Size Optimization for Retailers | White Paper by SAS

WHITE PAPER
Size Optimization for Retailers
How to effectively stock sizes of each item in each store – into
a profitable advantage

i
SIZE OPTIMIZATION FOR RETAILERS
Table of Contents
Executive summary .................................................................................... 1
The right merchandise in the right sizes to meet demand......................... 2
Size is important – so why has it been so difficult to get it right?............ 2
SAS® solves this problem for retailers today.............................................. 4
The pack optimization component – sending the right units to stores .... 5
Size profiling in action ................................................................................ 6
Conclusion ................................................................................................... 7

ii
The content providers for this white paper include: Elaine Markey and Greg Wilson,
SAS Retail Solution Strategy; Elizabeth Dove, SAS U.S. Retail Business Unit;
and Alexi Sarnevitz and Diana McHenry, SAS® Global Retail Practice and
Darius Baer, Global Analytical Services. For more information, please contact
Diana McHenry at Diana.McHenry@sas.com.

Executive summary
In a perfect world, retailers would always have just the right assortment of
merchandise – in the ideal mix of sizes – at each store. However, that has been
difficult to achieve on a consistent basis for assortments with tens of thousands
of stock keeping units (SKUs) across thousands of stores. Traditional methods
that rely on the use of “top down” chain averages have been overwhelmed by this
complexity. Even if detailed size profiles were available, there remained the question
of how to integrate this knowledge into assortment and allocation decisions.
“No one needs to pinch retailers and tell them this is a problem,” writes Susan
Reda, Executive Editor of STORES magazine. “They’ve heard the groans coming
from the dressing room and the complaints from shoppers who wonder why
there never seem to be enough sizes 0 and 2 and 14 and 16. The question is,
what to do about it?”1
SAS, the leader in retail intelligence, has incorporated its analytical strength into
scalable business solutions that allow retailers to address this question. SAS®
Size Profiling identifies store-level selling patterns across size ranges within any
classification of merchandise. SAS®
Pack Optimization identifies the best pack
or combination of packs to efficiently fill stores’ size demands. Through open
interfaces, this information can be integrated into existing planning, allocation
and replenishment systems to produce assortments that more accurately reflect
consumer size demand at the store level.
The potential benefit is enormous. Reducing out-of-stocks by just 1 percent of
revenues will add US$100 million in sales to a $10 billion-a-year retailer. That
figure doesn’t even include the cost of marking down overstock items, or the
labor cost of handling frequent allocations and ad hoc transfers of merchandise.
Longer-term benefits from an improved customer experience can be expected.
1
■ Maximize assortment productivity by
ensuring that store shelves have the
right merchandise in the right sizes
to meet customer demand. Reduce
stock-outs and end-of-season
markdowns due to size.
1Reda, Susan. “Sizing Up Sizing.” STORES (January 2006).

The right merchandise in the right sizes to meet demand
A customer, eager to buy the dress she saw in the window display, approaches a
fixture full of the dresses. There’s the right style, the right color; but her size
is missing.
The dress she wants – in her size – is on the rack in another store where that size
isn’t selling well. Right merchandise, wrong store. A potential customer walks away
disappointed, while the dress gets marked down later at the other store.
This scenario is frustrating for consumers and costly for retailers. For the retailer, it
brings two negatives: a forfeited sale and a markdown on the same item. According
to the National Retail Federation, stock-outs amount to losing 1.5 weeks of store
volume each year. That’s nearly 3 percent of sales. For retailers, that translates into
millions in lost revenue. For a $1 billion retailer, that can mean tens of millions in
lost revenue.
But it’s also costly to make every sale if you can’t be sure which sizes will be in
demand. You could overstock stores, but you’d pay too much for inventory and shelf
space – and end up marking down more merchandise at season’s end. You could
send allocations to stores more frequently or handpick individual sizes from bulk
packs as needed, but the processing costs would erode profit margins. You could
transfer the dress to the shopper’s preferred store, but that’s a time-consuming and
expensive proposition.
Size is important – so why has it been so difficult to
get it right?
Clearly, the answer is to have just the right assortment of merchandise – in the right
mix of sizes – at each store. That has been difficult, for several reasons:
An inability to process all the data: If a retailer starts amassing information at the
granular level of store and size, data volumes can quickly overwhelm the systems
that are supposed to make sense of it. Many systems used today to perform
rudimentary allocation by size simply cannot scale to manage the huge volumes
of data required.
An inability to consider variables over time and across products: Even if
the system could digest all the required data, does it have the analytic strength to
optimize the answer – not just automate it? Determining how to allocate the right
assortment – in the right sizes, in the right place, at the right time – is a complex
optimization problem that must factor in many costs and constraints.
2

There is no way to account for demand variances among stores: Even if the
retailer knows how many S, M, L and XL items were sold of a given color/style,
actual sales will not be uniform across all stores in the chain. Size demand will vary
by store. For instance, one store might sell 10 percent small, 40 percent medium, 40
percent large and 10 percent extra-large of an item. For another store with different
customer demographics, the distribution could be 30-40-20-10 or 5-30-50-15. If the
retailer shipped the average ratio of sizes in pre-packs to all stores, very few stores
would be appropriately served.
An inability to address multiple packs: Many retailers suffer from logistical
limitations. Systems, as well as distribution centers, are unable to consider and
process multiple pack configurations. Consequently retailers use generic pre-packs
and ship too many or too few of a size to stores.
Figure 1. Percent unit sales by cluster: Each line represents the size demand of an
item for a cluster of similar stores. The green line represents the average for all stores.
If the retailer allocates sizes based on the average, stores would be either under- or
overstocked on some sizes at some point.
There is no standardized methodology: Within many retail organizations,
analysts use their own approaches to try to create demand curves and apply them
to orders and allocations. They often use different data and processes, which are
typically not shared across departments. It is common for two analysts, even within
one department, to produce drastically different results for the same item. Which one
is credible?
It is no longer enough to allocate merchandise based on past regional performance
with one distribution ratio for size and color across all stores. Retailers must fully
understand demand variance – store by store and merchandise to merchandise – to
reduce stock-outs, markdowns and lost sales.
3
��
��
��
��
��
��
��
��
��
��
��
��
��
��
��
��
��
��
�� V� ��

How do you get the right fit – that is, the right sizes to stores?
Forward-thinking retailers are applying sophisticated analytics all the way to the
level of size and accounting for variances between stores. In the process, they are
generating, analyzing and acting on much more refined information than in the past.
These retailers are incorporating analytical methodologies more commonly seen in
industrial and financial arenas, such as data mining to identify sales patterns and
outlying anomalies, and complex algorithms to group (“cluster”) stores according to
size demand. The result is a fairly sophisticated answer to the problem of distributing
right-sized merchandise to stores that achieves the right balance between size profile
accuracy and process complexity.
SAS® solves this problem for retailers today
SAS Size Profiling and SAS Pack Optimization give retailers solid answers to the
perennial question of how to order and allocate just the right sizes of merchandise for
every store.
The size profiling component: determining which sizes sell – store by store
The SAS solution creates very accurate profiles of size distributions by merchandise
category by store (the foundation for more accurate ordering and allocations)
due to its ability to process huge volumes of data and balance many variables
simultaneously.
Size specifications can vary significantly even within a single department. A high
degree of variability of consumer size demand can exist across local store trading
areas. The SAS solution is flexible and customizable to reflect these realities. A
retailer defines its own optimization framework – the merchandise and stores,
variables, and constraints that reflect its unique operations. It is not constrained to
industrywide assumptions or across-the-board practices.
The behind-the-scenes logic that optimizes these decisions is highly sophisticated,
but the complexity is invisible to users. A quantitative analyst with business acumen
would establish the data management processes, analytical models and critical
relationships. Then, business users need only indicate the high-level variables – such
as which stores and how many units of a color/style – and then point the size
profiling system to that problem. The system returns optimized answers to support
highly accurate purchase orders, allocation plans, replenishment plans or reorders.
The SAS solution detects and addresses conditions that would erode the accuracy
of analysis, such as out-of-stock conditions or inaccurate/missing data from stores.
So, in the real world, where data collection processes might be imperfect, retailers
will have confidence in the accuracy of results with SAS.
■ The use of predictive analytics to
forecast shopper demand has grown
rapidly in the last five years and
incorporates more variables
across the supply chain, such
as outlet profiling, promotion
uptake propensity, and now:
size optimization.
4

At the end of the day, every store has an accurate and realistic size distribution profile
– either derived from actual sales data (accounting for all the above variables) or from
correlations with stores that do have reliable sales data. This size profile can then be
applied to create accurate purchase orders, or against allocation plans to determine
the best size distributions to stores.
The pack optimization component – sending the right
units to stores
SAS Pack Optimization identifies the best combinations of packing configurations
(such as pre-packs, case packs and bulk packs) to best fill likely size demand in
stores. This process draws on updated size profiles and factors in a host of variables
and constraints, such as seasonality, price, promotions, inventory levels and
minimum shipments.
Using the SAS Size Profiling solution, size profiles are created for each store
for specific categories of merchandise. These size profiles are applied against
assortment plans to generate accurate purchase orders, and now the merchandise
arrives at the distribution center in prepackaged cartons. Let’s assume that each
carton contains a standard ratio of sizes. Perhaps a pack of 100 men’s polo shirts
contains 10 small, 20 medium, 50 large, and 20 extra-large, packaged in sets of 10.
Given the standard pack configurations – and knowing how many of each size you
need at each store – what is the best way to distribute these packs to stores? The
calculation can be complex. Aligning product pre-packs with demand requires an
understanding of a broad set of logistical costs and constraints.
The SAS solution provides the serious analytics and processing power to answer this
question with confidence, even in cases where the vendor ships multiple standard
pack configurations, or where the retailer can choose to open packs or handpick
individual items from bulk.
The process starts by running the latest size profile for each store against the new
inventory. After all, it has probably been several months since the purchase order
was placed. Updated sales information might reveal new patterns that could affect
distribution. Or, the vendor may have shipped a different quantity than what
was ordered.
Then, the SAS solution evaluates all the constraints and variables that have been
incorporated into its logic and returns an optimized allocation plan for distributing
packs to stores. Should whole packs be sent? How many to which stores? Is it
worth it to break open cartons and send packs of 10? Is it worth it to break open
bulk packs and ship individual units? The pack optimization component answers
these questions while considering a host of interdependent variables.
n SAS®
Size Profiling is built on the
SAS®
9 architecture, a foundation
of data management and analytic
technology that can scale to handle
databases of any size, even those
containing massive amounts of
retail data. A powerful solution that
leverages the business intelligence
platform underlying SAS®
Retail
Intelligence Solutions, SAS®
9
provides an integrated, open and
extensible foundation for creating
and delivering accurate, in-depth
retail intelligence.
5

In short, the SAS solution solves a complex optimization problem where the
objective is to balance conflicting variables. The solution will minimize handling costs
associated with flowing the product through the distribution center, maximize sales
and minimize markdowns at the stores, and consider any number of costs
and constraints.
Size profiling and pack optimization: not just for size
The SAS® solutions are called SAS Size Profiling and SAS Pack Optimization, but
they are not limited to size.
“How many basketballs should we order in each color?” “How should we distribute
cases of red and white wines among restaurants in the chain?” If the attribute can be
quantified and characterized in a distribution, it can be optimized.
Similarly, size is not the only attribute for which you can optimize pack allocations. A
wholesaler, factory or vendor – any organization that ships merchandise in multiples
to multiple locations – could find the solution useful to optimize those shipments.
“If I can buy toasters in their own boxes, or in cartons with four toasters apiece,
or in large cases of 12 toasters – and there’s a discount on the larger cases
– which packaging choice is my best bet?” “The Chilean fruit comes in boxes of 10
packages, but only a few stores will sell that much. Should we break open boxes
and send individual packages, or will the processing costs be higher than the cost of
some spoilage?”
Wherever many interdependent or conflicting variables must be considered, the SAS
optimization engine can deliver effective and profitable answers
Size profiling in action
Kohl’s uses the SAS® solution to develop at the size level
Kohl’s is a family-focused, value-oriented specialty department store with some 800
stores nationwide. The fast-growing retailer plans to open 500 more stores in the
next five years. With billions in revenue at stake, Kohl’s wanted to take every measure
to improve in-stocks by size for seasonal basics and fashion merchandise. A major
obstacle to this was that existing systems didn’t provide sufficient visibility into the
dynamics and details of demand at the size level. Analysis criteria were loaded into
spreadsheets and manually executed – a cumbersome process. Kohl’s needed more
accurate representations of what each store was likely to sell in each size for each
type of merchandise.
6


Working with SAS, Kohl’s selected high-volume, size-intensive departments to run
as test cases in a size profiling solution. They created size profiles for representative
stores based on sales history, and applied these size profiles to assortment and
allocation plans. They then measured the impact on sales and margin over product
lifecycle compared to a control group.
Their findings were remarkable. Size distribution varied more significantly than
expected among stores, among departments within a single store, and even among
classifications or categories within a single department. The opportunity associated
with these differences was “financially significant,” according to Jon Nordeen,
Executive Vice President for Planning and Allocation.
For example, one cluster of stores showed a propensity for selling small sizes for a
particular category. These stores required 60 percent more small merchandise than
stores in the “extra-large” cluster. Stores that had a propensity for selling extra-large
sizes required 27 percent more extra-large items than stores in the “small” cluster.
Imagine the lost-opportunity cost of so great a mismatch if all stores had received
one standard distribution
of sizes.
If Kohl’s had sent products to all stores based on the average, very few stores would
get the right mix of sizes.
Convinced of the merits of size profiling, Kohl’s worked with SAS to establish clusters
of stores that shared the same size-selling characteristics and to develop hundreds
of merchandise-specific size profiles for each cluster of stores. These size profiles
have been integrated into processes and systems to optimize assortment planning,
ordering and allocation.
Kohl’s results are more satisfied customers, higher sell-through, and fewer markdowns.
Conclusion
The scenario seems so simple: Make sure consumers can find the size they want.
If their sizes are out of stock, you forfeit the sale and disappoint (and potentially
lose) the customer. Marking down overstocked merchandise to move it undermines
profitability.
Here is where SAS analytical strength can be invaluable. SAS Size Profiling and SAS
Pack Optimization effectively match size-level demand with pack-level supply – for
each store and merchandise category – improving the accuracy of purchase and
allocation decisions based on a multitude of defined variables. This type of analysis
has the potential to transform a difficult issue – how to effectively stock sizes of each
item in each store – into a profitable advantage.
SAS® Analytics will give Kohl’s
an accurate view of consumer
preferences by location – information
that can be integrated into business
systems and processes.
“SAS Size Optimization will be
used by Kohl’s to define our order
quantities by size, create multiple
pre-pack configurations and allocate
specific size quantities to each store.
Underlying each of those processes
is a customer demand profile by size
and category of business for each
store.”
Jon Nordeen
Executive Vice President, Planning
and Allocation
Kohl’s Department Stores
In “Sizing Up Sizing,” STORES
magazine, January 2006

SAS Institute Inc. World Headquarters +1 919 677 8000 To contact your local SAS office, please visit: www.sas.com/offices
SAS and all other SAS Institute Inc. product or service names are registered trademarks or trademarks of SAS Institute Inc. in the USA and other countries. ® indicates USA registration.
Other brand and product names are trademarks of their respective companies. Copyright © 2007, SAS Institute Inc. All rights reserved. 102868_ 421931.1007

Size Optimization for Retailers | White Paper by SAS

More Related Content

What's hot (19)

Similar to Size Optimization for Retailers | White Paper by SAS (20)

More from Louis Fernandes (13)

Recently uploaded (20)

Size Optimization for Retailers | White Paper by SAS