Markdown_Optimization_Chapter_ by stariya

VIEWS: 9 PAGES: 21

									Markdown Optimization Chapter – DRAFT AS OF 9-4-08 WITH SCF INPUTS &
EDITS


Markdowns – A New Problem?

“Retail is a simple business!” Thus is the position of most career merchants, executives
who have cut their business teeth on the buying of goods in the wholesale market for
ultimate resale, via retail stores, to consumers. This often daily affirmation is rooted in a
merchandising philosophy dating back to the early department store days at the turn of
the century and best articulated by Marshall Field, proprietor of Marshall Field’s
Department Store in Chicago – “Just give the lady what she wants!”


Retail, from its inception, was characterized by a close personal relationship between
merchant and customer. Most often the merchant would literally procure in the wholesale
market goods he knew his customer wanted to buy from him at retail. This level of
intimacy, facilitated by the proximity of the merchant to the end consumer, minimized
the need for guess-work, or, as we sometimes call it today, forecasting.


Yet even in these early days of formalized, yet local, retail store operations,
merchandising challenges existing which could lead to markdowns. In his review of this
era, Leon Harris writes in “Merchant Princes – An Intimate History of Jewish Families
who Built Great Department Stores” that Edward Filene, young proprietor of Filene’s in
Boston, was particularly concerned about the vagaries of demand for fashion
merchandise and the resulting markdown liability. Harris writes:


       “Edward was bewildered by the terrifying variety of merchandise offered for sale
       in New York, and even more by the inexplicable and sudden shifts of fashion that
       destroyed the value of merchandise in stores. He promptly determined that he
       would somehow devise scientific methods of merchandising so as to decrease the
       terrible risks these presented to the retailer. This determination would continue
       throughout his life and would long cause him to be an object of scorn among
       merchants who were smugly certain that success in merchandising depended not


                                              1
       upon science but upon such inexplicable mysteries as superior taste and a
       prescient nose.”


Scale versus Intimacy


Interestingly, every major retail operation one can think of today – be it Wal-Mart, Ann
Taylor, The Gap, or Staples – started out running its business much like E.A. Filene.
There was a store, a merchandising “point of view,” a set of customers within close
geographic proximity to the store, and a merchant whose job it was to procure goods in
which his customers expressed explicit interest. And the best merchants, particularly of
apparel, were those who also innately understood not just the explicit interests of his
customer but also her more latent desires. It was these needs that could be translated into
high priced “fashion” oriented purchases at a high gross margin.


And inevitably, as these retail operations on a small local scale became successful, their
management realized that they could easily replicate their success by satisfying similar
needs of similar consumers in other geographies via additional stores. In so doing, not
only would they grow their top-line revenue in relation to the number of stores they
opened, but they could also leverage many fixed costs of operations – the buying,
merchandising, accounting and other back-office functions that didn’t have to be scaled
linearly with the growth in stores. Importantly, merchants also found they suddently had
increased negotiating leverage with their vendors as they were procuring goods in greater
and greater quantities to support the broader base of stores. The result of this approach
was greatly increased average per store profitability as each store benefited from the
economies of scale in both purchasing and operations.


Inevitably, however, these same successful chains ultimately face an unexpected dis-
economy of scale where per-store performance begins to decline as the base of stores
grows. In some chains this happens after 50 stores, in others after 500, and in still others
after much greater scale. The cause isn’t a decline in the power of scale of course – it’s
instead a result of the loss of customer intimacy inherent in the move from small-scale



                                              2
local operation to large-scale regional or national chain. Where once a merchant was off
“to market” to buy dresses and gowns and shoes for Mrs. Cavanaugh and Mrs. Smith,
knowing full well both the tastes and checkbook capacity of each, now that same
merchant is buying in massive quantities for women “like” Mrs. Cavanaugh and Mrs.
Smith all over the country. And while certainly some customers will feel that the gowns
are so perfect they must have been procured just for them, many, inevitably, will find the
items are not quite right.


This loss of intimacy is what leads to a mismatch between supply of merchandise and
demand for that merchandise and what gives rise to the need for special pricing practices
to help stimulate otherwise tepid demand.


Contemporary Retail – Why Pricing is Hard


Contemporary retail is thus marked by the pursuit of increasing economies of scale.
Leveraging buying power and logistical networks across hundred or thousands of stores
has proven benefits in terms of direct costs, efficiency, and branding – witness the
success of Walmart. Starting with the buy from suppliers, decision-making is
concentrated in the center rather than in stores on the periphery. Centralization of pricing
decisions naturally results from a centralized buying organization. Large retailers (1000+
stores) tend to create merchant organizations managed via individual gross margin
targets. The merchant will buy for the entire chain, set initial prices, and then manage
promotional prices and subsequent markdowns in order to maximize gross margin for her
patch of square footage within every store in the chain. She does not trust her pricing
decisions to local store managers.


By taking on the pricing decision at the center a pricing diseconomy of scale can often
result. Consider the challenges that a merchant faces if she is determining prices for a
thousand stores versus just one store. The merchant-customer relationship is
depersonalized and abstracted by physical disconnection. The customer for a particular
store is perhaps barely glimpsed by the merchant on an annual store visit. Particular



                                             3
customer preferences or store-specific idiosyncrasies are now contained in the stream of
sales and inventory data fed to headquarters from the store’s POS and inventory systems.
Pricing is now done at a remove. The merchant cannot walk the store and directly
observe stock levels, what is selling, and what is driving sales in terms of promotions,
placement, and presentation. Scale is achieved at the price of customer intimacy.
Regaining it is not simple.


The burden of customer intimacy has been shifted to the IT systems that track and control
the inventory flow. Store-level differences can be lost in the sea of information generated
daily from 1000+ stores. A store’s sales for a particular item are a direct result of local
climate, demographics and preferences of customers living nearby. Without an approach
to discern and disentangle these factors from the data aggregated at the center to then
recombine into reasonable forecasts, large chains are often prone to the problem of the
wrong stuff at the wrong place at the wrong time.


Fashion merchandise adds another level of customer-intimacy difficulty. The fashion
item is, by definition, ephemeral. It did not exist last season and will not last beyond the
current season. Observations of discount-driven demand lifts are often only visible at the
critical first markdown or promotion. There is little opportunity to learn and recover from
mistakes. The paucity of direct item history results in less precise, inferential forecasting
based on the history of similar items from previous years. Not surprisingly, forecast error
is typically 2 or 3 times greater for fashion items than for basics. Fashion demand is
unpredictable, by defintion. Couple this with the loss of visibility to store-level
idiosyncrasies from the center and getting the correct price can be quite a challenge.


The Return of Customer Intimacy through Systems


The relentless, exponential reduction in computing power costs (Moore’s Law) over the
past 40 years has transformed retail and offers potential solutions to the pricing challenge.
Computerized inventory control systems have rapidly advanced in terms of speed,
accuracy, and depth. Data collection starts at the POS. Every customer transaction at a



                                              4
cash register is regularly captured and stored in a database. The costs and complexities of
storing terabytes of POS data are no longer only available to retailers the size of Walmart.
Advances in database software make aggregation, queries, and reporting against very
large datasets feasible. A slew of applications has been developed to provide effective,
central management of critical retail business processes. Automated replenishment
software is adopted widely within the industry and has created increasing efficiencies in
terms of inventory reduction and fewer excessive inventory build-ups in the retail supply
chain. Replenishment software focuses on basic merchandise and illustrates how
effectively customer intimacy issues can be overcome with the accurate, relevant history
of particular store/items. The replenishment system reacts to store-level demand “pulls”
via accurate short-term demand forecasts at the store/item level. Proper deployment of an
automated replenishment system can often reduce safety stocks by up to 20% (NEED TO
VALIDATE THIS NUMBER – I DIDN’T ACTUALLY LOOK AT THE ARTICLE)
(Marshall Fisher and Ananth Raman, "Rocket Science Retailing is Almost Here - Are You Ready?" Harvard
Business Review, July/August 2000.)   without any increase in unplanned stock-outs resulting in
potential lost sales. Planning systems have also been widely adopted. These systems help
to distribute inventory-purchasing budgets (buys) over time, location, and merchandise
categories. These systems offer powerful, top-down spreading and budget tracking
functionality, which can reliably plan for 1000’s of stores and 100,000’s of sku’s.
Allocation applications in widespread use demonstrate the power of automated decision
systems handling the problem of pushing inventory out to stores in anticipation of
demand. A typical approach is to treat the 1000 stores in a hypothetical chain as a
collection of average stores each representing a cluster or particular tier of stores
exhibiting common demand levels. Thus a 1000-store chain can be effectively
transformed into a 10-store chain by setting the application to 10 store clusters. Proper
ranking and clustering is a science in itself (see also Darrell K. Rigby and Vijay
Vishwanath, “Localization, The Revolution in Consumer Markets” Harvard Business
Review, 2006); but, done correctly, it will provide a valid approximation of store-level
differences. Rules determining how much should be pushed to each cluster can then be
set. Flexibility in creating and modifying these store groupings along with varying
allocation rules are common features. In apparel, another important dimension against



                                                  5
which stores may exhibit greatly variable sales demand, and thus require unique
clustering, is size. Work with apparel retailers such as Ann Taylor, Macys and American
Eagle has shown that demand for certain sizes of the same items varies greatly by
geography and store type. East and West coast stores often skew toward demand for
smaller sizes while Midwest stores tend see great demand for the larger sizes of an item.
And anomalies often occur – a chain of youth oriented clothing identified high demand
for smaller sizes in a cluster of Midwest stores consistently demonstrating larger size
demand. Further investigation revealed that this particular store was situation on a main
street of a college town and thus served customers with very different demographic
characteristics than the balance of its stores in that region. Effective clustering, taking
into consideration such anomalies which can be revealed through appropriate analysis of
size selling history, will allow for stores in regions or demographics where customers
tend to buy larger or smaller sizes versus the chain average to be clustered together to
receive appropriately sized apparel. The net result of the right sizes in the right stores, of
course, is a closer match between supply and demand and fewer stockouts and
markdowns.


Current Retail Pricing Approaches


Today, cutting edge information systems in retail go well beyond simple forecasting
techniques, top-down spreading, and rules-based pushes to complex analytical algorithms
handling sophisticated forecasting and optimization challenges. Much of this practice was
first pioneered in the academic discipline of operations research with notable initial
successes in industries outside of retail. Optimized seat pricing within the airline industry
was pioneered by SABRE in the early 90’s and demonstrated the benefits of automated,
computerized pricing of perishable inventory (seats) offered in thousands of local
varieties (flights) per day. Whereas retail handles the final links of the supply chain
ending in the consumer, much of the pioneering OR work around inventory management
focused on the beginning of the supply chain with factory floor inventory optimization
applications such as MRP and MRE. Producer forecasts and inventory optimization
began its widespread adoption in the 80’s with a dramatic reduction in overall inventory



                                               6
levels in supplier supply chains. Retail remained the last frontier given the enormous
SKU/location intensity, but given advances in software, driven (as noted earlier) by the
shrinking costs of computing, the complexities of retail optimization from the center are
now tractable.

Markdowns for apparel oriented merchants have risen dramatically over the past forty
years. The National Retail Federation reports that average department store markdown
levels which hovered below 10% in the 1970’s rose exponentially through the 80’s and
90’s to north of 30% by 1995. Interestingly, in the past few years, markdown levels,
while still high, have begun to stabilize. Recent advances in the analytical approaches and
technology supporting markdown pricing are at the heart of this phenomenon.


To best understand the most recent science of markdown pricing, a natural starting point
is to look at how initial prices are set. The standard approach relies on cost plus logic.
One starts with the unit cost and sets the price to attain a desired and hopefully attainable
gross margin. The measures used capture desired versus attained gross margins in terms
of Initial Markup and Maintained Markup.
                 Initial Markup = (Initial Retail – Cost)/Cost
Initial Markup (IMU) can be thought of as the maximum Gross Margin % attainable for
the product if no subsequent price discounts (promotions or markdowns) occur after
introduction of the product.
Maintained Markup tells us the actual Gross Margin % that was achieved after
discounting.
                 Maintained Markup = (Actual Selling Price – Cost)/Cost
With simple algebra and a cost input, gross margin expectations or targets can be
translated into prices. Thus many retailers think of initial prices in terms of gross
margins. Costco famously has set a constant low gross margin for all its merchandise so
initial pricing is a simple result of the Initial Markup definition. Other retailers vary gross
margins by a variety of factors including merchandise category (consumer packaged
goods versus apparel, for example) , consumer segment (affluent versus lower income)
and service level (high touch versus self-service) and so derive their initial prices.



                                               7
A critical challenge to overall gross margin is how to best set initial prices for loss
leaders. Setting zero or negative IMU for a particular item can be profit-maximizing in an
overall sense if it drives the purchase of a basket of related high IMU items or increases
general customer traffic enough to compensate for the loss leader. The calculation of
market basket impacts and cross elasticities is at the frontier of analytical feasibility given
the tremendous amount of potential cross-correlations -- imagine a 100,000 x 100,000
correlation matrix of SKU-level demand. Approaches relying on category-level cross
correlations (e.g. shoes and dresses versus size 4 red velvet Jimmie Choo pumps and size
2 Chloe evening gown) have been offered to cope with the complexity. Adoption of
advanced analytics for loss leader initial pricing is low and the majority of retail relies on
experiential intuition based on trial and error and common, time-tested industry practices.
Probably the most common approach is to handle loss leaders via promotions, which we
will discuss in more detail in the following sections.


Customer psychology is another important consideration in setting initial prices.
Kahneman pioneered the field of behaviorial economics and efficient pricing by noting
consumer biases such as irrational price point rounding and price anchoring. (see MRL on
this?) Studies have shown (need reference) that the common 99-cent price ending
exploits a subconscious rounding error where the last digits are essentially rounded down
to the dollar. This allows the retailer to gain the extra 99 cents of gross margin without
the rationally expected demand suppression. Anchoring prices at a high IMU is also
worthwhile. For many retailers, especially in apparel, the bulk of the inventory is sold at
subsequent markdowns to the initial price. The ability to proclaim markdowns of 50% off
and deeper while maintaining healthy MMU’s allows the retailer to profit from the
consumer anchored to the high initial price and leverage the consumer perception of a
“great deal” due to the high percentage discount. Efficient pricing techniques let the
retailer advantageously tweak the price sensitivities of their customers.

In many ways Every Day Low Pricing (EDLP) is a response to all of the noise efficient
pricing coupled with continuous discounting consciously presents to the consumer. Not



                                              8
all customers enjoy the challenge of chasing down a bargain. Walmart pioneered the
simple and consistent concept that prices would always be low for all items they carried.
Consumers were spared the effort of finding deals and were satisfied that any item bought
at a Walmart was a “good deal.” The brand message was powerful and the competitive
advantage derived from simply passing on to their customers their industry-leading cost
advantages and supplier discounts was formidable. Increased scale drove costs down,
which lowered prices, and that in turn drove growth and subsequently increased scale.
The virtuous circle eventually resulted in the largest retailer in the industry. Many
retailers followed in Walmart’s wake such as Costco and The Home Depot. Despite the
enormous growth of this strategy in the past 30 years, limits are apparent. The competing
pricing strategy of high/low has not been supplanted and is alive and well.


High/Low pricing can be defined as an attempt to reach multiple customer tiers in terms
of price sensitivity. The clearest examples lie within Geffen goods, named for Sir Robert
Geffen, a British economist in the second half of the 19th century who defined the
paradox of certain items for which demand increases, irrationally, as the price rises – the
most well-known cases being fine art or expensive wines. In this case, price elasticity is
the reverse of the conventional wisdom that the lower the price the greater the demand;
instead, the higher the price, the greater the demand. Price is perceived as a signal of
quality and thus enhances demand. Starbucks is a prime example of first degree price
discrimination where high prices for a readily available, inexpensive commodity
enhances demand. An expensive cup of high-quality coffee becomes an emblem of
affordable luxury. The Starbucks experience of baristas, couches, and espresso machines
can be summarized as a successful attempt to support IMU’s several times higher than
traditional competitors. Fashion apparel pricing is greatly influenced by the same inverse
price elasticity. High IMU can be seen as a fair charge for finding the size you need when
the assortments are first introduced. It is interesting to note that apparel is the one general
category where Walmart has consistently struggled (could go into a couple of more
sentences on this interesting situation), underlining limitations to the general applicability
of EDLP. The Low side of Hi/Low opens up an even greater opportunity for potential
demand. Many customers do, after all, enjoy the thrill of a good bargain and will respond



                                               9
dramatically to aggressive price discounting. Crazy Eddie was famous for “insane” prices
implying that the bargain-hunting consumer could take “advantage” of the retailer. Many
high/low retailers have segmented consumer demand, which activates in stages as a range
of diminishing prices is explored over a product’s lifecycle. Hi/Low especially lends
itself to short-lifecycle fashion merchandise. The limited time the particular item is
offered allows little opportunity for competitive activity and repeated price changes to
find the stable price equilibrium seen in many basics stocking the big box EDLP retailers.


Another pricing factor to be considered is scarcity pricing. Seasonal merchandise often
takes advantage of inventory scarcity relative to demand by timing high IMU’s to capture
initial demand peaks. Air conditioners in the summer or snow shovels in the winter are
priced as such. High IMU’s at peak demand coupled with aggressive discounting off-
peak is usually the most effective approach to optimize gross margin for seasonal
merchandise and is a common driver of high/low pricing practice.


After time and merchandise, the remaining price variation dimension is location. Tiered
pricing is the practice of offering different prices to different markets for the same
merchandise. Given variations in climate and demographics one would imagine tiered
pricing to be a common practice, yet its adoption is limited mostly to local promotions
and store-driven final clearance markdowns. Many retailers still adhere to national
promotions and markdowns for their major pricing moves. Inhibitors to adoption include:
fear of negative reactions (witness Amazon’s recent foray and immediate retreat from
tiered pricing) from customers who cross-shop between stores, operational challenges in
execution of different prices and signage in different stores, and, most importantly, the
problem of determining store-level price variations from the center. Some pricing
variation by store is obvious, such as locations with high tourist traffic. Tourists usually
have only one chance to shop a store and have little opportunity to time visits or
purchasing decisions around discounting as much as a regular, local shopper. In general,
most retailers find surprising homogeneity to price response despite a broad geographic
footprint. A sale in an affluent suburb in Atlanta may drive a similar lift in sales as one
held in an affluent suburb in Chicago. Determining location-specific price elasticity is



                                             10
extremely difficult given the increase in noise around the demand signal at store-level
resolutions. Only the most advanced analytical teams have successfully surmounted this
challenge. Demand seasonality is another opportunity for geographic price variation, but
even here opportunities are somewhat limited. The major seasonal traffic drivers are
national holidays and have little geographic variance. Climate does vary geographically;
but, from a pricing perspective, suffers from the unpredictability of local weather
conditions. At the end of a product’s lifecycle, the benefits for differentiated pricing by
store become more obvious as some locations begin to stock-out while others are stuck
with relatively large pools of inventory. This problem is exacerbated by sub optimal
allocations of merchandise to begin with. The ability to discount deeply if a store/item’s
end-of-season weeks of supply are high, and not discount if low, will create additional
gross margin opportunity. In fact, most retailers, even hardcore EDLP ones, allow store
managers to take over pricing at the very end of an item’s lifecycle allowing them to
optimally clear local inventory imbalances.




Markdowns versus Promotions


An important consideration to highlight is the distinction between markdowns and
promotions. Markdowns are often called clearance pricing implying that the markdown’s
primary function is to clear out remaining inventory, whereas promotions are associated
with driving traffic into stores. A markdown is defined as a permanent price discount
unlike a promotion, which is a temporary discount. In retail accounting (which values
inventory at current retail ticket price instead of cost), a permanent price cut has a
dramatic impact. For example, a 20% markdown is experienced as an immediate charge
or write-down of 20% against all remaining inventory. The accounting measures the
amount of potential revenue lost once a permanent price cut is taken. For major
department stores, according to the National Retail Federation, overall markdown rates
average greater than 30% so the accounting charges are enormous. In contrast, a
promotion is accounted for as reduced sales as the units are sold. Unsold inventory is not
written down since the price will rise again in the future once the promotional period,



                                              11
defined by the retailer, is complete. Given all the uncertainties around how well an item
will sell upon introduction, limiting the potential revenue upside via an early markdown
should be approached cautiously. Instead, promotions are typically used early in the
season to drive demand when inventory levels are high, or to further increase traffic
during peak demand periods such as holidays. Promotions also offer a quick response to a
developing inventory imbalance. In contrast, markdowns tend to be used later in the
season. It often seems rational to the merchant to defer markdowns since the cost of a
premature markdown, which is irreversible and results in an immediate accounting
charge to earnings, would seem to outweigh the cost of a tardy one. A second deeper
markdown can follow immediately if the first was not deep enough. An overly deep
initial markdown cannot be undone by raising the price again. The balancing act lies in
not deferring until too late into the season where demand no longer supports a healthy
sales-increase response to a markdown. This tension between prematurely destroying
revenue potential versus waiting beyond the point of markdown efficacy is the
fundamental challenge of the markdown decision. Actual analysis of more than 50 major
retail data sets conducted by the analytical teams at ProfitLogic (now Oracle Retail), a
leading retail price optimization business, reveald the opposite, however – that most
retailers wait far too long to take their initial markdowns and in so doing, end up
requiring far deeper discounts later in the product lifecycle to stimulate demand and clear
the inventory. The result is sub optimal gross profit.


Inventory clearance is as important as GM% attainment; or, put another way, inventory
turnover drives overall gross margin dollars as much as GM%. Old inventory in the
stores blocks the flow of new, fresher inventory with higher revenue and margin
potential. Store square footage productivity will suffer. The most reliable turn accelerator
at a retailer’s disposal is discounting to liquidate at the end of the season. A host of
practices has arisen to ensure the flow of merchandise does not clog. Apparel vendors
typically provide markdown dollar support to department stores to encourage the timely
execution of markdowns and re-order of newer items, helping to keep their merchandise
fresh on the selling floor. An ecosystem of off-price retailers will buy leftover, end-of-
season merchandise and resell through their channels. Jobbers and off-price retailers



                                              12
(some extremely large such as TJX) all exist to profitably take unsold apparel
merchandise off the hands of department stores and apparel manufacturers. Many
retailers have created their own separate channels in the form of outlet stores situated far
from their mainline stores to drain off old inventory at extremely discounted prices. Some
retailers enforce clearance discipline by putting an internal expiration date on the
inventory via a “penny markdown”. At a preset date (the outdate) all remaining inventory
will be marked down to a penny and written off. Merchants need to sell before time is up
or literally lose the inventory at a complete loss. Sample sales and employee sales are
other methods commonly used to stimulate demand to clear out the aged inventory.


Given all of the considerations mentioned above, there are several basic and commonly
used approaches for setting retail prices, especially markdowns. The simplest approach
from an operational perspective is to set initial prices and then the timing and amount of
subsequent discounts as part of the pre-season planning process. Knowing a fixed pricing
schedule from the beginning of the season allows for precise store operational planning
for signage and merchandise placement. As an example, prior to its implementation of
markdown optimization, a leading children’s apparel retailer would follow a rigorous
schedule of markdowns for each new “floor set” (a new collection of merchandise that
arrived on the selling floor every four weeks). The merchandise in a floor set would all be
markdowned to 25% off after 3 weeks on the floor, then moved to the middle of the store
and marked down to 50% off after 5 weeks and to the back of the store and 70% off after
8 weeks. The obvious problem in this approach is is disconnection from actual demand in
season and the associated difficulty of forecasting in-season demand pre-season. Fashion
items are especially difficult given that they are subject to the caprices of consumer taste
season to season. The fixed approach allows for no reaction to growing inventory
imbalances and the prices execute regardless of the variance of the pre-season demand
forecast from in-season actuals. The other end of the spectrum is to put pricing into the
hands of local store managers. JC Penney and Nordstrom’s both followed this practice
and, tellingly, have evolved away from it in the past 10 years. The approach should
benefit from sensitivity to local inventory imbalances and a better sense of local demand.
It appears to be a direct way to address the diminishing customer intimacy problem



                                             13
discussed previously. In practice, store manager pricing control is very difficult to
execute well. The demand variability across merchandise typically far outweighs
geographic variability. Thus a better result can be achieved by having merchandise
specialists (buyers and planners at headquarters who have responsibility for a particular
category of merchandise across the chain) making pricing decisions rather than what
amounts to geographic specialists (store managers) who are in touch with local market
conditions but unlikely to be as familiar with the margin structure and demand
characteristics of each item. In other words, given the tradeoff, greater accuracy appears
to be achieved by averaging across stores instead of averaging across merchandise. The
ideal might be to operate in the old-school fashion of local personnel with deep
understanding of the local customer base and merchandise making buying and pricing
decisions. This approach requires highly skilled personnel in each and every store, thus
pressuring an already fragile retail economic model and challenging labor pool. Once a
retailer capitalized on the benefits of scale through centralization, customer intimacy
cannot be cost-effectively regained by moving decision-making back to the periphery.


Consolidating pricing decisions geographically implies a fine-grain along the
merchandise and time dimensions. (We will revisit the geographic dimension, but for
now we will focus on chain-level pricing.) The challenge at the center is how to respond
to the unpredictable weekly demand revealed over the course of a season for a particular
item. Rules-based systems are a reasonable approach. A typical system bases the
discounting decision on the current weeks of supply (WOS) projection versus weeks to
outdate (WTO), the date by which all merchandise should be sold. The more WOS
exceeds WTO, the larger the called-for discount. The rules capture the commonsense
logic that inventory build-ups need to be dealt with deeper discounts and reacts in season
to unexpected imbalances. Despite these virtues, rules-based systems can suffer from
both over-simplification and excessive complexity. The focus tends to be on getting the
rules right – “Is it 50% off at 6 WOS or 5 WOS?” – while both the forecasting and
optimization aspects of discount price setting are simplistic. Typically the forecast is a
moving average of the last four weeks of sales and prices are set to a fixed schedule of
perhaps three or four possible markdown percent discounts (e.g. 20% off, 30% off, 50%



                                             14
off and 70% off). The limitation of rules-based strategies is that the complexity of pricing
constraints and optimization rules will directly increase the complexity of the system.
Imagine parsing a rules-based system with hundreds of nested “IF THEN” statements.
(Think of an example.)


Modern Markdown Optimization Systems


Markdown optimization takes the virtues of rules-based systems and more elegantly
generalizes by using advanced OR and forecasting techniques pioneered in airlines and
manufacturing pricing systems to avoid the limitations. A causal sales forecast is
typically generated every week taking into account merchandise and location specific
price elasticities and seasonalities. This is compared against current inventory to calculate
if inventory will clear at current prices by the targeted outdate. If not, the minimum
discount (gross margin optimal) is calculated to clear by the outdate. A flexible
framework of business rules is incorporated in which the optimizer will only search the
space of “legal” price/week combinations. Complexities such as markdown free weeks
(e.g. during heavily promotional holidays), complex price ladders with numerous price
points, or minimum weekly spacing of markdowns can all be taken into account without
increasing the complexity of the underlying system. Legal pricing constraints such as
(CALIFORNAI LAWS AROUND PRICE CHANGES HAVING TO BE
PERMANENT?…??????….) can also be handled. The rapid gross margin pay-offs from
deploying Markdown Optimization systems are unmatched in the industry and recent
adoption has been dramatic. Oracle Retail, a leading vendor of markdown optimization
systems, reports over forty major retailers using such solutions with improvements in
gross margin dollar performance of 5%-15%. Public announcements just in the past few
yeas include Loehmann’s, Wal-Mart, Jones Apparel, Goody’s Family Clothing, The
Children’s Place, Famous Footwear, and Talbots. To put this level of benefit in
perspective, take a $1 Billion specialty apparel retailer like Ann Taylor. If its average
maintained gross margin is 40%, that implies $400 million in gross profit dollars each
year. A 5% improvement through more timely and appropriately priced markdowns
would result in $20 millions of increased gross profit annually. This level of



                                             15
improvement should more than pay for any associated systems and business process
related work required for implementation.


Markdown optimization systems are powerful and flexible enough that they are
increasingly used successfully for store-level pricing (list retailers??? Store-level pricing
is a big topic in itself – actually quite a few pros and cons). They have proven to be the
only viable approach to handle the third dimension of fine-grain geography in pricing
decision-making. The results indicate optimization systems are the best path for large
retailers to restore the benefits of merchant/customer intimacy enjoyed by small local
retailers.


Implementation Considerations


Since 2000, a sufficient number of retail operations have deployed markdown
optimization to create a body of learning and best practices that inform today’s successful
implementations. While ProfitLogic was the pioneer of these approaches dating back to
the late 1990’s, today many vendors offer solutions, including Oracle Retail (who
acquired ProfitLogic in 2005), SAP, JDA, and SAS. Often specialized retail focused
consulting organizations are hired to manage both the technology implementation and,
more importantly in this case, the business process and change management efforts
required for success.


Across the industry, general merchandise and apparel retailers have gained great
experience attempting to deploy markdown optimization solutions and change business
practices to capitalize on the opportunity of more efficient markdown pricing. Among
the most important lessons gleaned from on-the-ground experience in the implementation
of markdown optimization systems is that senior level sponsorship in the merchandising
organization is an absolute prerequisite for success. Markdown optimization is not an
Information Systems department project. While it involves deployment of a system,
actual training on use of the new system is trivial, often requiring no more than a half-day
tutorial session. More important is philosophical change required by merchants and



                                             16
merchandise planners, the managers in a retail organization who have been responsible
historically for making markdown decisions. Often these individuals have vast
experience buying and managing pricing for their category of merchandise. They believe
intuitively, through their hard fought experience and success that they can recognize the
signals associated with growing, or declining, consumer demand for their merchandise.
And often, because they are the same individuals who chose the merchandise in the first
place and believed in its merits, they, like any humans, let emotion play a role in their
decision making. The result is typically less objectivity in the markdown decision that
might otherwise be the case. Leon Harris, in his book “Merchant Princes”, described the
situation as it occurred to Edward Filene in the early 1900’s aptly:


       “One of the few certainties in retailing is that some of the merchandise bought
       enthusiastically in the wholesale market, where it looks eminently saleable, will,
       when it reaches the store, prove stubbornly unsaleable. But despite this
       inevitability, buyers, like the rest of humanity, are reluctant to admit and address
       their errors. They find endless excuses for the slow sale of merchandise: the
       weather's still too hot; the weather's still too cold. Easter's late this year; Easter
       was early this year. It hasn't been advertised; the ad was lousy; the ad ran on the
       wrong day; the ad ran in the wrong newspaper.


       The danger in these rationalizations is that they usually increase markdowns. A
       coat that does not sell early in the fall may still tempt customers if it is marked
       down early by as little as 25 percent. But as the season ends, it must be marked
       down far more drastically to tempt a customer who, having done without a new
       coat so long, may otherwise reasonably decide to wait until the next season.”


This phenomenon was true in 1910 when Edward Filene was running his family stores in
Boston, and it has remained true throughout the last century of evolution on the retail
landscape.




                                              17
The implication is that real behavior change is required for success. Merchants must
essentially hand-off the decision making on the timing and depth of markdowns to what
amounts to a “black box.” While such decision support has proven successful in a variety
of other industries, retail, and fashion retail in particular, has the unique distinction of
being populated by people who are attracted, not surprisingly, to fashion! More precisely,
the merchandising part of the retail organization (as opposed to the logistics, finance or
store operations organizations) is made up more of left brain, creative thinkers than right
brain, analytical thinkers. The result tends to be gut-feel intuition oriented decision
making which works reasonably well when choosing fashion for the coming season,
creating merchandise assortments, or designing branding and marketing campaigns. This
approach breaks down, however with more analytically oriented tasks such as optimizing
markdown pricing. As a result, adoption of decision support tools built on a basis of
relatively complex mathematical models tends to be inconsistent unless supported by top-
down mandate and well thought out new business processes. Without these, full adoption
by merchants and planners is unlikely, yet experience suggests that full adoption is
critical in order to achieve the full level benefit available. The following case studies
from retailers who have worked with ProfitLogic and Karabus Consulting, a leading
retail I/T and strategy consulting organization, illustrate the importance of
implementation planning and change management in driving successful results.


Case Study Examples1:

A leading fashion retailer, with over 700 stores, acquired a markdown optimization
solution. Realizing that success required more than just providing a new tool to support
existing habits, they retained a change management consulting organization for the
design of the new processes that would maximize the value of the markdown
optimization solution, the definition of new organizational roles, and the personalized one
on one coaching and training of buying and planning team members to ensure the
promised ROI would be realized. The effort began with an assessment of the retailer’s
current processes related both directly and indirectly to markdowns to identify
opportunities to implement best practices prior to starting the markdown optimization
1
    See Karabus Management, www.karabus.com, “Case Studies”


                                                18
implementation. As the technology was configured to incorporate the retailer’s business
rules, the implementation team ensured that the retailer’s unique requirements
were met with creative solutions and designed explicit future state processes for the
pricing analysts, buyers and planners who would be making the analytic-driven zone
pricing decisions. The team then managed the change to Markdown Optimization
through the launch process, providing comprehensive training on both
process and technology and live coaching of hands-on users and executive stakeholders
to establish markdown optimization as a key component of in-season merchandise
management.


The result, as described by this retailer, was that this was the most successful technology
implementation they had ever experienced. Not only was the project completed on time
and on budget, but it was also an extremely effective exercise in change management. It
involved shifting an entire organization to move beyond its historic merchandising
processes - where decisions relied on first hand knowledge of every style and every store-
to new processes that leverage attributes, forecast analytics and powerful technology to
complement the buyers’ passion for fashion and free them up from manual spreadsheet
analysis.


In another example, after many quarters of declining sales and profits, a multi-billion
dollar national retailer needed nothing short of transformational change to reverse
downward trends. As part of a multi-year project to transform their merchandising
practices, markdown optimization was selected as the first priority to realize rapid ROI
and help fund longer term initiatives. In three months, the retailer implemented an interim
solution - rules based markdowns - to drive almost immediate benefit, while planning the
markdown optimization initiative. The consulting organization in this case played a lead
role in the project, spanning markdown strategy, business process, organizational
changes, testing, and training while leveraging their deep expertise in many previous
projects. Key implementation activities included:
•Introducing Rules-Based Markdowns as a means of rapidly realizing benefits, several
months in advance of the rollout of markdown optimization



                                            19
•Defining Future State operating processes, accountabilities and timelines for the
introduction of performance-driven markdowns
•Developing a pricing zone strategy for markdowns that reflected different consumer
demand by zone
•Creating and leading business-focused training covering markdown concepts, business
process changes, application usage, and the markdown decision process
•One-on-one coaching with Planners, Buyers, and executives during initial markdown
cycles to ensure proper use and adoption of the new markdown approach


As a result of these efforts, the project has been a resounding success:
•Business strategy, not budget, drives acceptance or rejection of markdown
recommendations
•Well over 95% of firm Markdowns were accepted through MDO in the first season
•Gross Profit results exceeded plan by a significant margin
•Sell-through also increased significantly




                                             20
21

								
To top