Skip to main content
56 min readmarkdown optimization softwaremarkdown optimization

What Is Markdown Optimization Software?

Markdown optimization software is the system that decides when, how deep and where to reduce the price of each product so that it sells through by its exit date at the highest achievable margin, by forecasting price response per style and location, scheduling the markdown cadence and sizing the markdown budget against the plan's reductions.

Markdown optimization software is the system that decides when, how deep and where to reduce the price of each product so that it sells through by its exit date at the highest achievable margin, by forecasting price response per style and location, scheduling the markdown cadence and sizing the markdown budget against the plan's reductions. It is the system of record for the price-reduction decision: the exit date each style has to clear by, the forecast of how each style sells at each candidate price, the constraints the business places on the price path, and the markdown dollars every reduction consumes against the planned reductions in the merchandise financial plan. The merchandise plan tells it how much margin the season was built to make and how much markdown it was built to absorb; the allocation tells it where the units are; the sales feed tells it how fast they are leaving; it hands back a price per style per location per week, and a number for what that price path will cost.

This guide defines the software category. It is not a guide to marking down well, which markdown optimization in retail covers for apparel, with the causes of markdowns, a trigger table, the rules-based and model-based approaches and a worked cadence, and which markdown and exit strategy by vertical covers across ten categories. It answers a narrower question that arrives in two forms, "what is markdown optimization software?" and "do we need one, or is this a planning problem?". It covers the decisions the software owns; the inputs it reads and the outputs it writes; how it works mechanically, as a price-response model and an optimizer under constraints, without claiming any vendor's method; how it differs from markdown planning, promotional planning, dynamic pricing and clearance management; where it sits against the merchandise financial plan, the open-to-buy, the assortment, the allocation and the in-season reforecast; who uses it; how a suite module, a point solution and a spreadsheet markdown calendar differ; what each of ten verticals needs it to carry natively; one fully worked example of two markdown paths on one style; the data it needs; the criteria for choosing one; and the ways implementations fail.

What "markdown optimization" means in a retail organisation

A markdown is a permanent reduction in the retail price of a product, taken because the product will not sell the quantity the business owns at the price it was ticketed by the date it has to be gone. The word carries two facts that the software exists to hold together. The first is a quantity and a date: a seasonal product has a finite window, set by the calendar, the floor set that replaces it, the model year that supersedes it or the dating that expires it, and whatever is left when the window closes is worth less than it was. The second is a price and a margin: every dollar taken off the ticket on a unit that would have sold anyway is a dollar out of the margin the season was planned to make at the initial markup, and every dollar not taken off a unit that then does not sell is a unit worth its exit value instead of its ticket.

Markdown optimization is the discipline of resolving those two facts for every style the business owns, every week, at the lowest total cost. "Optimization" is the right word, not a marketing one, because the decision has an objective, margin dollars kept, and constraints, the exit date, the price points the brand is willing to show, the number of times a ticket can be changed, the relationship between the store price and the online price and the price promised to wholesale accounts. A merchant does this in their head for a dozen styles; a planner does it in a workbook for a class; a markdown system does it for every style-color at every location, and does it again next week when the sales come in.

The function usually lives between two owners. The planner owns the money: the merchandise financial plan carries a line for planned reductions by class and month, and the markdown decisions of the season have to land inside it or be explained. The merchant or buyer owns the product: which styles are fashion and which are core, which can carry forward, which have a wholesale commitment behind them, and which exit route the residual can take. The markdown decision is where the two meet, usually in the weekly trade meeting, and the software's job is to bring both the forecast and the budget into that meeting so that the decision is made on the same numbers each week. The markdown and exit strategy by vertical guide covers the governance of that meeting; this guide covers the system that feeds it.

What the software is for

The discipline is older than any software. In a brand without a markdown system the decision is made in a workbook: a tab per class, a row per style-color, columns for units bought, units sold, sell-through to date, weeks of supply at the current rate, the week the style landed and the week it has to exit, and a column of markdown percentages that somebody keys in on Monday after the trade meeting. The workbook works until four things happen at once: the style is selling at different rates in three regions, a promotion ran in two of the eight weeks of history and the rate in those weeks is not the rate at full price, half the doors are broken in the core sizes so the chain rate understates demand, and the class has already consumed its planned reductions for the month and the planner cannot see it until the month closes.

Markdown optimization software exists because the markdown decision is a forecast made against a position that changes every week, under a budget that is spent as it is made. The question is never "should this style be marked down?" in the abstract; it is "at this rate, with this many units, in these doors, with this many weeks to the exit, what price path from here keeps the most margin, and what does it cost against the plan?". Every term in that sentence moves between one trade meeting and the next. A workbook holds one state of the position and one planner's guess at the price response; a markdown system holds the position as a live object, estimates the response from the history it holds, searches the price paths, and shows the planner the recommended one with its cost, so that the Monday question gets a list of exceptions instead of a morning of arithmetic.

Three consequences organise the rest of this guide. First, the software is defined by the forecast it makes and the constraints it enforces, not by its screens: a tool that cannot hold an exit date per style as a hard boundary, a margin floor as a hard boundary and the markdown budget as a running total is a discount calculator, however good the grid looks. Second, it is a system of record for the price-reduction decision, not for the plan, the buy or the goods: it reads the planned reductions from the merchandise financial plan, the position from the ERP and the allocation, and the price history from the sales feed, and each of those neighbours owns something it must not. Third, the markdown is the last commercial decision made on a unit before it leaves the business, and the first one whose cost is measured directly in margin, so the value of the software is mostly in how well it closes the loop: whether the markdown dollars it spent are read back against the buy that made them necessary, so that the hindsight can tell a demand miss from a buy-depth miss from a markdown-timing miss.

The decisions the software owns

Six decisions recur in every markdown system and in every markdown workbook one replaces, because they are the decisions the function requires. Products package and name them differently; the decisions are stable. The table gives the one-line version and the subsections give what each decision produces and what it hands to the next.

DecisionThe question it answersWorks at the grain ofHands on
First-markdown timingIn which week does this style take its first reduction, if any?Style-color by location groupA review week per style and a trigger that fired or did not
DepthHow far below the current price does the next step go?Style-color by location groupA new price, on the brand's price ladder
CadenceHow many steps, how far apart, from the first reduction to the exit?Style-colorA price path by week to the exit date
Store, zone or channel differentiationDoes the reduction differ by door, region, climate zone or channel?Style-color by locationA price per location group, and the parity rule it obeyed
Clearance exitWhat happens to what the last price will not clear by the exit date?Style-color by locationA residual quantity, an exit route and its recovery value
Promotional vs permanentIs this reduction a temporary event that restores the price, or a markdown that does not?Style-color by weekA flag that decides which history the response model reads

First-markdown timing

The timing decision is the one that matters most, because a reduction's effect depends on the traffic the style still has, and traffic decays through the window whether or not the price moves. The software answers it with a forecast rather than a rule: at the current rate and the current position, does the style clear by its exit date at full price? If not, at what week does the first reduction have to land so that the remaining steps can clear it without the deep final cut that costs the most? The rules-based version of this, a trigger table by week of life, is worked in the tactics guide; the software's version replaces the fixed week with a calculation per style, and holds the result as a review list for the trade meeting.

Depth

Depth is the size of the step, and it is where the price-response estimate does its work. A shallow step on a style that responds strongly to price can move more units than a deep step on one that does not, and the software's job is to know which kind of style it is looking at. Depth is constrained by the brand's price ladder, the endings and intervals it is willing to show, and by the margin floor, the price below which the business would rather take the exit route than sell the unit. The output is a price, not a percentage, because a percentage off an original retail that differs by channel is a different price in each.

Cadence

Markdown cadence is the sequence of steps from the first reduction to the exit: how many, how far apart, and where price stops being the lever. The software plans the whole path rather than one step at a time, because the right first step depends on how many steps are left and how far the exit is. A path with three steps and four weeks between them is a different first step from a path with one step and eight weeks, on the same style with the same position. Cadence is also where the operational constraints bite: a store that re-tickets by hand can take a step every two weeks and no faster, and an online channel can take one every day and should not.

Store, zone or channel differentiation

The same style is not equally slow everywhere. The software decides whether the reduction applies chain-wide, by region or climate zone, by store grade or door by door, and holds the parity rules that bound the decision: the gap the brand will tolerate between the store price and the online price, between the full-price channel and the outlet, and between the owned channel and what wholesale accounts were promised. Differentiating saves margin in the doors that were not slow; it costs execution in the stores that have to ticket, and it costs trust when a customer sees two prices. The decision is a trade, and the software's job is to show what each side of it is worth.

Clearance exit

Price does not clear everything. Whatever the last step on the price ladder will not sell by the exit date needs another route: a transfer to an outlet, a sale to an off-price buyer, a bundle, a carry-forward into the next season, a return to the vendor or, at the end, donation or destruction. The end-of-season exit strategies guide covers the five routes and their margin arithmetic. The software's part is to forecast the residual at the exit date for each price path, value it at the route it will take, and count that value in the objective, so that a path that leaves more residual at a higher price is compared honestly with one that leaves less at a lower one.

Promotional vs permanent

A promotion is temporary and restores the price; a markdown is permanent and does not. The two are different decisions with different owners, different accounting and, critically, different signals. The software has to flag which is which in the price history it reads, because a week in which a style sold at 20% off in a sitewide event is not evidence of how it would sell at a permanent 20% markdown: the event brought traffic that the markdown will not. Promotional planning decides the events; markdown optimization has to read around them, and a system that cannot tell the two apart in its own history estimates the wrong price response.

Inputs and outputs

The inputs are what the software reads each week; the outputs are what it writes back into the planning cycle. The inputs are mostly owned by other systems, which is why the integration list matters as much as the method. The three to check for first are the exit date, the planned reductions and the promotion flag, because a system can pass every screen-level test without reading any of them.

InputGrainOwned byWhy it matters
Sales by style, color, size, location and week, at the price soldSKU by location by weekPOS, e-commerce, marketplace feedsThe response model's history; the current rate the forecast starts from
On-hand, on-order and in-transit inventorySKU by locationERP, WMSThe quantity the price path has to clear, including what has not landed yet
Original retail, current price and unit costSKUProduct master, costingMargin per unit at every candidate price; the markdown dollars each step consumes
Exit date per style or classificationStyle or classPlanning calendar, assortment planThe boundary every path must reach; a style without one cannot be optimized
Planned reductions by class and monthClass by monthMerchandise financial planThe markdown budget the season's decisions have to land inside
Promotion and event calendarStyle or class by weekPromotional planningSeparates temporary lift from permanent price response in the history
Stockout and broken-run flagsSKU by location by weekAllocation, inventoryWeeks where availability capped sales are not evidence of weak demand
Store attributes, grades, zones and channel rulesLocationAllocation, merchandisingWhere differentiation is allowed, and the parity rules that bound it
Price ladder, step limits and margin floorBrand or classMerchandising policyThe constraints the optimizer searches inside
Product attributes and several seasons of historyStyleProduct master, hindsightThe basis for pooling similar styles when one has too little history
Wholesale commitments and markdown allowancesStyle by accountWholesale planningOwned-channel reductions below a promised price, and markdown money owed, are costs the objective has to count

The outputs are fewer and more consequential.

OutputGrainConsumed by
A recommended price per style per location group per week, to the exit dateStyle-color by location group by weekThe trade meeting, then the pricing system, POS and e-commerce
The forecast at that path: units by week, revenue, margin, and the residual at exitStyle-colorThe in-season reforecast and the open-to-buy
The markdown dollars the path consumes, against the planned reductionsClass by monthThe merchandise financial plan
A review list: styles whose trigger fired, whose path changed or whose constraint boundStyle-colorThe planner and the merchant
A record of the decision taken, the recommendation it replaced and the reasonStyle-color by weekThe hindsight

How it works, explained by mechanism

Every markdown optimizer, whatever its vendor calls the method, has the same two parts, and a planner evaluating one should be able to describe both in plain terms before signing anything. The first part is a price-response model; the second is an optimizer that searches price paths against an objective subject to constraints. Neither is a black box in principle, and a product that presents either as one is asking for trust it has not earned.

The price-response model

The model's job is to answer, for one style or one group of similar styles, "if the price were X next week, how many units would sell?". It estimates that from history: the weeks the style, or styles like it, sold at different prices, and how many units moved in each. The estimate is summarised as a price elasticity, the proportional change in units for a proportional change in price, or as a lift multiplier at each depth on the ladder, which is the same information in a form planners find easier to read. The price elasticity and promotion planning by vertical guide works the arithmetic of estimating it from a brand's own history, and why the number differs by category.

The hard part is not the estimate; it is separating the price effect from everything else that moved at the same time. Four things move with price in a seasonal product's history and have to be modelled out. The style's base rate decays through its window as the people who wanted it have bought it, so a markdown taken in week ten sells into a smaller base than the same markdown in week six, and a model that does not carry the decay reads the later markdown as weaker than it was. Seasonality in the category moves the base rate independently of the style. Promotions bring traffic that a permanent markdown will not, so promotional weeks have to be flagged and read separately. And stockouts and broken size runs cap sales below demand, so a week in which the core sizes were gone is not a week of weak demand at that price; it is a week of no supply, and a model that reads it as demand underestimates the response.

A single style rarely has enough history of its own at enough different prices, so models pool: they estimate the response for a group of styles that share attributes, a class, a price tier, a fabrication, a lifecycle shape, and apply it to each member, adjusting as the member's own sales come in. Pooling is where new styles inherit a response they may not deserve, and it is the first thing to ask about: how are groups formed, how does a style's own history take over from the group's as it accumulates, and what does the system do with a style that behaves unlike its group?

The optimizer

The optimizer's job is to search the candidate price paths from this week to the exit date and choose one. A price path is a sequence of prices by week, each on the brand's ladder, each no higher than the last. For each candidate path the model forecasts weekly units, and the optimizer values the path: units sold at each price times that price, less the cost of those units, plus the value of whatever is left at the exit date at the route it will take, less any markdown money or allowance the path triggers. The path with the highest value is the recommendation.

The constraints are what make this a planning tool rather than a pricing experiment, and the constraints are where the brand's judgement lives in the system. The exit date is the hardest: the position must clear, or reach a planned residual, by a fixed week, and no path that misses it is admissible however good its margin. A margin floor keeps every step above a minimum margin per unit or a minimum rate for the season. The price ladder restricts the search to the endings and intervals the brand shows. Step limits cap the number of markdowns per style and set the minimum interval between them, which is where the store's re-ticketing capacity enters. Channel rules hold parity or a defined gap between stores, online and outlet. Wholesale commitments forbid owned-channel reductions below what accounts were promised in the window they were promised it. And the planned reductions by class and month bound the whole class: a system that recommends the best path for each style independently and lets the class blow through its markdown budget has optimized the wrong thing.

Two practical notes. The optimizer re-runs every week, because the position, the rate and the weeks remaining have all changed, and a recommendation is a plan from here, not a commitment. And the planner overrides it: the system has to record the override and the reason, because the hindsight needs to know whether the overrides, taken together, helped or hurt. A system that forgets its own recommendations once they are overridden cannot learn, and neither can the team using it.

Markdown optimization vs markdown planning vs promotional planning vs dynamic pricing vs clearance management

Five disciplines touch the ticket price of seasonal product, and vendors blur them because the same screen can show all five. They differ in direction, duration, objective and the moment in the calendar they act. A brand buying one should know which of the five it is actually buying, and which of the others the product's demonstration is quietly implying.

Markdown optimizationMarkdown planningPromotional planningDynamic pricingClearance management
Direction of the price moveDown, permanentlySets the budget for downward movesDown temporarily, then restoredEither directionDown, or out of the channel
When it actsIn season, weekly, from the first trigger to the exitBefore the season, in the merchandise financial planAcross the season, on a calendar of eventsContinuouslyAt and after the exit date
Unit of decisionStyle-color by location groupClass by monthEvent by category or channelSKU by channel, often by dayResidual quantity by route
ObjectiveMargin dollars kept while clearing by the exitPlanned reductions that fund the planned marginTraffic and lift inside a margin budgetMargin or revenue on replenished productRecovery value and speed of disposal
Core inputPrice response per style, exit date, positionLast season's markdown rate by class, the buy depthEvent history, lift and pull-forwardDemand, competition and inventory signalsResidual units, route prices, logistics cost
Applies toSeasonal product with a finite windowEvery class in the planAny product, seasonal or continuityContinuity product that is refilledWhatever the last price did not clear
Where it is coveredThis guidePlanned reductions, open-to-buyPromotional planning, price elasticity and promotion planningDynamic pricing for fashionEnd-of-season exit strategies

Three of the five boundaries matter most in practice. Markdown optimization and markdown planning are the same money at two moments: the plan sets the planned reductions by class and month before anything is bought, and the optimizer spends them in season, style by style; a brand with a good optimizer and no plan has no budget to optimize inside, and a brand with a good plan and no optimizer discovers in the hindsight how the budget was spent. Markdown optimization and dynamic pricing share a model and differ in everything else: both estimate price response, but one moves a finite position downward to a date and the other moves a replenished position in either direction indefinitely, and dynamic pricing for fashion sets out why the second is the wrong tool for most seasonal apparel. Markdown optimization and clearance management meet at the residual: the optimizer forecasts how much will be left at the exit date at each path and values it; clearance management decides what happens to it and executes the route. A system that optimizes the price path without valuing the residual will always recommend holding price too long, because in its arithmetic the leftover units are free.

Where it sits with the plan, the assortment, the allocation and the reforecast

Markdown optimization is a downstream system. Almost everything it reads was decided earlier, by a planning system, and almost everything it writes feeds back into one. Where it sits in the planning stack decides how much of a brand's markdown spend it can reach, because the markdowns a buy decision has already made necessary are ones no price path can fix.

The merchandise financial plan and the open-to-buy. The merchandise financial plan carries planned reductions by class and month: the markdown dollars the season was built to absorb, set from last season's rate, the buy depth and the margin target. The open-to-buy carries them as a term in the receipt equation, because retail value that leaves the stock as markdown has to be replaced by receipts to reach planned ending inventory. The markdown system spends that budget and reports the spend; when a class is running ahead of its planned reductions, the news has to travel back to the open-to-buy, where the planner can decide whether to pull receipts or re-plan the margin. On the wholesale side, the plan may also carry markdown money, the allowance a vendor pays an account to fund the account's markdowns, and the optimizer that recommends an owned-channel reduction has to count what that reduction triggers in the account's claim. The what is merchandise planning software guide covers the plan the budget comes from.

The assortment plan. The assortment decides which options are bought, in what depth, for which clusters and channels, and it is where the exit date is first set: a floor set that replaces the option, a delivery that supersedes it, a model year that closes it out. It is also where the one determinant of markdown exposure that no later price path can change is decided, which is depth. An option bought at twice the depth its history supports will not be saved by any price path; the optimizer can only choose the cheapest way to pay for the buy. The hindsight that attributes markdown dollars to the buy decision that caused them is the loop that makes the next assortment better, and a markdown system that cannot report its spend by the planned depth it was spent on has left that loop open.

The allocation. The allocation decides where the units are, and markdown optimization decides what they are worth where they are. The two interact in both directions: a style that is slow in the chain average may be slow in twenty doors and sold out in ten, in which case the right answer is a transfer or a consolidation before any markdown, and a markdown optimizer that cannot see the door-level position recommends a price cut for a placement problem. The weeks of supply and sell-through the optimizer reads have to be read at the grain where the decision is made, which is door and size, not chain and style.

The in-season reforecast. The reforecast is how a markdown decision reaches the rest of the plan. When the optimizer recommends a path, its forecast of units, revenue and margin by week replaces the plan's for that style, and the sum of those replacements, by class, is the new in-season view of the season. How often to reforecast sets the cadence; the markdown system is one of its main inputs, and a brand running one without the other has a markdown decision that never updates the number the business is steering by.

Who uses it

Four roles touch a markdown system, and the design question in choosing one is which of them it is really built for.

  • The merchandise planner owns the planned reductions, runs the reforecast, and reads the markdown spend against the plan by class and month. For the planner the system is a budget-tracking and forecasting tool, and the output that matters is the margin and residual forecast by class.
  • The merchant or buyer owns the product and the exit route, and makes or confirms the decision per style in the trade meeting. For the merchant the system is a review list with a recommendation and a reason, and the output that matters is the path per style, the alternative paths and what each costs.
  • The pricing or promotions team, where one exists separately, owns the ladder, the event calendar and the channel parity rules, and sets the constraints. For them the system is a policy engine, and the output that matters is which constraints bound and how often.
  • Store operations and e-commerce execute the price: the re-ticketing in stores, the price change online. For them the system is a feed, and the output that matters is a clean price file per location per effective date, with the operational limits already respected.

The test of fit is which role's Monday the demonstration is built around. A system built for the pricing team shows a price grid; one built for the planner shows the spend against the plan; one built for the merchant shows the list of styles that need a decision this week, with the forecast and the budget beside each. In a brand of the size that reads this site, the merchant and the planner are often the same meeting, and the system has to serve both views of the same decision.

Suites vs point solutions vs a spreadsheet markdown calendar

Three kinds of thing answer to "markdown optimization software" in a selection, and the difference is mechanism rather than brand. A suite module shares its data model with the plan; a point solution shares an integration with it; a spreadsheet shares a planner's time with it.

Suite modulePoint solutionSpreadsheet markdown calendar
Where the position, the plan and the exit date come fromThe same data model the merchandise and assortment plans live inAn integration from the ERP, the planning system and the calendarA weekly export, keyed or pasted
Price-response modelUsually simpler; pooled by class; shares the plan's forecastUsually the product's core; style-level; its own forecastThe planner's judgement, or a fixed trigger table
Constraint handlingPlanned reductions enforced natively; channel rules shared with allocationConstraints entered and maintained in the tool; the budget read from the plan on a scheduleA column the planner watches
Feedback into the planThe reforecast and the open-to-buy update when the path is confirmedA file or an API call back to the planning systemRe-keyed, or not
Fits best whenThe brand's main markdown problem is a planning problem: depth, timing, budgetThe brand clears many styles per week across many locations and the price path is the main leverThe style count is in the low hundreds, the fleet is small and the cadence is slow
Fails whenThe model is too coarse for a brand whose styles respond very differently to priceThe integration is late or thin, so the optimizer runs on last week's position with no budgetAnything happens at the door or size level, or more than one person edits the file

The honest answer to "which?" depends on where the brand's markdown dollars actually come from, which the hindsight can show. If most of the season's markdown spend traces to depth bought beyond the style's history, a late delivery or a broken size run, the money is upstream and the right investment is the planning layer that sets depth, timing and the budget, with markdown handled as a disciplined cadence inside it. If the buys are sound and the spend traces to the price path itself, to steps taken too late, too deep or in the wrong doors, the money is in the optimizer. Where the hindsight shows some of each, the order in which they are fixed matters, because an optimizer pointed at a planning problem will produce very precise recommendations for how to pay for mistakes it cannot see.

What the software has to handle natively, by vertical

Every vertical marks down, and every vertical has a clock, a unit and an exit route that a system built only for apparel will get wrong. The question for each is what sets the exit date, what unit the price response has to be estimated on, and what the residual is actually worth. The markdown and exit strategy by vertical guide covers the tactics; this section covers what the system has to carry.

Apparel

Apparel has two populations and the system has to know which it is looking at. Fashion product has a finite window set by the floor set and the delivery that follows it, a base rate that decays, and an exit date the whole path has to reach; core and never-out-of-stock product is replenished, has no exit date, and is a candidate for promotion or for a permanent price reset but not for a clearance cadence. A system that runs fashion rules on core product will mark down a basic that was simply between replenishments. Within fashion, the size run is the unit of truth: a style-color at 40% sell-through in the chain can be sold out in the core sizes and carrying the whole residual in the fringe, in which case the right decision is a size-restricted markdown or a consolidation rather than a chain-wide step. The system has to hold sell-through, weeks of supply and the response estimate at size level, and the flags that say when a week's sales were capped by a broken run.

Footwear

Footwear's unit is the pair, its grid is size and width, and its central markdown problem is the broken size run. A style that has sold through in the core sizes and widths holds its residual in the tails, where demand was always thinner, and a chain-level markdown on it discounts the tails without creating a buyer for them. The system has to estimate response at the size level, recommend consolidation of broken runs into fewer doors or the online channel before any price move, and distinguish the exit for a carryover style, which continues into the next model year and may be a candidate for a price reset rather than a clearance, from the exit for a closeout, which is superseded by the new model year and has a hard date set by the changeover. The planning a model year changeover guide covers that date; the system has to read it.

Accessories and bags

Accessories run on two clocks. The seasonal colorway has an apparel-like window, and the evergreen core in the hero color has none and carries over. The system has to hold the two separately, so that a slow seasonal color is marked down against its exit date while the core colorway beside it, which is slow this month because it is between gift peaks, is not. Carryover is the vertical's defining decision: a bag that costs little to hold and will sell at full price next season may be worth carrying rather than clearing, and the system has to let the residual be valued at next season's full price, less the cost of holding it, as one of the exit routes, so the optimizer can compare carrying with cutting honestly. The planning accessories lines guide covers the two populations.

Home and furniture

Home and furniture carries three things an apparel markdown system does not. The first is the floor sample: the display unit that is sold off at the end of the program, often as the last step, at a price that reflects wear, and the system has to treat it as a distinct SKU with its own route rather than as stock. The second is container timing: receipts land in container quantities on ocean lead times, so a program's position jumps by a container at a time and a markdown taken the week before a container lands clears the old units and then meets a full new position at the reduced price. The system has to read the on-order and the in-transit as part of the position and schedule the path around the landings. The third is landed cost, which is the cost base the margin floor is set on and which moves with freight and duty; the planning margin on a moving cost base guide covers it, and the system has to hold the current landed cost per SKU, not the standard.

Outdoor

Outdoor's clock is the model year. A style is superseded on a changeover date set by the line, the dealer prebook has already committed the next year's units, and the current year's residual has to be out before the new one lands, through the owned channel, the outlet or the dealer floor. Two constraints are specific. MAP, the minimum advertised price, bounds the owned channel's public price for the current model year and, under many brands' policies, loosens or lifts when the year closes out, so the ladder changes by date, and the system has to hold a date-dependent floor. And the dealer channel carries a parity expectation: an owned-channel reduction below what dealers are selling at damages the relationship the prebook depends on, so the constraint is a cross-channel rule with a cost the objective has to count. Counter-seasonal categories mean the system is running a spring changeover for one half of the line while the other half is mid-season.

Sporting goods

Sporting goods ends its seasons by sport rather than by calendar, so the system has to carry an exit date per category rather than per season: the team sports close when the league calendar does, the snow categories when the resorts do, the water categories when the weather does, and the prebook for the next season has often been placed before the current one's residual is known. Launch pricing is a specific case: a new model launches at a protected price for a window and steps down on a schedule that is often set by the brand rather than by the retailer's response, and the system has to hold a path that is partly given. Equipment carries a replenishment cadence and a long tail of sizes and specifications where the broken-run logic from footwear applies at the SKU level.

Beauty

Beauty's exit clock is the expiry date and the lot's shelf-life dating, not the floor set; period-after-opening starts when the customer opens the unit and is not the retailer's boundary. A unit past its dating ceiling cannot be sold at any price, so the residual's exit value falls to zero on a date the system has to hold per batch, and markdown is the last lever rather than the first, behind transfer to a channel that sells faster and return to vendor. The unit of the markdown problem is the shade range: a foundation or lipstick line sells through unevenly by shade, the slow shades are the ones with the most dating risk, and a range-wide markdown discounts the shades that were selling. Gondola resets and retail partner calendars set when a shade is discontinued, which is the exit date for its stock, and gift sets have an apparel-like window around the gifting peak. The system has to hold dating per lot, response per shade and the partner's reset calendar.

Toys and games

Toys clear after the gifting peak, and the markdown problem is the weeks after it: the sell-through to the peak was the season, what is left after it has a short window before the next line arrives, and the base rate falls off a cliff rather than decaying. Licensed product adds a window set by the licence: a property whose film or series window has closed loses demand on a date the brand can see in advance, and the system has to hold the licensed window as an exit date and read whatever sell-off terms the licence sets, on timing and where it sets them on price, as constraints. The planning a licensed product window guide covers the window; the system has to carry it per SKU.

Baby and juvenile

Baby and juvenile hard goods have long lives and a demand that is driven by the registry: a car seat, a stroller or a crib is bought when a due date approaches, often from a registry built months earlier, and the registry is a forward demand signal the system can read: itemised, explicit about intent, and not a committed order. A markdown on a registered item changes the price at which months of visible, uncommitted demand will convert, and the vertical's regulatory exits, a standard that changes or a recall, remove an item from sale on a date that no price path can move. Soft goods and apparel in the vertical behave like apparel with a size run that is an age range. The system has to hold registry demand as a forward signal, as the planning with registry demand guide describes, and regulatory dates as hard exits.

Jewelry and watches

Jewelry's cost base moves with the price of the metal, so the margin floor is not a fixed number per SKU but a function of what the unit would cost to replace, and a piece bought at last year's gold price is carrying a different margin at today's. The system has to hold the metal content per piece and a current metal price, so the floor it enforces is the live one. Watches run on a model year and a dealer channel, with the same parity and MAP constraints as outdoor, and precious pieces have an exit route apparel lacks: the remount or the melt, where the residual's value is its metal content rather than its ticket. The merchandise planning for jewelry and watch brands guide covers the cost base; the system has to read it weekly.

A worked example: one style, two markdown paths, and the hold they are measured against

The arithmetic below is the first three decisions, timing, depth and cadence, done by hand for one style-color at one review, with the clearance exit valued so the paths compare honestly. Every figure is illustrative, chosen so the division comes out cleanly; none is a benchmark, and the price-response multipliers in particular are inputs a brand would estimate from its own history, not numbers to plan with. The point is not the result but the shape of the calculation, the way each constraint and each input moves the answer, and the fact that a markdown system has to redo all of it every week for every style.

The inputs:

  • A women's woven blouse, bought at 1,200 units for a 16-week selling window, at an $80 original retail and a $30 unit cost. Planned exit is the end of week 16, when the floor set changes.
  • It is the end of week 8. The style has sold 540 units at full price, a sell-through of 540 ÷ 1,200 = 45%. Full-price revenue to date is 540 × $80 = $43,200 and margin to date is 540 × $50 = $27,000. On hand is 1,200 − 540 = 660 units, with nothing on order.
  • The rate over the last four weeks has averaged 50 units a week, so weeks of supply at the current rate is 660 ÷ 50 = 13.2 weeks, against 8 weeks to the exit. At full price the style does not clear; that is the trigger.
  • The illustrative base-rate forecast at full price, carrying the decay of a seasonal style, is 44 units a week in weeks 9 to 12 and 36 a week in weeks 13 to 16.
  • The illustrative price-response inputs, as lift multipliers on the base rate: 1.5 at 25% off, 2.25 at 40% off and 2.75 at 50% off.
  • The brand's constraints: a price ladder of 25%, 40% and 50% off; no owned-channel price below the $30 cost; at most two steps, at least four weeks apart; and whatever is left at the end of week 16 goes to an off-price buyer at an illustrative $12 a unit.

The hold, for reference. If the price is held at $80 for all eight weeks, the forecast sells 44 × 4 + 36 × 4 = 176 + 144 = 320 units for 320 × $80 = $25,600, at a margin of 320 × $50 = $16,000. The residual at exit is 660 − 320 = 340 units, which the off-price buyer takes for 340 × $12 = $4,080, at a margin of 340 × ($12 − $30) = −$6,120. Margin from week 9 is $16,000 − $6,120 = $9,880, and the check is revenue less the cost of the 660 units: $25,600 + $4,080 − 660 × $30 = $29,680 − $19,800 = $9,880. The hold shows zero markdown and the worst result, which is the markdown percent trap the tactics guide describes.

Path A: 25% off in week 9, 40% off in week 13. The first step takes the price to $60, a $20 reduction, for weeks 9 to 12. Base 44 × 1.5 = 66 units a week, so 66 × 4 = 264 units, for 264 × $60 = $15,840 of revenue and 264 × ($60 − $30) = 264 × $30 = $7,920 of margin. The second step takes the price to $48, a $32 reduction from the original, for weeks 13 to 16. Base 36 × 2.25 = 81 units a week, so 81 × 4 = 324 units, for 324 × $48 = $15,552 of revenue and 324 × ($48 − $30) = 324 × $18 = $5,832 of margin. Units sold in the eight weeks are 264 + 324 = 588, so the residual is 660 − 588 = 72 units, sold off-price for 72 × $12 = $864 at a margin of 72 × (−$18) = −$1,296. Margin from week 9 is $7,920 + $5,832 − $1,296 = $12,456. The check: revenue is $15,840 + $15,552 + $864 = $32,256, less $19,800 of cost, is $12,456.

Path B: hold to week 12, then 50% off in week 13. Weeks 9 to 12 stay at $80: 44 × 4 = 176 units, for 176 × $80 = $14,080 of revenue and 176 × $50 = $8,800 of margin. Week 13 takes the price to $40, a $40 reduction, for weeks 13 to 16. Base 36 × 2.75 = 99 units a week, so 99 × 4 = 396 units, for 396 × $40 = $15,840 of revenue and 396 × ($40 − $30) = 396 × $10 = $3,960 of margin. Units sold are 176 + 396 = 572, so the residual is 660 − 572 = 88 units, sold off-price for 88 × $12 = $1,056 at a margin of 88 × (−$18) = −$1,584. Margin from week 9 is $8,800 + $3,960 − $1,584 = $11,176. The check: revenue is $14,080 + $15,840 + $1,056 = $30,976, less $19,800, is $11,176.

The result, and the check.

LineHoldPath A: 25% then 40%Path B: hold, then 50%
Units sold weeks 9 to 12176 at $80264 at $60176 at $80
Units sold weeks 13 to 16144 at $80324 at $48396 at $40
Units sold in owned channels320588572
Residual to off-price at $123407288
Units accounted for660660660
Owned-channel revenue weeks 9 to 16$25,600$31,392$29,920
Off-price revenue$4,080$864$1,056
Total revenue weeks 9 to 16$29,680$32,256$30,976
Cost of the 660 units$19,800$19,800$19,800
Margin from week 9$9,880$12,456$11,176
Markdown dollars taken in owned channels$0$15,648$15,840

On every path the units sold plus the residual sum to the 660 on hand: 320 + 340, 588 + 72 and 572 + 88 are each 660, and with the 540 already sold each path accounts for the 1,200 bought. Path A keeps $12,456 − $11,176 = $1,280 more margin than Path B and leaves 16 fewer units for the off-price buyer. Both paths beat the hold by a wide margin, $2,576 and $1,296 respectively, because the hold's zero markdown is paid for in a residual of 340 units sold below cost. The markdown dollars, the reduction per unit times the units sold at that reduction in owned channels, are 264 × $20 + 324 × $32 = $5,280 + $10,368 = $15,648 for Path A and 396 × $40 = $15,840 for Path B, on the point-of-sale convention the tactics guide uses; a brand that books a markdown on every unit on hand when the ticket changes, as the formula page's example does, records $17,952 and $19,360 instead, and the comparison between the paths is unchanged: the earlier, shallower cadence spent slightly fewer markdown dollars and kept more margin, because it sold more units into a larger base. Against full-season owned-channel net sales of $43,200 + $31,392 = $74,592 for Path A and $43,200 + $29,920 = $73,120 for Path B, the markdown percent is 21.0% and 21.7% respectively, and either figure is what the planner checks against the class's planned reductions for the month, which this example does not invent.

What the example shows about the system. Every number above depends on three inputs the software has to hold and a workbook usually does not: the decay in the base rate, without which the hold looks better than it is; the lift multipliers, which are estimates from history with error around them, so the real comparison between A and B is a comparison of distributions, not of two numbers $1,280 apart; and the off-price value of the residual, without which every path that holds price looks free. Change the exit route to a carry-forward valued at next season's full price less a holding cost, and Path B may win; change the second step's lift from 2.25 to 2.0 and Path A's second step sells 72 × 4 = 288 units instead of 324, its residual rises to 660 − 264 − 288 = 108, its margin from week 9 falls to $7,920 + 288 × $18 − 108 × $18 = $7,920 + $5,184 − $1,944 = $11,160, and Path B now wins by $11,176 − $11,160 = $16. The system's job is to make those sensitivities visible every week for every style, which is the whole case for the category.

Data requirements

A markdown optimizer is only as good as the history it reads, and the first work of an implementation is data rather than method. The four data properties that decide whether the model can estimate anything are price history by week, the promotion flag, the availability flag and an exit date per style. Without the first the model has no response to estimate; without the second it reads events as markdowns; without the third it reads stockouts as weak demand; without the fourth it has no boundary to optimize to.

  • Weekly sales by SKU by location at the price actually sold, for at least two comparable seasons, so that similar styles can be pooled and the current season's new styles have a group to inherit from. The grain has to reach size, or the broken-run flag cannot be built.
  • Permanent markdowns separated from temporary promotions in the price history, with the event calendar that explains each promotional week. A point-of-sale record carries a price, not the reason it was that price, so the separation is a data-preparation project before it is a modelling one.
  • Stockout and broken-run flags by SKU by location by week, derived from on-hand and from the size run's state, so that weeks capped by availability are excluded from the response estimate or adjusted for.
  • An exit date per style or classification, held in the planning calendar or the assortment plan, and maintained when the floor set, the delivery or the model year moves.
  • Original retail, current price and unit cost per SKU, with the cost base the margin floor is set on: landed cost where freight and duty are material, replacement cost where the metal price moves it.
  • On-hand, on-order and in-transit by SKU by location, read from the ERP and the WMS on a cadence at least as fast as the review cadence.
  • Planned reductions by class and month from the merchandise financial plan, read as a budget the recommendations are checked against, not as a report.
  • The constraints as data: the price ladder, the step limits, the channel parity rules, MAP where it applies and the wholesale price commitments by account and window.
  • Product attributes sufficient to pool: class, price tier, fabrication or material, lifecycle shape, licence or model year where relevant.

The data readiness for merchandise planning guide covers the general preparation; the markdown-specific additions are the promotion flag and the availability flag, and a brand that cannot produce both from its history should expect the first season of any optimizer to be the season in which it learns to.

Selection criteria and the questions to ask

The questions to ask planning software vendors guide and the apparel planning software selection guide cover the general evaluation: architecture, implementation model, integration, data model and commercial terms. The markdown-specific criteria follow from the mechanism, and the test for each is to bring one of the brand's own styles with its own history and watch the system handle it in the demonstration, rather than the vendor's.

  1. Can it hold an exit date per style as a hard constraint, and show what happens to the recommendation when the date moves? A system that treats the exit as a preference rather than a boundary will recommend paths that miss it.
  2. Does it separate promotions from permanent markdowns in the history it reads, and can it show which weeks it excluded or adjusted and why? Ask to see the response estimate for a style that was on promotion in two of its weeks.
  3. Does it carry stockout and broken-run flags, and does the estimate change when they are applied? Bring a style that sold out in the core sizes mid-window.
  4. At what grain is the response estimated, how are styles pooled, and how does a style's own history take over from its group's as it accumulates? Ask what the system does with a new style that behaves unlike its group after four weeks.
  5. Does it value the residual at its exit route, and can the route and its value be set per style? A system that does not value the residual will hold price too long.
  6. Does it read planned reductions by class and month from the plan, and does the recommendation change when a class is ahead of its budget? Ask to see the class view, not only the style view.
  7. Which constraints does it hold as data: price ladder, step limits, minimum interval, channel parity, MAP by date, wholesale commitments by account and window? Each one missing is a rule the planner enforces by hand.
  8. Does it record the recommendation, the decision taken and the reason, and can the hindsight read overrides against outcomes? A system that forgets its own recommendations cannot be evaluated after the season.
  9. How does a confirmed path reach the reforecast and the open-to-buy, and how fast? A path that lives only in the markdown tool is a decision the rest of the plan cannot see.
  10. How does the price reach the store and the site? Ask for the file or the feed, the effective-date handling and what happens when a store cannot re-ticket on the recommended day.
  11. Which of the vertical requirements above does it carry natively? Size and width grids, floor samples, dating per lot, model-year dates, licensed windows, registry demand, metal cost: a system that holds any of these as a note rather than a field will not enforce them.
  12. What does the first season look like? Ask how long the history has to be, what the system does before it has it, and how the team's existing trigger table is carried while the model earns trust.

Common failure modes

The failure modes below recur, and they are failures of position in the stack and of data rather than of method.

  • Optimizing a style the plan should never have bought that deep. The system produces a precise path for clearing 1,200 units of a style whose history supported 700, and the markdown dollars are attributed to the price decision rather than to the buy. The mechanism is that the optimizer sees the position as given; the fix is the hindsight that reads markdown spend against planned depth, and a planning layer that sets depth from the style's history before the optimizer ever sees it.
  • A markdown with no exit date. The style has no date it has to be gone by, so the optimizer has no boundary, holds price as long as the margin per unit is positive, and the residual meets the next delivery on the floor. The mechanism is a missing field; the fix is an exit date per style in the assortment plan, maintained when the calendar moves.
  • Ignoring size-run brokenness. The chain-level sell-through says the style is slow, the core sizes are sold out in half the doors, and the markdown discounts the fringe sizes without creating a buyer for them while the doors with the core sizes sell them at the reduced price unnecessarily. The mechanism is estimating at the wrong grain; the fix is size-level position and flags, and consolidation as a lever the system can recommend before price.
  • Store-by-store optimization with no replenishment logic. The system recommends a different price in every door for a style that is still being replenished from the DC, so a door marked down this week receives full-price stock next week and the chain's price position becomes unreadable. The mechanism is optimizing locations independently of the inventory flow; the fix is reading the allocation and replenishment state, and differentiating by group rather than door where the flow still runs.
  • Reading promotions as markdowns. The response estimate is inflated by event weeks, the recommended depths are too shallow, and the style misses its exit. The fix is the promotion flag, built before the model is trusted.
  • Spending the class's budget one style at a time. Each recommendation is optimal for its style and the class is through its planned reductions by the second week of the month. The fix is the class view and the budget read as a constraint.
  • An optimizer that cannot be overridden, or one whose overrides are not recorded. The first is abandoned within a season because the merchant's knowledge has nowhere to go; the second cannot be evaluated because nobody can say whether the system or the overrides produced the result.
  • Chasing the markdown percent. A team measured on markdown rate holds price, shows a low markdown percent and loses the most margin at the exit, as the hold in the worked example does. The fix is measuring margin dollars kept and residual at exit alongside the rate, which the tactics guide sets out.

Where RetailNorthstar fits

RetailNorthstar is a merchandise planning platform, not a dedicated markdown optimizer; a system that estimates price response per style and searches price paths under constraints, as this guide describes, is a different system category, and a brand that needs one should evaluate it with the criteria above. Where the two meet is the plan on either side of the markdown decision. RetailNorthstar's platform page describes seasonal open-to-buy modelled with receipt planning, markdown reserves and carry-forward inventory, which is the planned-reductions side of this guide: the budget a markdown decision is spent against, held with the buy rather than discovered at the end of the season. Its market and performance intelligence page describes in-season sell-through tracking at the style, color and size level from all sales channels, SKU performance scoring on velocity, margin contribution and weeks of supply, AI-assisted anomaly detection in sell-through, inventory or margin performance, and margin scenario modelling under different pricing, markdown and promotional scenarios, which is the review side: the position and the rate a markdown trigger is read from, and the margin consequence of a candidate path.

The evaluation criteria in this guide apply to RetailNorthstar's planning side as they apply to any product in its category, and a brand should run them against it: bring a style with its history, its exit date and its planned depth, and see where the plan sets the budget and where the review surfaces the trigger. Whether the brand's markdown dollars are upstream, in depth, timing and budget, or in the price path itself is the question this guide is meant to help settle before any demonstration, this one included.

Related resources

See how RetailNorthstar keeps the open-to-buy, the receipt plan and the markdown reserve in one connected plan, with in-season sell-through at the style, color and size level surfaced at the point of decision.

Start Free Trial →

No credit card. No commitments.

Common questions

What is markdown optimization software?

Markdown optimization software is the system that decides when, how deep and where to reduce the price of each product so that it sells through by its exit date at the highest achievable margin, by forecasting price response per style and location, scheduling the markdown cadence and sizing the markdown budget against the plan's reductions. It is the system of record for the price-reduction decision: the exit date each style has to clear by, the forecast of how each style sells at each candidate price, the constraints the business places on the price path, and the markdown dollars every reduction consumes against the planned reductions in the merchandise financial plan.

What decisions does markdown optimization software make?

Six decisions recur in every markdown system and in every markdown workbook one replaces: when a style takes its first markdown; how deep each step goes; the cadence of the steps from the first reduction to the exit; whether the reduction differs by store, zone or channel; the clearance exit for whatever price alone will not clear; and whether a reduction is a temporary promotion or a permanent markdown. The software recommends each decision per style and location from a price-response forecast run against an exit date, a margin floor and the other constraints the business sets, and a planner confirms or overrides it.

How does markdown optimization software work?

Mechanically it has two parts. A price-response model estimates, for each style or group of similar styles, how weekly unit sales change at each candidate price, separating the price effect from the natural decay of a seasonal style's rate, seasonality, promotions and weeks where availability capped sales. An optimizer then searches the candidate price paths from now to the exit date and chooses the one that maximizes an objective, usually margin dollars including the value of any residual at its exit route, subject to constraints: the exit date, a margin floor, the price ladder, the maximum number of steps and the minimum interval between them, channel parity rules and wholesale commitments. The constraints are where the brand's judgement enters the calculation.

What is the difference between markdown optimization software and dynamic pricing software?

Markdown optimization moves the price of a product downward, permanently, on a schedule, to clear a finite quantity by a fixed date; dynamic pricing moves price in either direction, repeatedly, in response to demand, competition or inventory signals, on products that are replenished rather than cleared. The two use a similar price-response estimate and different objectives: markdown optimization maximizes margin on a position that has to reach zero or a planned residual by the exit; dynamic pricing maximizes margin or revenue on a position that is refilled. Seasonal product with an exit date is a markdown problem, and continuity product that is replenished is where dynamic pricing applies.

What data does markdown optimization software need?

Weekly sales by style, color and size by location at each price the product actually sold at, with permanent markdowns separated from temporary promotions; on-hand, on-order and in-transit inventory by location; the original retail, the cost and the current price per SKU; an exit date per style or classification, from the planning calendar; the promotion and event calendar; stockout and broken-run flags, so weeks capped by availability are not read as weak demand; product attributes and several seasons of history, so new styles can be pooled with similar ones; the planned reductions by class and month from the merchandise financial plan; and the price ladder, step limits and channel rules the business sets as constraints.

How does the worked example in this guide compare two markdown paths?

One style bought at 1,200 units at an $80 retail and a $30 cost has sold 540 units at full price by week 8 of a 16-week window, a 45% sell-through, with 660 units on hand and 13.2 weeks of supply against 8 weeks left. Path A takes 25% off in week 9 and 40% off in week 13 and sells 588 units for $31,392, leaving 72 units for off-price; Path B holds full price to week 12 and takes 50% off in week 13, selling 572 units for $29,920 and leaving 88. With the residual sold off-price at $12 a unit, Path A's margin from week 9 is $12,456 and Path B's is $11,176, so the earlier, shallower cadence keeps $1,280 more margin and clears 16 more units. Every price-response figure in the example is an illustrative input, not a benchmark.

Is RetailNorthstar markdown optimization software?

No. RetailNorthstar is a merchandise planning platform; a dedicated markdown optimizer that estimates price response per style and searches price paths under constraints is a different system category. Where the two meet is the plan: RetailNorthstar's platform page describes seasonal open-to-buy modelled with receipt planning, markdown reserves and carry-forward inventory, and its intelligence page describes in-season sell-through tracking at the style, color and size level, SKU performance scoring on velocity, margin contribution and weeks of supply, and margin scenario modelling under different pricing, markdown and promotional scenarios. That is the planning-side work this guide describes as the upstream and downstream of a markdown decision, not the optimizer itself.

RetailNorthstar Editorial Team
RetailNorthstar ·

Share this guide with your team

Copy a link or a pre-written message for Slack, Teams, or email.

// Know where your operation stands

Apply this to your planning operation.

The free Apparel Planning Maturity Assessment benchmarks your operation and tells you exactly which gaps to fix first.

Take the assessment →

Apply these insights with RetailNorthstar.

See how modern apparel brands use RetailNorthstar to put this planning framework into practice.

No credit card. No commitments.

Connected merchandise planning — live in weeks, not quarters.