Understanding Data Analytics in E-Commerce Retail - SwiftERM

Understanding Data Analytics in E-Commerce Retail. In the 2nd Quarter of 2020, US retail e-commerce sales have shot up to $211.5 billion, up from $160.4 billion in the 1st Quarter. The lockdown caused by Covid-19 has certainly contributed to this boost in e-commerce sales.

The below graph looks at the Q2 e-commerce sales as a percentage of the total retail sales in the US, in the last 5 years, so that you get a good feel of the rise of the e-commerce potential.

Understanding Data Analytics in E-Commerce Retail

Moreover, e-commerce sales in the US are set to surpass $1 trillion to make up 18.1% of total retail sales by 2024.

Let’s take a look at the performance of the top CPG (Consumer Packaged Goods) companies in 2020 so far, as compared to the last year in terms of the percentage share of e-commerce sales in total retail sales.

Understanding Data Analytics in E-Commerce Retail

So, now you know the importance companies are given to boost their e-commerce sales.

Now, let’s dive into the different avenues where analytics has a high potential of helping consumer goods companies scale up e-commerce sales.

Product Recommendations

Do you remember the time you visited the general store when the friendly assistant helped you throughout your shopping experience? He/She knew the  items you bought, the preferred brands you purchased, or the style your kid loved. They knew your shopping budget and would make recommendations. And you did end up purchasing what they recommended quite a few times. Didn’t you?

That relationship with the customers is what the companies need.

A product recommendation engine is the solution. What is a recommendation engine?

It is simply an automated form of the ‘friendly assistant’. It is a tool that uses machine learning algorithms to suggest products to customers that they might be interested in. The recommendations can happen through the following channels:

  • Within websites
  • Email campaigns
  • Online ads

The recommendation engines are of two types:

a) Impersonalised Recommendation Engine:

This is useful for sending out recommendations via welcome emails or bulk newsletters, without using customer data. Here are some useful categories:

  • new products like a new chocolate ice-cream that has been recently launched
  • trending products like a fast-selling coffee brand to help recruit new customers
  • slow-moving products like a luxurious bar soap at a discounted price

b) Personalized Recommendation Engine:

This is gold! It uses customer data to recommend the best offers and promotions. It uses deep learning algorithms and Natural Language Processing (NLP) to make the customer experience as personalized as possible. The algorithms use data from the following spheres:

  • Purchase history
  • Recent activity
  • Online behavior
  • Social media browsing

These algorithms are really powerful and can greatly boost e-commerce sales by:

  • improving conversions rates
  • increasing the average purchase value of customer
  • reducing bounce rates & cart abandonments

    Here’s what Amazon’s recommendation engine shows currently. It looks like it knows that you love to workout.

Understanding Data Analytics in E-Commerce Retail

Market Basket Analysis

Market Basket Analysis is a technique that identifies the strength of association between a pair/group of products that are purchased together. In simpler words, it’s based on the concept that if a customer buys a product, they are more or less likely to buy another related product. Market Basket Analysis also uses machine learning/deep learning algorithms, like product recommendation engines.

I’ll explain this with an example from the project at Colgate-Palmolive, mentioned earlier. We bundled a children’s toothbrush (unprofitable product) with a deodorant, a dad’s product (fast-moving product), and were able to increase the total weekly profit of the combination by 8%.

Association rule learning is a machine learning method that is used for discovering associations between products. It does not consider the order of the purchase of products. Let me explain this with a demonstration.

The analysis is run on millions of baskets, but we’ll demonstrate it with 7 baskets and explain the 3 important measures used. Let’s explore the association between shampoo and conditioner (Basket 1):

Understanding Data Analytics in E-Commerce Retail

Support

Definition: No. of transactions that include both the products x and y as a percentage of the total number of transactions

Business meaning: It indicates how frequently both the products are purchased together.

Understanding Data Analytics in E-Commerce Retail

Lift

Definition:Lift is the ratio of confidence to expected confidence. Expected confidence is simply the probability that product y is in the basket.

Business meaning: The lift of a rule indicates how strong is the association within the rule, i.e. how better a rule is at predicting the result than just assuming the result in the first place. Higher lift values indicate stronger associations.

Understanding Data Analytics in E-Commerce Retail

Price Optimisation

80% of people say that the most important aspect of shopping at an online e-commerce store is competitive pricing.

Segmentation of Customer & Products

This is the first step in price optimisation, where you’ll group similar products and customers. Classification (supervised) and clustering (unsupervised) algorithms are useful for this segmentation. The best technique will depend on the business requirements and the type of data.

The analysis used K-mean clustering to form groups of products with similar sales trends at Colgate. It had helped to build one predictive model per cluster instead of having many models.

Regression Modeling

Once the segmentation is done, regression model on the various data points around sales, conversion rates, seasonality, product attributes, marketing channels etc. helps to fix optimal prices for the products. Let’s give you a hypothetical business example from the CPG industry. The regression model built by the analytics team at Unilever for Personal Care products can help it to predict that a 5% decrease in the price of Dove deodorant can boost sales by 15%.

Dynamic Pricing

Do you remember the last time the price of your favorite product dropped during the big sale at Amazon, and the other e-commerce stores also dropped their prices?

This is dynamic pricing. It uses competitor data as an input into machine learning algorithms for pricing to determine the product price at different times. This helps the products to remain viable in the competitive market with frequent price fluctuations.

A/B Testing can be used to refine the pricing model, but it needs a lot of technical considerations.

Demand Forecasting

Let’s imagine Benjamin Franklin and Abraham Lincoln to be e-commerce business tycoons for a minute.

Franklin seems to have warned against underestimating the potential of demand forecasting in e-commerce retail sales,

“If you fail to plan, you are planning to fail.”

And Lincoln seems to have provided the perfect strategy for including demand forecasting in the analytics scene at your e-commerce company.

“Give me six hours to chop down a tree, and I will spend the first four sharpening the axe.”

Demand forecasting refers to the use of analytics techniques to predict the demand for a product and forecast sales. When you know the sales trends in advance, it gives you an advantage over your competitors in the following ways:

a) Better Inventory management

An underestimation of inventory levels can lead to inventory getting out of stock and hence cause customer dissatisfaction. An overestimation of inventory stock levels can lead to overstocking of inventory and hence cause unnecessary storage costs.

Demand forecasting helps in informed inventory planning and saves you from both selling out of popular products and warehouse space wastage due to slow-moving stock.

b) Better cashflow management

Since the money is not tied up in slow-moving inventory, It helps you plan the budget properly and use the cash optimally. Thus, it helps you to reduce financial risk.

c) Better pricing strategy

Effective demand forecasting also helps in improved pricing strategy. You can charge more for the products having a high demand forecast, and vice versa. You can better plan the marketing budget, advertising investment, and discount plans.

d) Better customer service

The prior knowledge of the product demands and their fluctuations helps in planning better customer service. It’s time for a business example.

Amazon patented ‘Anticipatory Shipping’ in 2013 to use predictive modeling to ship products to you before you would even make the purchase. Amazon’s algorithms would predict your purchase, and it would send the product to your nearest warehouse. Then, Amazon would wait until the actual purchase. Amazon believed that this would drastically cut the delay from the time of purchase to the actual delivery and boost the overall company sales.

SwiftERM combines the best aspects of predictive analytics in product selection on an individual consumer by individual consumer basis to maximise their CLV and AOV. No two people are alike, nor are their purchases so nor are the offerings made to them. Our Case-Studies illustrate an average ROI of 1500%.

 

We hope you enjoyed this article, intended to help improve our client’s profitability. It reflects the care SwiftERM offer. If you haven’t already done so, then please enjoy a FREE month’s trial and let us know what you think. Register, call us on 0207 998 3901, book a call with us https://calendly.com/swifterm/15min or Zoom ID 964 515 7464 

 

Other articles of interest below:
(Index to all articles here)

How to create an authentic brand identity for your business
Personalisation Drives Retention and Monetisation
Why data-driven marketing is so important