Assortment Optimization

Assortment Optimization is an extension of or a mature form of assortment planning. The task of a retailer to plan the assortment mix in a given store is complicated by various factors such as the differences in the format of the stores, size of the stores, demographics of the population around the stores, the product categories’ performance in the given region, the life cycle of the product/product category etc. Retailers start from a strategic plan with respect to the assortments they want to carry and then refine the product offerings into different categories, sub categories and SKUs. The assortment plan will be made category by category, channel by channel, cluster by cluster. (Cluster would mean the store clusters formed based on certain characteristics – Please refer Store Clustering).

For instance, the retailer may want to classify the products broadly under food and non-food. Food products can be classified as shown below:

The “Types” shown above can branch out to different SKUs. The task in front of the retailer is to come up with the demand forecast for the different SKUs and stock the right mix of SKUs or product categories in the right stores/channels in right quantities. Demand forecasting is being done by various retailers using various tools. These tools employ a variety of scientific techniques that range from a simple “moving average” method to more complex algorithms like “Exponential Smoothing with trend and seasonality”, “Auto Regressive Integrated Moving Average” (ARIMA) etc. Keeping aside the science & techniques involved in demand forecasting, let us try to understand certain other factors that influence the merchandise mix.

Factors influencing Assortment decisions

The decision to select on the appropriate merchandise mix will be primarily based on few attributes of the SKUs and product categories. Some of those attributes are given below:

a)      As we all know, 80% of the sales in a store will be driven by 20% of the SKUs. So the SKUs or product categories can be ranked in the descending order of sales and the retailer would try to retain those products that drive high sales volume and try to discard those SKUs which do not contribute to sales.

b)      There may be some products which will be sold in very less in number of units but their profit margin may be significantly high. Obviously, retailers would try to retain those products that have a high profit margin and discard that do not.

c)       There may be some products which may not really drive huge sales in a store but then it may be of significant importance to some of the loyal customers to that store. Discarding those stores may result in dissatisfaction of some of the loyal customers. Hence retailers would try to retain those products or SKUs.

d)      There may be certain SKUs which have a tendency to drive the sales of other products. I.e. Those SKUs will influence the sales of other accompanied products. (Market Basket Analysis can help in figuring such associations between products)

e)      Certain products may be key differentiators for the retailer and may provide a competitive edge over its competitors. A retailer would wish to retain such products in the offering.

f)       Few other key performance metrics can be collected for product categories and used to determine, whether or not to retain those products. For instance, percentage growth of the product category or the SKU can indicate the performance. Whether or not the product category/SKU has achieved the sales target can also be a performance index. Such performance metrics should be measured store by store or cluster by cluster, format by format to decide upon the assortment mix.

The effect of substitution within product sub categories or across different SKUs should be taken into account as well. For instance, if a pack of cream wafers in orange flavour is not available in a store shelf, the consumer may choose to buy cream wafers in strawberry flavour. So, while deciding on the quantity of strawberry flavoured wafers to stock in a store, the retailer should be aware of the distinction between the original/uninfluenced sales of strawberry wafers and the sales influenced by the absence of orange flavoured wafers. Another factor that should be considered is the lifecycle of the product category itself. I.e. A product category like an MP3 player is in “Decline” phase. A product category like a “Laptop” is in “Mature” phase. A product category like an “iPad” is in “Growth” phase. However, the product life stage may not be the same in all market segments. For instance, laptops may in a mature phase in a tier-I city in a country like India but may be in a growth stage in a tier-II city.

Thus, based on the demand forecasting (that was determined from historic sales) for different product categories and based on the different attributes listed above, the assortment mix for different formats and different store clusters can be arrived at and optimized.  Assortment optimization also covers the channel of distribution as well. In the world of e-commerce, the retailers need to have mechanisms to optimize the assortment mix in both the physical stores and in online catalogues. Thus the retailer is supposed to monitor the performance of products in different channels as well and decide upon the assortment mix accordingly.

Significance of Assortment Optimization

Once the assortment mix is determined by retailers, for different clusters and for different channels, that information will become a vital input for store space planning or planogram design. For instance, the product categories which the retailer wants to sell more should have more shelf space and the facings per shelf should be more as well. One can also appreciate the fact that the assortment mix will in turn drive other key decisions of a retailer including the supply chain and marketing promotions.  If certain product categories have a huge potential to sell than the retailer can offer in a certain store, it may pave way for opening new stores as well. At the same time, if certain SKUs are not selling well, the retailer may consider devising markdown strategies for those products, to clear off the stock. These factors underline the importance of the assortment optimization.

There are sophisticated tools available in the market to do the assortment optimization but to deploy it in a retailer’s IT landscape, understanding the issues mentioned above is of utmost significance. The tools can even provide provisions to understand the impact of increasing or reducing or deleting certain product categories from the merchandise mix. With such advanced technological tools, the retailer can do analysis on product revenues and other performance metrics for different SKUs or product categories and use the information for future product launches. Such information will be very handy and practical to use, when new products are launched by the retailer and also to tune the assortment mix to adapt to changes in various environmental factors.

I believe this post gives the first hand information about assortment optimization and the factors that one should keep in mind, while trying to optimize the assortment mix.

 

Markdown Optimization

Markdown Optimization is one of the enticing phenomena for many retailers and also one of the buzzwords that are of interest to many retail analytics firms. Before talking about markdown optimization, let us try and understand what is markdown and what is the need to optimize it? Markdown, as most of us are aware is marking the selling price of an SKU below the normal selling price, in an effort to increase the sales of that particular SKU. More often than not, such an act is performed to clear the inventory from the stores. Though retailers claim to have many sophisticated techniques to forecast the demand and policies that govern the inventory replenishment, quite often they see some SKUs piling up in stores and feel the heat to reduce the price so as to get rid of them as early as possible.

Let us use some numbers to get a better frame. Assume that a pack of dozen donuts cost $3 for a retailer. The retailer sells it for $5 under normal circumstances, which will give him a profit margin of $2. Assume that the donuts expire in 3 days after the stock arrives in store. If the donut doesn’t sell by the date of expiry, the retailer has to just dump down the remaining stock and in this case, the loss would be $3. In the best scenario, the retailer gains $2 and in worst scenario, the retailer loses $3 for every pack of donuts. Hence as the date of expiry approaches, the retailer would take any measure to reduce his loss. Eventually he’ll end up marking down the price of donut. Let us assume the price is brought down to $4 on the day before the expiry date and to $2.5 on the day of expiry. If the retailer stocks 100 pack of donuts, he would expect to get a total revenue of $500. What if the actual sales don’t go the way he expects? We’ve taken 3 different cases depicting 3 possible sales figures on 3 days.  Selling price on Day 1 is $5, $4 on Day 2 and $2.5 at Day 3. All the remaining donuts will result in a loss of $3 for the retailer.

Let us take the last row – the retailer invests $300 (for 100 packs) and gets only $295. (355 – 60). If the retailer has 100 such stores, he will encounter a loss of $500, whereas he would have expected a profit of $20000. ($2 * 100 * 100). Instead, if he had kept the selling price at $4 on Day 1, and if that makes the retailer sell all 100 on the same day, he would have got a profit at $10000 ($1 * 100 * 100). (In addition to these loses, for most products, whenever there is a price change, that will involve generation of new price labels, updating the IT systems, moving the SKUs from one place to another etc., All this will cost to the retailer’s operations.)

Thus, the decision of how much to markdown and at what time is of utmost importance for a retailer. You may be wondering as to how the retailer will know that if he reduces the price to $4, he will get rid of all the donuts on the first day itself. This is where the concept of price elasticity comes into picture. This is where the factor of seasonality comes in. This is where the purchase behaviour of the consumer comes in. The increase in demand of a particular SKU for a unit change in price is called Price Elasticity. For instance, a retailer may think that if the price of a pack of donut is reduced by $1, that will increase the sales by 50 units. Donut sales may not be uniform in all seasons. It may be on the high during winter or spring. The consumer may prefer to purchase donuts may be on a Saturday morning or a Sunday morning but may not buy on a Monday morning. How does a retailer can know all these? Scientifically, that can be obtained by previous sales data. How the sales of a particular SKU varied with respect to day of the week, season of the year, variation in price etc., And for SKUs like leather shoes or a TV or a refrigerator, the sales may get influenced by general economic conditions.

The success of a retailer’s price optimization or markdown optimization depends largely on understanding the above mentioned factors. Ie.

a)      Price elasticity of an SKU

b)      Purchase behaviour of consumer with respect to that SKU

c)       Exact implication of seasonality on that SKU

d)      Macroeconomic conditions that prevail

If a retailer is able to figure out these, then most of the work in devising a markdown strategy can be considered complete. Unfortunately, many retailers fall into the trap of intuitively guessing the sales and end up generating mark down prices, which are not optimal.

Another set of retailers blindly go behind sophisticated products in the industry. There are quite a few retail analytics suite of products available in the market. All of them provide provisions to set up rules, do What-If Analysis, generate insightful charts etc., But the power of such tools can be extracted only if they are configured with the right data and right set of rules and assumptions. Hence it is imperative for the retailers to understand the factors influencing the sales and only that will help them in finding out what to markdown, when to markdown and how much to markdown.