price is maintained. This rule can require a minimum number of markdown steps below regular price and keep us from having too many.
  • Duration of markdown steps: Minimum or maximum duration for each markdown step, for example two to four weeks at each markdown level.
  •  “Coherent” product-store combinations: Markdowns synchronized across products or stores, or both. A retailer might want to keep all the stores the same for a particular product so shoppers don’t go from store to store to find the most favorable markdown. (Such a shopper might find a better deal from a competitor!) Also, a retailer might want to mark down all the sizes of a fashion item with the same discount schedule.



 Not all the inventory is at the store during a markdown. Some of it can be at a central location, a distribution center (DC). While this makes a markdown more complicated, DC inventory gives the retailer another dimension of choice.

The retail-science question of DC inventory is which stores get the most value from the central supply. Retail science can tell how to distribute the inventory based on demand forecasts. Stores can have different inventory levels as well as different demands, so spreading DC inventory equally across all the stores is seldom the best way to allocate the inventory.

There are really two DC-to-store problems in markdown. The first is deciding price schedules for a product at the DC-supplied stores based on the available DC inventory. We can figure out the best DC-inventory distribution and base price schedules on that inventory being available.

The second DC-to-store markdown problem is figuring out when to ship inventory from DC to store. The uncertainty of markdown demand makes it wise to retain some inventory at the DC for a while in case actual demand for product-store combinations turns out different from the forecast. The same uncertainy makes us want to keep some extra inventory at the store in case sales are higher than originally forecast.



Managing a markdown is a wonderful opportunity for retail science to add value. The inflexibility of fixed inventory and predetermined schedule create more-complex decisions over specific products, stores, and dates. The opportunity to optimize the markdown schedule is increased by the opportunity to re-optimize it during the markdown event.