Posts Tagged ‘clustering’

Hardlines Optimization—Part 1: Clustering is Not Localization

by Quantum Retail • August 17th, 2011 • Write a comment »

There was a time when it was enough to have individual store managers make the decisions of which products to carry and how much to order. As hardlines retailers have grown, their offerings have become more elaborate and competition has become more sophisticated. Customers and their behavior have become more complex as well. It’s no longer possible for people to keep track of all the variations and trends making each location unique.

For the last two or more decades, conventional wisdom has been that clustering stores is the way to address this problem. While, conceptually, grouping stores with similar characteristics makes sense – the reality is that it’s very difficult if not impossible to do this in a consistent and effective manner. From volume clusters to attribute clusters and even more recently devised mechanisms for deriving “intelligent” clusters using BI tools, all suffer from flaws that keep them from achieving the elusive goal of localization. Let’s look at some of these to understand why that is.

Inherent problems with volume based clusters:

The most widely used of these decades-old methods to cluster stores is volume clustering. Typically using historical sales for a group of merchandise presumed to be similar then finding apparently logical breaks as that volume is ranked high to low. The breaks represent boundaries which define which cluster stores fall into. Most commonly the result is a set of 10 or fewer cluster groups (although we’ve witnessed organizations that do many more). The three most common flaws in this process are:

  1. The clusters are virtually always based on historical sales
  2. The clusters are not updated frequently enough
  3. The clusters don’t incorporate any attributes of the merchandise or the stores

Typically a retailer looks at a store’s performance over the last year or compares like for like seasons to group stores that performed similarly together. The problem is that usually its only historical sales that are used to group the stores together. This creates a bit of a self-fulfilling prophecy. If a store underperformed in a department because of an inordinate amount of out-of-stocks, could the store have performed better?

The better option is to use either historical demand, or better yet, a demand forecast. When we refer to demand we look for what would have happened had the store been in stock. This allows us to understand any missed opportunity that may have resulted from out of stock situations. Of course using a true, forward looking demand forecast which includes understanding those missed historical opportunities to cluster would be the best of both worlds. Not only are you incorporating lost sales, but you would be incorporating future trends of store behavior as well.

Another problem is timing of when clusters are used and how often they are evaluated. Store assignment to clusters should be evaluated as often as possible. Customers change their shopping behavior constantly and this changes the behavior of individual stores constantly. As a result, store clusters need to be re-evaluated as frequently as possible using the most current understanding of behavior. At a minimum cluster should be updated prior to each merchandising activity that is consuming them. For example, updates should occur pre-season when an assortment plan is created (which products in which locations), again when a buy plan is made (soft commitments for how much), again when the actual order is placed (time has passed, things have changed), again when commitments to DC shipping locations (DC splits) are confirmed, again if the buy is pre-distributed to commit stock to stores, another when the goods hit the DC, and so on.

Furthermore, clusters should never be used to drive replenishment decisions. Once the goods have hit the stores, any re-allocation/replenishment activities should be based on actual, individual item/store behavior. There is simply no good reason to use clusters after the first allocation – and even then a sound argument can be made as to whether they should be used at all. This is especially true for hardlines and commodity products that don’t have the volatility or size complexity seen in hardlines. Today’s technology is capable of identifying and leveraging the uniqueness of individual products in individual stores, so each store can be treated as it needs to be. This is where true localization can truly begin to provide huge benefits.

Clustering beyond volume

Retailers use more than just volume clusters but as of yet no standards have been adopted. Climate, an attribute of physical location, is one of the most commonly seen attribute clustering criteria. While useful, categorizing a location as one of three or four climate groupings is an inexact science at best. And there are surely many other attributes of location and product that can refine the quality of the result. In traditional execution, however, more attributes leads to more groups needing to be managed. These groups are often nested into artificial hierarchies and quickly become difficult to navigate and virtually useless when complexity reaches the point that execution is impossible.

More recently many have tried to utilize Business Intelligence solutions and statistical analysis to find more refined groupings of stores that behave similarly. One of the most common flaws of this approach is that the resulting clusters have no definition of what makes them similar. While it may be accurate that they have similar behavior, that conclusion alone is useless unless a merchant understands what makes a group unique. Without that there is nothing available to guide a cluster specific decision.

Where does that leave us?

Ultimately, newer technologies that have been focused on solving these problems in retail have refined the utilization of clusters significantly. Better solutions constantly review and update clusters based on current behavior and the processes that consume clusters are able to accept and modify their conclusions accordingly and without unnecessary user intervention. They analyze and update activity across a variety of attributes proven to impact the products and locations within the grouping and offer flexibility to navigate across those without being tied to unmanageable hierarchical relationships. They also derive learning about individual products to the point that the need for clusters is either significantly reduced or eliminated in many processes throughout the management of inventory (such as replenishment). If your processes are not supporting that level of locating management and practical clustering, it’s simply not possible to achieve what is expected when discussing localization in retail.

Consider the following as you think about the quality of your clustering practices:

  1. Do you cluster only on volume or on other attributes or KPIs too?
  2. Do you cluster only on historical sales or do you incorporate missed opportunities / lost sales?
  3. Do you cluster based on a forecast of demand for the periods you’re using the clusters to represent?
  4. Do you re-evaluate your cluster assignments as often as possible?
  5. Do you re-cluster in-season?
  6. Do you cluster the stores as low on the merchandise hierarchy as is reasonable for your business?
  7. If you change cluster definitions can the systems that use them accept that change cleanly?
  8. Do you drive re-allocations or replenishment by clusters?

If the answer to any of the above questions is “no”, there is room for some self-evaluation and improvement to your localization strategy.

Look out for next week’s blog on SKU Rationalization.

Sign up to receive updates throughout this series.

2 people like this post.
Unlike

Fashion Forward – Part 1: Clustering is not localization

by Quantum Retail • May 25th, 2011 • 5 Comments »

It has been quite a while since your typical fashion retailer was able to merchandise their stores individually. It has long been common practice to group stores together based on some commonality. On the surface this makes practical sense. One merchant or allocator can’t handle hundreds of stores themselves, so grouping the stores together is a logical way to deal with the problem of location proliferation.

The problem is thinking that clustering via methods developed in the 1980s and perpetuated heavily today is the best way to localize a chain’s assortment. It simply is not. It’s high time we reevaluated clustering to see where improvements can be made. Limitations of yesterday’s merchandising systems were a legitimate reason to use simple clustering to manage the complexity. However, today’s systems are much more sophisticated and need not cripple a merchant with unnecessary constraints like before.

The most common way to cluster today consists of clustering stores based on volume. Typically, this volume is sales volume and the number of clusters varies, but basically is what the retailer is able to handle. The average number of clusters that a fashion retailer uses falls in the 6-8 range, but you could find many that use significantly more.

Unfortunately, volume based clusters have three inherent problems:

  1. The clusters are almost always based on historical sales
  2. The clusters are not updated frequently enough
  3. The clusters don’t incorporate any attributes of the merchandise or the stores

Typically a retailer looks at a store’s performance over the last year or compares like for like seasons to group stores that performed similarly together. The problem is that typically only historical sales are used to group the stores together. This creates a bit of a self-fulfilling prophecy. If a store underperformed in a department because of an inordinate amount of out-of-stocks, could the store have performed better?

The better option is to use either historical demand or a demand forecast. Demand incorporates what would have happened had the store been in stock. Of course, using a true demand forecast which includes understanding those missed historical opportunities to cluster would be the best of both worlds. Not only would you be incorporating lost sales, but you would be incorporating future trends of store behavior as well.

Another problem is when clusters are used and how often they are evaluated. Store assignment to clusters should be evaluated as often as possible. Stores should be clustered pre-season when a buy plan is made, again when the order is placed, another when the buy is pre-allocated, another when the goods hit the DC, and so on. Sometimes this isn’t practical, such as direct to store prepacking for cross-docking, but in the absence of store commitment, stores should be re-clustered. This increases the data used for clustering and makes it more current and more accurate. Furthermore, clusters should never be used in the replenishment process. Once the goods have hit the stores, any re-allocation/replenishment activities should not consider clusters. Store behavior is too erratic, especially in fashion, to rely on clusters in-season. There is simply no good reason to use clusters after the first allocation and even then it’s arguable whether they should be used at all. Today’s technology is capable of identifying and leveraging the uniqueness of individual products in individual stores, so each store can be treated as it needs to be. This is where true localization can really provide huge benefits. Refining a store’s size allotment, replenishing better sellers and, as importantly, not replenishing items that are destined for markdowns, should be store specific decisions.

There are certainly a lot of retailers using clusters that are more sophisticated than simply volume, but it is nowhere near universal adoption. Climate is by far the second most common store attribute on which stores are clustered. But are there other attributes of stores and/or merchandise that would refine the localization strategies better? I would argue both yes and no. Pre-season, it would be beneficial to add some attributes such as price point (merchandise), economic status (store), fashionability (merchandise) or similar attributes to the buy decision to better assort the stores based on consumer behavior. On the other hand, I’ve stated earlier that in-season clusters are useless. They only result in normalizing out store uniqueness. It matters not why a store is selling an item well, it only matters that it is!

I challenge you to question your clustering cadence and ask yourself these questions:

  • Do you cluster only on volume or on other attributes or KPIs too?
  • Do you cluster only on historical sales or do you incorporate historical or forecasted demand?
  • Do you re-evaluate your cluster assignments as often as possible?
  • Do you cluster in-season?
  • Do you cluster the stores as low on the merchandise hierarchy as is reasonable for your business?
  • Do you drive re-allocations or replenishment by clusters?

If the answer to any of the above questions is “no”, there is room for some self-evaluation and improvement to your localization strategy.

Look out for next week’s blog on optimizing size/packs by Dr. Ziad Nejmeldeen, VP of Science, Quantum Retail.

Sign up to receive updates throughout this series.

1 person likes this post.
Unlike

Busting the Myths of Retail: #4. Clustering and Localization

by Quantum Retail • November 24th, 2010 • 1 Comment »

MYTH #4: I can localize my merchandise by clustering based on store attributes

TRUTH: Grouping stores together seasonally based on store size, geography or sales performance will only deliver average results at best. Retailers can hone the localization of their stores with intelligent clustering at lower levels of merchandise including merchandise attributes and the actual current performance of products.

Clustering stores that performed in similar ways still perpetuates the problems associated to the use of aggregated data. Instead of pretending you have, say, 1,000 stores that all perform alike, you now pretend that in each of maybe 10 different store groups, you now have around 100 homogeneous stores. The reality we know is that every store performs differently. Though it is not realistic to manage the creation of an assortment plan for each of your stores, you can begin to step in the direction of localization by creating merchandise specific clusters that group products together based on their true historical demand (including lost sales).

For example, if your clusters were ranked from A to F by volume of sales per category, one store could be an A level in women’s jeans. However, if you went one level lower to womens fashion jeans or even lower to womens fashion jeans by True Religion, for example,  you could see that store move to an F in fashion jeans but a C in True Religion jeans. By going across the chain and identifying which merchandise performs at which level at each store, you can send a more accurate volume of each product to each store. The more specific you can be by product and attribute cluster, the closer you can get to creating a localized plan for every store in your chain. And the more constantly you update those clusters, by week or month at least, the truer you will stay to actual store demand.

Three major problems to look out for in clustering

With few exceptions, the vast majority of clustering has three major problems associated with the process. These three issues are seriously inhibiting the retailer from truly localizing their stores.

  1. Clusters are almost always based on historical sales performance alone.
  2. Clusters are typically locked in for a season or similar time period. If the recent economic climate has taught us anything, it is that store behavior changes and it changes rapidly.
  3. Clusters that group an entire store are incredibly inaccurate, even clustering by department or category won’t achieve the goal of true localization. Significant value can be gained if stores are clustered at or near the individual product level based on merchandise and location attributes which have been proven to have a direct correlation to product behavior. This must be executed using historical, current and expected product demand.

Plan for future demand, utilize historical demand

Clustering based off history alone is a mistake that almost every retailer makes. But the typical merchandiser does not have much of a choice. History is the only thing that they have at their fingertips on which to cluster. But, this means of clustering misses the quite obvious fact that product performance last year will not equate to product performance this year. It also misses the not so obvious fact that each store likely missed sales opportunities during periods of stock-outs.  That’s why history is not the best base for clustering products together.  A consideration of expected future behavior must be made alongside a historical adjustment for what could have happened in the past, had the inventory been there to meet demand. Clustering on a trend or, better yet, a forecast at the merchandise level is a better way to cluster products.

Stores are dynamic

The second issue mentioned is that clusters are typically created and locked in for a season or more.  An individual store that performed as an A cluster last year during the spring season in women’s tops will be clustered again as an A store for the entire spring season this year. However, we know that store will frequently not repeat the same performance year over year especially with same or similar products across all departments. Stores need to be able to move within a cluster to more closely align their actual performance with unique merchandise.

If stores don’t move with their performance, they aren’t being localized. Stores will underperform and be left with merchandise to markdown or overperform and stock out. If, however, stores actual current product performance defines and modifies the cluster and thereby their merchandise levels and breadth of assortment, these problems are less likely to happen and the store is being localized more effectively. By doing that, we are continuously introducing small amounts of change into the way that products are being assorted into stores, which is more manageable and timely in reacting to the way that customers are really acting in the stores. That’s an incredibly powerful piece of the merchandising puzzle.

The best way for stores to be localized given that it is impractical to expect an individual assortment per store is by having dynamic clusters. The assortment planning process should include a periodic review of each store’s product performance versus the product’s cluster and make a recommendation to move a store to a different cluster or modify it’s depth based on a variety of criteria. This allows merchants to fine tune the assortment that will perform best in a store given the store’s behavior right now, not last year.

Clustering at the merchandise level

The last issue is clustering by attributes or merchandise rather than by merchandise hierarchy alone.Clustering solely on store attributes, geography and demographics misses a significant opportunity for store localization based on how merchandise attributes collectively perform at an individual store.

An example of this can be found in price point. Let’s look at jeans again. If you cluster jeans based on the price tier (for example: A through F by sales volume), the stores that perform best with higher priced jeans would be assigned the A cluster grade within that product category, whereas a store that has mediocre performance for high priced jeans would be given a store grade F, which would equate to a lower level of inventory sent for that price tier – or perhaps even changing how much of that product the store is eligible to carry. This is even more accurate than grouping the stores based on demographics such as income level. Just because a store is in a nicer neighborhood does not mean that higher priced merchandise will sell better in that store. Honestly, if the retailer creates product-specific clusters with stores that actually perform better in the type of merchandise, the demographic information hardly matters!

In Summary

Today, nobody expects every store to receive its own assortment plan. Every store, however, can receive its own localized, unique assortment even when clusters are being utilized.

Recap:

  1. Cluster on more than just volume and history but include demand, both past and future in the clustering equation.
  2. Constantly update the store cluster assignments based on actual store behavior.
  3. Create localized clusters based on how merchandise attributes collectively perform at an individual store.

By following these guidelines, a merchant can have a positive impact on their chain’s performance and will be able to create localized plans for the individual stores.

Beyond clustering, executing at the SKU level: Quantum’s approach

Dynamic SKU level awareness //

Q serves as a comprehensive demand platform for retailers. It calculates demand and selling behavior for each product, at every store individually, and does not dilute the sku/store demand signals through traditional averaging and aggregating techniques. Q dynamically creates unique SKU/store profiles, for topics including seasonality, day of week, time of day, life-cycle, etc., and manages against supply and demand side events and constraints. This enables Q to make the atomic level adjustments necessary to capture extra full prices sales while greatly reducing inventory investment.

Continuous learning  //

Unlike other solutions, the value and intelligence of Q constantly improves as Q learns from product behaviors. Item profiles in Q are constantly updated by the system whenever new data is available, allowing it to accurately predict how that product will act with the knowledge of how it has acted before, while taking into consideration how it is acting now. This constant learning is unlike anything offered by other vendors.

Real demand visibility //

Q gives you visibility of the real demand in your stores now. With the understanding of lost sales for every product, Q prevents you from missing opportunities so you can capitalize on every potential sale. Q learns from how your product is moving right now, so you do not need a year’s worth of data to predict how a product will perform. By continuously tracking 35 performance metrics, such as: average demand, average sales, seasonal affects to product life-cycle patterns, shelf life, maximum sales per day, average inventory by date, in stock and in transit, this lets retailers calculate potential consumer activity and demand every day.

Download a PDF of this blog series HERE»

Sign up to receive updates throughout the series.

To learn more about clustering with localization, visit: http://quantumretail.com/2010/08/24/the-profit-lab-clustering-with-localization-in-mind/

Debunking the cluster myth in allocation: http://quantumretail.com/2010/06/04/the-profit-lab-the-cluster-myth/

Be the first to like.
Unlike

The Profit Lab: Clustering with localization in mind

by Quantum Retail • August 24th, 2010 • 2 Comments »

THE PROFIT LAB // 4 Strategies to Optimize Assortment Planning

WEEK 2

Years ago the store owner knew his customers by name. He could pull their goods in advance of them coming in to the store. If the customer wanted something that the store owner didn’t carry, the customer could request a specific item to be added to his assortment and he could choose whether it would be worth it or not. As store chains developed, non-centralized planning and merchandising allowed the store manager to keep his finger on the pulse of his customers. What were they asking for? What did they like? What did they not like?

Today is a much different picture. These same chains have expanded their store counts by hundreds, if not thousands, and now rely on buyers and planners that sit in headquarters trying to determine how to localize the assortments to maximize the potential revenue and margin that each individual store has the ability to provide. How can today’s merchant personalize and localize an assortment the way the store owner or store manager would have done when they were responsible for just one store? The obvious answer would be to assort each store independently, but that just isn’t realistic. There are not enough people to do that. The answer lies in clustering.

The Beginnings of Clustering

Clustering started decades ago as chains began reaching the high double digits in store count and merchandising became more centralized. Back then, everybody did it the same way. Stores were ranked in terms of sales and grouped, usually by percent of average. The “A” stores may be those stores that perform at 200% of the “average store.” Of course, there was no “average store,” but it was the total sales divided by the number of stores that represented the average. “B” stores could be 160% to 199% the average store, and so on. The number of clusters were somewhat a semblance of how many stores were being managed, but also the number of clusters a buyer or planner could manage was a factor as well.  The more clusters there were, the more precise the assortment could be, but the more difficult it was to merchandise. Trade-offs were common. This was the beginning of clustering.

Merchandise Hierarchies

Next the merchants started to group the stores by merchandise hierarchies. Categories, departments, and classes now were getting their own clusters of stores, a logical transition. An “A” store could be a fantastic store in women’s career apparel, but terrible in men’s accessories. This allowed merchants to be increasingly specific in building assortments that would perform better in certain stores.

This is about where your typical retailer is today. A majority of retailers dissect their stores into volume (sales) based clusters in this manner at a merchandise hierarchy level. That merchandise hierarchy varies, but it’s typically at the level that planners are building an assortment plan, most likely to be class. While a majority of retailers are at this point, a few have successfully moved beyond this stage and made a variety of improvements.

Nested Clusters

Some clusters are nested, building clusters not just on sales volume, but also on a variety of store attributes. Climate is probably the most common and most logical. This has a big impact on a variety of categories. Outerwear will sell better (and earlier) in Minneapolis than in Miami. Store size is another somewhat common attribute that merchants use to cluster as is demographic information such as race, religion (for some classes heavily influenced by holidays), or income. All of these make sense, but they are far from being universally adopted.

Statistical Clustering

A mathematician would tell you that what I have previously referred to in this article as clustering is actually “grouping of stores.” Pre-determining both the break points as well as the number of groups doesn’t allow stores to truly “cluster” together, but instead to simply “group.” By applying statistical methods to clustering, stores that are truly more alike will end up in the same cluster. The number of clusters becomes statistically relevant as well, and not something as simple as 26 clusters because that’s how many letters there are in the alphabet. You laugh, but I’ve seen it more than once in my career.

An Evolving Process

So, the evolution has begun, clusters are now really clusters, as opposed to groups. Stores are being clustered together based on more options than sales volume alone and being clustered with statistical accuracy. Consideration for demographics or store attributes such as climate are now commonplace.  However, there is a big piece missing that I haven’t hit on yet.

Three major problems of clustering

While there might be exceptions out there the vast majority of clustering has three major problems associated with the process. These three issues are seriously inhibiting the retailer from truly localizing their stores.

  1. Clusters of stores are almost always based on historical performance.
  2. Clusters are typically locked in for a season or similar time period. If the recent economic climate has taught us anything, it is that store behavior changes and it changes rapidly, especially at the merchandise levels.
  3. Clusters are hindered by store attributes. Significant value can be gained if stores were clustered based on merchandise attributes.

Plan for Future Demand

Clustering stores based off history is a mistake that almost every retailer makes. I understand why, the typical merchandiser does not have much of a choice. History is the only thing that they have at their fingertips on which to cluster. But, this means of clustering misses the quite obvious fact that stores performance last year will not equate to store performance this year. That’s why history is not the best base for clustering stores together.  A consideration of expected future behavior must be made. Clustering on a trend or, better yet, a forecast at the store/merchandise level is a better way to cluster the stores.

Stores are Dynamic

The second issue mentioned is that store clusters are typically locked in for a season or more.  An individual store that performed as an A cluster last year during the Spring season in Womens Tops will be clustered again as an A store for the entire Spring season this year. However, as often as not, that store will not repeat the same performance year over year especially in every department. Stores need to be able to move within a cluster to more closely align their actual performance with merchandise levels.  If stores don’t move with their performance, they aren’t being localized. Stores will underperform and be left with merchandise to markdown or overperform and stock out. If, however, stores actual current performance dictates the cluster and thereby their merchandise levels, these things are less likely to happen and the store is being localized more effectively. By doing that, we are introducing continuous small amounts of change into the way that products are being assorted into stores, which in itself is more manageable and timely in reacting to the way that customers are really acting in the stores. That’s an incredibly powerful piece of the puzzle.

The best way for stores to be localized given that it is impractical to expect an assortment per store is by having dynamic clusters. The assortment planning process should include a periodic, typically weekly, review of each store’s performance versus its cluster and make a recommendation to move that store to a different cluster based on a variety of criteria. This allows merchants to fine tune the assortment that will perform best in a store given the store’s behavior this season, not last year.

Don’t Forget the Merchandise!

The last issue that I have called out is clustering solely on store attributes. There is clearly value in merchandising based on some store attributes. Climate is the best and most obvious example as this not only affects the breadth and depth of the assortment, but also the flow of the merchandise. I remind you of my earlier Miami and Minneapolis example in outerwear. You’re not only going to have more choices in jackets in Minnesota, but you’ll have more inventory as well as an earlier flow of merchandise. However, clustering solely on store attributes missed a significant opportunity for store localization based on how merchandise attributes collectively perform at an individual store.

An example of this can be found in price point. If you cluster a class of merchandise based on the price tier (“good, better, best” is common representation of this), the stores that perform better with higher priced merchandise will be grouped together. I would argue that this is even more accurate than grouping the stores based on demographics such as income level. Just because a store is in a nicer neighborhood does not mean that higher priced merchandise will sell better in that store. Honestly, if the retailer creates clusters with stores that actually perform better in the type of merchandise, the demographic information hardly matters!

In Summary

Today, nobody expects every store to receive its own assortment plan. Every store, however, can receive its own localized, unique assortment even when clusters are being utilized.

Recap:

  1. Cluster on more than just volume and history, by incorporating attributes of not only stores but of the merchandise.
  2. Constantly update the store cluster assignments based on actual store behavior.
  3. Create localized clusters based on how merchandise attributes collectively perform at an individual store.

By following these guidelines, a merchant can have a positive impact on their chain’s performance and will be able to create localized plans for the individual stores.

<< Previous post in series | Next post in serious >>

Learn more

To learn more about assortment planning, be sure to check back weekly, or sign up below to receive email notices when this blog is published.

Subscribe to receive weekly updates of this series HERE»

Download this blog as a PDF»

For resources on assortment planning, visit: http://quantumretail.com/solutions/assortment-range-planning/resources/

1 person likes this post.
Unlike

The Profit Lab: You’re biased, but it’s not your fault

by Greg Wilson • June 9th, 2010 • 2 Comments »

THE PROFIT LAB // Top 10 Ways to Pull Profit from Allocation

Strategy #3: Make time to analyze your assumptions for like products and see if they were true in the past

In the last post of this series we discussed location clusters in the allocation process. Here we’ll take that a step farther and talk about merchandise grouping.

In virtually all allocation processes it’s necessary to select a product, merchandise hierarchy level or group of products to use as a base of historical data to make allocation decisions. Assuming a constant of source data (historical sales, historical demand etc.) the decision of what products or product groups can be the single most influential factor affecting the results of your allocation.

Choose a base of products that’s truly reflective of the item you’re allocating and it’s likely you’ll get a decent result. Choose a base that’s not and you almost certainly fail.

Ultimately the goal of this process is to select product(s) that have acted in a way that we expect the allocated product to act. In the process of making this decision we use our judgment to make assumptions based on what we believe to be similarities. Since many products we allocate are new, we can’t use the same item’s actual history. Even when we can there is rarely enough activity at SKU / store level to make good decisions. As a result we look for similar items to group together and give us enough data for the decisions we need to make.

In the process of deciding what to group, we tend to think about similarities of products. Similar fabrications, silhouettes, colors etc. We assume that these similarities mean items will act similarly. Sometimes that’s true. More often than most people realize, it’s not true. We’re biased in our assumptions, but we have to be. Unless you are unique in the retail industry, as an allocator you’re not afforded the time to prove out the assumptions you make – you have to execute so often and so quickly that there’s just not enough time to validate every decision for each allocation. So what’s an allocator to do?

What you can do now

Make time to do some analysis. A better allocation puts merchandise that would have been sold at markdown into a store that will sell it at full price rather than being out of stock and missing the sale completely. The truth is that making a better decision about the product base of your allocation consistently can often lead to as much as a 1% increase in revenue and margin. It may not sound like much at first but for a $1B company that’s $10M in sales and somewhere around $3-4M in profit in a typical case. That can quickly pay for a few more people, better technology or other improvements.

Take a day as an analysis day and look at products based on the characteristics you tend to group by. When you look at all blue tops individually, do they sell similarly? Is this true by store or cluster? If so, it’s a keeper criteria; if not, maybe fabrication or silhouette will have more impact as a group. The influencers will be different for different product groups, so take different passes for different product groups. Expect the things you find and results you get to be subtle. Using the better criteria will usually only shift a few packs within a given distribution. When that shift is consistently for the better and it’s multiplied across all stores, all products and for all receipts it quickly adds up to significantly better performance for your company!

What you should consider when looking for new capabilities

When looking at updated capabilities in allocation look for systems that are constantly analyzing product behavior by store. The best systems today are looking across multiple levels and groups of merchandise and location and learning about how each relates to individual product behavior within individual stores. Doing this allows them to do the analysis described above at much more detailed levels and to constantly update the results which become the basis of your allocation decisions.

Done right, results from using technology with this capability can quickly push that 1% gain mentioned above into 2-3% gains. That’s potentially $9-12M of profit per year. Enough to pay for the investment within just a few months.

<< Previous post in series | Next post in series >>

Learn more

Follow this series to learn all 10 strategies for improving allocation. We will be deconstructing the allocation process and exploring opportunities to improve within your current allocation processes and technology limitations. We will also review key areas to think about if you are considering investing in improved allocation capabilities.

Subscribe to receive weekly updates of this series HERE»

Download this blog as a PDF»

For resources on allocation, visit: http://quantumretail.com/solutions/allocation-replenishment/resources

1 person likes this post.
Unlike

The Profit Lab: Debunking the Cluster Myth

by Greg Wilson • June 4th, 2010 • 1 Comment »

THE PROFIT LAB // Top 10 Ways to Pull Profit from Allocation

Strategy #2: Learn the unique behavior of each store before clustering it

Location clusters. We all know them and some love them. They have their place, especially in larger retail organizations. It’s important to think about why they exist before determining when and how to utilize them.

For the most part they only exist to allow merchants to look at data in a way that reflects some localized perspective, but makes it more manageable than evaluating each individual location. This is fine for review purposes, but when it comes time for execution, we need to consider the uniqueness of each individual location.

There are a few different ways clusters are utilized in Allocation

Some of the most common include:

Setting constraints: Min, max, calculations and caps, inventory target ratios, presentation quantities etc.
Making adjustments: When reviewing allocations making cluster or store group level changes to allocation quantities.
When allocating: Sending X number of cases to “A” stores, Y number to “B” stores etc.

There are two realities that generally keep clusters from being effective – especially in the context of allocation

First – Cluster definitions must be created. To keep this process manageable this is executed at aggregate levels of merchandise – either hierarchy levels (categories or classes) or by manually selected groupings. This process involves making assumptions about what products will behave similarly and in the case of using hierarchy levels, forces items known to be different to participate in the groupings. This can be convenient as it tends to smooth or “normalize” some of the volatile activity out of the result. Unfortunately it virtually always colors the answer to the point that clusters have similarities to some of the products you’re applying them to but are not truly reflective of the unique behavior of any of the products you’re applying them to.

Second – Store behavior ALWAYS changes. As a result, no matter when you’ve clustered and how perfect your clustering criteria are, the cluster definition is almost immediately out of date. Conventional wisdom says that there isn’t enough change in a season or quarter to warrant revising clusters. The reality is that unbiased analysis shows us that this assumption is undeniably wrong. The change is constant and relentless. This is especially true in the schizophrenic economic conditions we’ve been experiencing in recent times.

Localization

If you’re company has localization as a goal, you can’t truly achieve that if you’re driving the allocation process with clusters as described above. Clusters can still be a part of a successful allocation strategy, but a process to uncover the unique behavior of each location is the only way to achieve true localization in the allocation process.

What you can do now

If you’re currently allocating by setting the number of cases you send to a cluster, STOP! Even the most technologically challenged organizations can usually find some source of recent historical data by store and build a spreadsheet that can use that individual store understanding to improve results. You can still use cluster level constraints (i.e. min / max) to buffer the volatility that comes from looking at store detail.

If you already have technology that looks at store history and you’re setting criteria at cluster level, evaluate the clustering criteria more frequently. Stores are changing their behavioral characteristics faster than ever before. We need to keep up with the change or results will suffer. Even if your system dynamically generates clusters based on indexes, the assumptions that went into those index definitions have probably all changed – and they’re different for different merchandise. Put the time into updating them.

What you should consider when looking for new capabilities

Modern allocation systems are capable of learning much more about how individual stores and products behave. In addition to considering demand rather than just historical sales (as discussed in the last post in this series), these systems understand a variety of levels that have been proven to influence the behavior of products within locations. They constantly monitor that behavior and utilize it to set store specific objectives, eliminating much of the need for clusters from the Allocation process. Clusters are still used to set some constraint criteria, however the individual store awareness means that this criteria is less critical to achieving the desired results. The range of these criteria is often softened (e.g. reducing mins and increasing maxs) or in many cases eliminated due to the fact that the detailed store understanding accommodates them.

<< Previous post in series | Next post in series >>

Learn more

Follow this series to learn all 10 strategies for improving allocation. We will be deconstructing the allocation process and exploring opportunities to improve within your current allocation processes and technology limitations. We will also review key areas to think about if you are considering investing in improved allocation capabilities.

Download this blog as a PDF HERE»

Subscribe to receive weekly updates of this series HERE»

For resources on allocation, visit: http://quantumretail.com/solutions/allocation-replenishment/resources

3 people like this post.
Unlike