Connect With Us
Assortment Planning and Up-selling based on “This Sells With That”
Walmart’s (formerly Wal-Mart) announcement of a SKU rationalization project contained in this year’s 10-K filing with the Securities and Exchange Commission confirms the importance of this initiative for all retailers. In SKU (Stock Keeping Unit) rationalization, a retailer examines the profitability of items and vendors as a whole. When done in a linear fashion it results in lost sales and bringing back the SKUs.
SKU rationalization projects look for “What items are bought together” so that retailers and distributors can improve assortment planning. As shoppers, we all know that we buy items in groups. It is the job of the retailer to figure out what kind of stuff we buy together, so that they can optimize their assortment planning. Simple example – If I cannot buy both bagels and cream-cheese at the same time, I will go to a store where I can find it!
SKU Classification Based on Frequency of buys and Product Relationships
SKU analysis for assortment planning is based on two key metrics:
- The frequency of buys. This is a metric that measures true popularity of an item based on how often customers buy this product. For measuring popularity, it is better metric than volume as it is not skewed by one-time large volume purchases by a few customers.
- How often this item is bought with other items. This metric is a measure of how strongly correlated this item is with other items that you sell. If an item is always purchased with another item (like bagels and cream-cheese), it is very important to know the “often bought with” items, and ensure that they are stocked together and in the right proportions. Not having one item from a basket of high affinity products will result in loss of the customer.
These two metrics also apply for Amazon-esque suggestive selling for online sales. Items that have high correlation with other items are candidates for suggestive selling, up-selling, cross-selling and add-ons. For example, this would be a way to detect that cables, cartridges and paper that are bought with a particular printer. So when that printer is bought, you can automatically suggest the other items as add-ons. (Not to get too technical here, but the suggestions are not symmetrical. So – you cannot suggest a printer when a customer buys paper!)
The implications of these product relationships cannot be emphasized enough on your merchandising strategy and your supply chain planning. Manufacturers, distributors and retailers struggle to manage thousands of SKUs. This SKU classification presents a methodical approach for assortment planning to maintain the most profitable portfolio.
SKU Categorization For Merchandising, Up selling and Cross selling
The second chart presents a more detailed discussion of the SKUs based on frequency of buys and affinity with other products. (Affinity simply means “this items sells with that”. )
I - Items that have low-frequency/ high correlation are important to detect. These are trouble-maker SKUs. As companies goes though SKU rationalization projects, these items often end up on the chopping block, only to brought back again because they caused lost sales. These items are difficult to identify and there is a need for sophisticated analytics to easily identify these items.
II – Items that are bought in high quantities, but always with other items are great candidates for merchandising and bundling. They are a natural for creating sales lift and revenue lift. It is often counter-intuitive, but your #1 top seller may not be in the #1 pair of top selling items. That is why linear analysis of the SKUs based on volume or frequency results in incorrect merchandising.
III – The low frequency/ low correlation items are the targets for SKU rationalization projects. However, these items are very difficult to identify. Hence SKU projects typically end up cutting the wrong SKUs. We call these items Low-Loners. If you are a distributor, you do not want to carry these items. They are perfect candidates for drop-ship.
IV – Items that sell in high frequency, but usually on their own, require high service levels. We call these Hi-Loners. Examples of these items are cigarettes and gas at a convenience store. And by the way, beer also falls in this category. And please do not believe the beer and diapers myth! It is a myth!
The challenge with SKU management is that companies make decisions based on product relationships from hear-say, industry veterans or tribal knowledge. I think that’s how the beer-diapers myth was started! Across thousands of SKUS, and with fast changing demand patterns, this results in errors, and not a sustainable process for assortment planning and SKU management. There is too much at stake to base a companies sales and revenue on hear-say.
As SKU management is getting a lot of attention, there is need for robust solutions based on real customer buying behavior, to help companies maintain their SKUs on an continuous basis. The value is high sales, higher margins and improved customer service.
Comments Off
Increasing Order Size With Basket Analysis
I came in to buy milk and I am walking out with 10 things in my basket. The man behind me had only one item in his basket. “How do you do that?” I asked. “It depends on what you come in to buy,” he responded.
There are a few “seed items” in the store that drive additional sales because of key concept ‘this is often bought with that’. These items are often found together in customer baskets and orders. Smart retailers will put these items as far away as possible, so that you have to walk through more aisles to get from one item to the other, in hope that you will buy more along the way. Bread and milk is a good example of that. The reverse is also true. For items that are often bought together, if the store does not carry both, they will lose the customer.
Every retailer knows that it is very profitable when a customer comes in to buy one item, but ends up with many more in his basket. Understanding the product relationships in the market basket is key to driving up the order size or basket size.
Understanding the Customer basket make-up
A retailer typically carries thousands of items. A small convenience store may carry 1,500 items. A grocery store typically carries 15,000. And the super stores like Wal-Mart and Targets carry well over 25,000 SKUs in each store.
Insight Into Customer Baskets and Product Relationships Based on Buying Behavior
The SKU management is a tremendous challenge because the buying pattern is truly a long tail. Retailers know their top sellers; these are easy to identify, but the frequency of buying falls of very sharply. The chart shows an example of one retail store operation over a 3-month period. The store carries 25,000 SKUs, has 100,000 transactions per month. The analysis covers a 3-month period, and shows the distribution and popularity of SKUs based on the frequency of purchase.
Here are some quick stats for insight into the baskets and buying behavior – The most popular SKU has a frequency of 3,435. That means is has been bought in 3,435 baskets. The frequency of the 100th most popular item drops off to 225. That means it is only in 225 baskets over the 3-month period. There are 4,000 SKUs that are bought only once. But the really interesting fact is that 1,800 SKUs are bought together 98% of the times. None of these 1,800 SKUs are top sellers! But when they are purchased, they are very often paired with other items. This intelligence is key to increasing basket size and ensuring the store is carrying the right items. SKU rationalization analyses that view each SKU as an independent item, that is bought in isolation, will result in incorrect merchandising and lost sales.
There basket analysis also showed the low-frequency/high-correlation SKUs. Every retailer knows the challenge with these items. These items sell rarely, they sit on the shelf for along time, and when it is placed in a basket it will only sell if the paired item is available! These are problem SKUs because they are capital hogs and always show up in inventory issues.
Insight into the basket make-up and the product affinities based on buying behavior is key to merchandising and increasing order size. Merchandizing, up selling, cross selling and add-ons based on buying behavior results in increased sales and enhanced customer experience. On the other hand, suggestive selling based on tribal knowledge and ‘he said/she said anecdotes’ will result in poor results and loss of customer good will.
Adding one more item to 10% of the baskets can increase sales by 5%
Sales Impact Of Increasing order size
The basket size or order size analysis shows the revenue potential of increasing the order size. The chart shows a typical basket size analysis and the upside opportunity of increasing order size. The results from this case study showed that adding one more item to 10% of the baskets can increase sales by 5%.
Manufacturers, distributors and retailers offer thousands of products. There is a significant opportunity to increase sales across all channels with knowledge of product relationships (what items sell together), when and where. It is commonly agreed that B2B purchase behavior is “need based” while a large percentage of B2C sales is emotion based. Hence, in B2B commerce, the product relationships have to be highly accurate to be relevant.
Quick review of definitions:
Frequency – Number of orders that contain this item
Volume – Number of items sold.
The volume of an item may be high because one customer bought a lot. However, frequency is better measure of popularity and is not skewed by a one-time large volume sale. In fact, SKU analyses will often remove large volume buyers to reduce this bias.
SKU Rationalization Demands Market Basket Analysis (aka Customer Buying Patterns)
SKU Rationalization Demands Market Basket Analysis (a.k.a Customer Buying Patterns)
Wal-Mart Stores Inc., the world’s biggest retailer, is bringing back some products it had removed from shelves last year as shoppers turn to competitors for a wider selection of merchandise. A failed SKU rationalization effort?
The company met with suppliers about reinstating items to keep customers from going to other stores, said Leon Nicholas, a director at consulting firm Kantar Retail who has spoken with manufacturers about the move.
Wal-Mart is telling suppliers it cut too much in some areas and wants to bring some items back, Smith said. The retailer is noticing that consumers are visiting other stores and no longer going to Wal-Mart for everything they buy, he said.
“I’m learning this from my suppliers who were down to one SKU in the store,” said Smith, who helps vendors hire account managers and other representatives to call on Wal-Mart merchants. “Now they’ve got a seat back at the table.”
Sales at Wal-Mart’s U.S. stores open at least a year declined 1.6 percent in the fourth quarter, more than its forecast of a sales decline of no more than 1 percent. Declining store traffic reflected disruptions caused by store remodeling, Wal-Mart Chief Financial Officer Tom Schoewe said last month.
Why did Walmart’s SKU rationalization effort fail? Because Walmart ignored the market basket effect. It is not an issue of cutting too many SKUs; it is an issue of cutting the wrong SKUs because you do not know the product associations in buying patterns. A low frequency items can be profitable and may be often bought with other low frequency items. If you cut one of these SKUs, you will lose the customer. On the other hand, there are SKUs that are bought in low frequency in 1-item baskets. The loners! These are typically low margin, high capital utilization SKUs. These SKUs can be easily identified with Customer Buying Patterns Analytics.
What most retailers ignore in SKU rationalization is the market basket effect. Profitable customers may take their entire basket elsewhere, if they can’t find certain items (even if those items are “slow-moving”). The market basket analysis across hundreds of thousands of SKUs requires advanced analytics. Based on testimonies from Wal-mart customers, people were in fact choosing to go elsewhere for many of their shopping trips. This is why Wal-Mart has changed their tune very quickly.
“They are calling me back and saying, ‘We need to hire somebody who has experience in this category and knows this buyer — it looks like we are back in business,’” said Smith, who is based in Rogers, Arkansas. I think Walmart does not get it. They need help not just in categories; They need help across the categories. They need help on what items are typically in a basket – also called Customer Buying Patterns!
Comments Off
What is Pattern Based Analytics?
Emcien's Pattern Based Analytics Automatically Reveals Choice Combinations and Trends in Sales Transactions
A fast emerging area of business analytics is Pattern Based Analytics (PBA). This has been launched due to the very large amounts of data and need for analytics that can reveal meaningful patterns that businesses can act on. A typical reaction to the large amount of data is “If I had seen this coming sooner, I could have acted faster, decreased my risk and enhanced my opportunities for growth. Pattern Based Analytics typically requires focus on a business areas, e.g. Sales, Marketing, Finance, etc. The key to Pattern Based Analytics is automatically revealing intelligence that is hidden in the data/information.
This is a fast growing area because of key value points:
Instant Use - The inherent nature of Pattern Based analytics is that it does not require models and it accepts unstructured data. Hence, one of the greatest value points is Instant Use!
Accepts unstructured data – A key value point that drives down implementation time, barriers and cost, and dramatically increases applicability of the analytics. The ability to detect patterns in unstructured data makes it very easy for applications from sales data, marketing data, to twitter strings.
Big Problems are easy – Problem size and data size are not an issue with PBA. On sales data, Emcien’s PBA will easily solve buying patterns on 250,000 to 500,000 SKUs in a few minutes. This offers the ability to solve problems that were too large/expensive to solve previously. This is a game changer, when the closest alternate solution requires complex models and has serious size limitations of a few hundred SKUs.
Works on problems big and small – On problems big and small, PBA is a natural fit. PBA dramatically lowers the price of analytics, enabling smaller companies to gain immediate value from business analytics.
No data-models, No data-cube, No set-up – This is one of the single biggest value points for PBA. This eliminates the need for specialized analysts, statisticians and technical staff to interact and maintain the system. The ability to accept unstructured data and not require a model means No Setup. This also means you can go live now! No more 18-month implementation cycles!!!
Intuitive for non-technical users – Pattern Based Analytics can present results naturally in a very intuitive way. This is because the patterns that are pop are typically the top categories that need attention. There is not need to drill down and ask questions – the ultimate bain of every BI user.
When Pattern Based analytics is pointed at sales data, the patterns that pop are “what are the top selling items”, “what is the pattern of choices combination”, “where is this happening”? Any non-technical business user can use this report to stock better and drive more sales.
Always up to date – Patten Based Analytics does not use models and cubes. Hence there are no cubes to maintain and update. Even as time passes, the analytics are always up to date, due to the ability to input non-structured data.
Gartner has rightfully established Pattern Based Strategy as the next frontier for capitalizing on large volumes of data and deriving value fast and continually.
Comments Off
We don’t need to get new customers!
We don’t need to get new customers!
Very bold statement from the Macy’s CMO Peter Sachse in his keynote speech at the Retail Innovation & Marketing Conference talking about a shift in company focus.
Here’s how it all started: Last year, Macy’s embarked on an intense research project to better understand their current customers. They conducted dozens of focus groups. Talked with nearly a thousand people walking out of their stores. Leveraged data from NPD Group for a holistic understanding of their customers. Combed through all of their transactional data to find themes in buying patterns and shopping habits.
The overwhelming finding? For Macy’s, “What we don’t need to do is get new customers,” Sachse said. Instead, “we realized that all we need to do is take care of those who already love us.”
The company has set out on a goal to encourage each existing customers to visit the store one more time each year. “Half the battle is won if we can get them to walk into our store,” Sachse said. “And if we convert them during that visit, our comp store sales will explode.” To accomplish that goal, he said, “We had to get a lot closer to the customer,” which has led to the company’s new strategy of customer-centricity.
I could not agree more! Macy’s needs to understand the buying patterns of its current customers and serve them better. This will result in higher customer satisfaction, higher repeat sales and higher profits. If you do not know the buying patterns of your current customers, getting more customers is NOT going to help. Mr Sachse is absolutely right!
Companies today spend tons of money trying to get more customers. Very few companies have a finger on the pulse of the buying patterns and trends of their current customers. What is the point of getting more customers if you cannot serve the one that you already have? Is it just busy work? Or is it because with thousands of SKUs, companies do not know how to keep up with customer buying patterns?
Congratulations Mr Sachse! I look forward to walking into your store and finding the right stuff.
A Race Towards Pervasive Analytics
Gartner’s top 10 trends for 2010 set the stage for Cloud computing and Analytics.
Analytics is context driven, and presents actionable results to the business user. BI allows the user to slice and dice data. BI is good if you know what you are looking for. The reason Gartner placed analytics above BI is because of the needs of businesses today to act on data, as opposed to merely having access to it. There is way too much data. We do not need systems that create more data – we need intelligence from the data, which is what Analytics does. Hence the positioning of Analytics on the Gartner charts.
BI has become pervasive, as it should be. It has even entered open source with Pentaho and Jaspersoft, a sure sign of being pervasive! However, this was inevitable as every business user needs easy access to their data. A recent survey conducted by B Eye Network involving more than 1,000 respondents from around the globe found that only 12 percent said they had no plans to use open source software in some form for business intelligence applications or data warehouses.
However – converting the data to intelligence, and actionable intelligence in the next frontier. That is why Gartner placed Analytics in their top 10 trends chart, and moved BI out! As we have watched with Google analytics, the analytics on web traffic data is pervasive. There are a myriad of products that provide analytics on web stats, but Google provides a universal product ensuring that everyone has access to it.
Analytics on corporate data will also become pervasive. Companies are demanding this. The analytics will be contextual, as this is required for analytics to automatically make sense out of data. The analytics will be agile and companies will be able to pour their data in, and watch the results take shape. Much like Google analytics on web traffic data.
Emcien’s analytics offers analytics on sales data. The context is sales and customer buying patterns. Companies can now pour the sales data and watch the customer buying patterns emerge. No data mapping and model building. No long implementation cycles! The ability to “just turn on and use” is key to being pervasive.
The future is here!
Comments Off