Shopping Carts Optimization – Ideas for Implementation

Commercial retail companies across the globe strive during traffic hours to provide the best experience to its customers. Large supermarkets/hypermarkets deploy extra manpower, more digital techniques like self checkout to provide its customers a hassle free shopping experience. This aids the retailer to build a good brand value by ensuring user friendliness, which in turn improves sales and margins.

Table of contents

What is shopping cart optimization?

Why is it required?

How is it done?

How is it useful?

What is shopping cart optimization

A shopping cart is a physical trolley inside a store which enables the customers to carry goods until checkout. It enables customers with the logistics within the store, to carry their basket along and shop for more products at once.

It is an indispensable tool at the time of a rush hour, and a non-availability of the same may result in a tussle thereby rendering a poor customer experience. It is therefore necessary for a retailer to stock an optimal number of shopping carts within a store in order to cater to the customer ‘s needs at all times.

Finding an optimal number of carts indulges us to look into a plethora of factors, some of them may be pertaining to store formats (speciality, large format, supermarket or hypermarket), geography (population density, shopping preferences, shopping time, best fit campaigns), demography (median income, age group, competitors in the vicinity, selling frequency etc. to assess basket size and spending power), customer likeliness of products (towards FMCG, apparel, general merchandise etc.)

By arriving at an optimal number of carts using historicals via forecasting techniques, it would enable the retailers to leverage their resources at best, and ensure that the traffic is appropriately managed inside a store, rendering a quality experience to its customers

Why is it required

Let’s consider a usual scenario where you are in desperate need of stocking up on eatables, and decide to visit the nearest hypermarket after work hours which is fortunately open at that time. You realize you would need to shop for a big basket that would include multiple products such as grains, cereals, bread, vegetables, milk, ready to eat products etc., which would certainly require you to borrow a shopping cart, while inside the store. However, you see a huge crowd already waiting inside for their turn to check out.

Upon entering the store, you notice apart from traffic, the carts are all occupied and it would take you quite a few minutes, if not an hour, to obtain one. Based on your personal best judgment, you plan to drive to another nearby store having relatively a lot fewer items in stock, but having relatively low traffic with more unoccupied shopping carts. You decide to compromise, but rate the previous retailer for a poor experience and crib about the unproductive time spent inside that store.

If every other customer does the same, the retailer would soon lose its reputation and brand in the market, and even if it has some excellent products to sell, it may not project itself as a user friendly partner for its customers.

In order to mitigate such risks, a retailer must be cautious about the resources that it makes available and handy, for its end customers. Shopping carts are one of the primary logistic support mechanisms that the customers want, after entering the stores. If they find one easily available, they would tend to relatively shop more in a single visit, moreover visit multiple times providing an opportunity for the retailer to upsell, as it comes at their ease. This would increase the footfalls and sales.

How it is done

Majority of retailers deploy standard statistical mechanisms to calculate the optimal number of carts required at a given point of time, in a store (on their sales floor). The initial task is to group similar stores to perform analysis on a cluster rather than doing this exercise individually for every store.

There are multiple methods by which stores can be clustered, few of them being unsupervised methods like K-means (non-hierarchical – if all key variables are given importance for the cluster) or supervised techniques like tree regression/CHAID (where one focuses on determining the relevant factors driving the number of carts in a cluster of stores).

The key factors to be considered for grouping may be the transaction data, footfall information, sales, and other store related characteristics related to geographics/demographics or macro-economics as discussed before. These factors are assigned importance based on the way they were aligned to the shopping cart needs in history.

Once the clusters (of stores) are formed, they must make sense in terms of the business i.e we should be able to clearly define and distinguish between the clusters. We may apply manual tweaks over the statistical outcomes to make it in line with the business needs. For instance, a cluster can be defined as something that has a customer with ‘preference towards apparel’, ‘falls in age bucket of 18-30’, ‘lives within 10 miles from the store’, ‘is loyal customer based on recency, frequency, monetary measures’, ‘high income’, etc.

Hereby, we may decide to either apply a conventional or an advanced approach in order to predict the number of shopping carts necessary to stock in a store, at any given point in time.

A more conventional approach would demonstrate the comparison of every store’s carts (say ‘x’) against the mean of the shopping carts for that cluster, on an hourly basis. We can further segregate the time window by clubbing hours as peak and non-peak, for simplicity.

In other words, calculate deviation in carts = (x-mean)/standard deviation for each of the stores within a cluster. And the stores that show up as outliers (depict a high deviation), cap them at 95% limit. This method holds good if the distribution of carts across stores in a cluster follows a normal distribution, which it generally follows if there is a good mix of stores in a cluster in line with the overall population for that company. Probably, it would be a good idea to validate the same before performing this analysis. As we have fixed the deviation to be not going above or below 95% compared to the mean, we would need to restrict ‘x’ (no of shopping carts) to this limit. This would need a retailer to add or subtract carts from their kitty to reach the optimal service level. ‘Service volume at any given point in time can be defined as the ‘no of transactions/total number of carts’ in a store. This is a little vague and one may argue that there may be smaller transactions that may not require shopping carts at all. However, on an average this would still be a reasonable measure of the efficiency at which the store carts would cater to the shopping needs of the traffic, at a given point of time. Again, depends on the nature of the industry we are dealing with. This would be more relevant for groceries/general merchandise but less for apparels. If a retailer notices that the ‘service volume’ exceeds 1, they must treat it as a warning signal and try to optimize the number of carts, from next time onwards.

A more advanced way of dealing with this would be to predict the optimal number of carts in a dynamic way (keep updating the time window to include the most relevant historical data into the model), formulating a regression model to calculate the carts at a cluster level. If at any point in time, the actual number exceeds the predicted, one must try to identify the hidden factors that might be driving it like macroeconomics. Another example is the seasonality parameter which can drastically affect the footfalls. For regular days, the retailer can deploy BAU techniques with the seasonality and macroeconomics flags turned off, and for exclusive events turn them on to predict the number of carts.

The whole objective of this exercise is to efficiently cater to the needs of the buyers, rendering a positive experience.

Benefits and outcome

As narrated in the previous sections, the voice of the customer apart from margins is a significant driver for the success of a retailer. Specialized hypermarkets/superstores deploy this technique to understand the customer needs and address it appropriately.. Generally, shopping carts are considered to be merely a logistic support and often neglected by the retailers, however one needs to acknowledge that it’s a key driver to increase sales and enhance the customer perception of a brand. Hence, retailers must leverage this as a tool to drive their store’s footfalls, transactions and sales.

Leave a Comment

Your email address will not be published. Required fields are marked *