More than 20 years ago, e-commerce was just a novel concept until Amazon sold their very first book in 1995. Nowadays, e-commerce market is a significant part of the world’s economy. The revenue and retail worldwide expectations of e-commerce in 2019 are $2.03 trillion and $3.5 trillion respectively. This market is developed and diverse both geographically and by business models. In 2018 two biggest e-commerce markets were China and the United States with revenue $636.1 billion and $504.6 billion respectively, while Asia Pacific shows better growth tendency of e-commerce retail in relation to the rest of the world. Companies use various types of e-commerce in their business models: Business-to-Business (B2B), Business-to-Consumer (B2C), Consumer-to-Consumer (C2C), Consumer-to-Business (C2B), Business-to-Government (B2G), and others. This diversity is since e-commerce platforms provide ready-made connections between buyers and sellers. This is also the reason that B2B’s global online sales dominate B2C: $10.6 trillion to $2.8 trillion.
Rapid development of e-commerce generates high competition. Therefore, it’s important to follow major trends in order to drive business sales and create a more personalized customer experience. Big data analytics is such a trend, although for many companies, data science techniques have already become customary tools of doing business. There are several reasons for the big data analytics efficiency:
· Large datasets that makes it easier to apply data analytics;
· High computational power of modern machines that allows to make even real time data-driven decisions;
· Well-developed methods of data science.
In this article we are going to show the impact of using data science in e-commerce and the importance of data collection, starting from the initial stage of your business.
Market Segmentation Analysis
Strategic marketing planning is the core of the development of any business. Market segmentation is one of the key building blocks of strategic marketing. In marketing segmentation, customers are subdivided into groups based on their attributes (age, gender, etc.). The aim of the segmentation analysis is not only to identify different market segments, but also to develop a detailed profile, description and nature of each segment selected for targeting. Such an analysis could be done using data-driven analytics. The benefits of segmentation are obvious:
· Company could find segments that better fits to its strong sides at compared to competitors.
· Company could identify what is work for particular segment and what doesn’t, the last one is more important, and then improve product or service.
· Company could better match of its products or services to each segment needs.
The advantage of analysis based on data science methods is that it is possible to analyze a variety of data. Earlier, segmentation was based on geographical data (country of origin, address), demography (age, gender, education). Now, we add to our analysis psychographic criteria (beliefs, interests, aspirations, etc.) and behavioral data (based on customers activity: amount spent, frequency of purchase, etc.). This makes analysis more complex but more detailed and precise. For example, instead of just “young females” or “high-educated French males” segments, data-driven segmentation gives “fashionista young females, who enjoy fine dining” or “auto addicted male, who play golf, enjoy wine-tastings” segments. So, the company knows better their customers and recognize more segments, up to 10-20. This is called micro-segmentation. Micro-segmentation is the basic key to understanding the motivation drivers behind customers actions (for example why customer abandoning its basket). The micro-segmentation-based management includes different dimensions, such as visit time, device preferences, that allows to make predictions about customer behavior.
One of the hottest trends in e-commerce is a recommendation system. The recommendation engine filters the information and predicts customer’s preferences in order to come up with relevant products. The filtering of information is personal for each customer and based on his past searches, other customer’s searches, purchase data. Of course, processing of huge amounts of data coming from hundreds (or even millions) of users needs modern artificial intelligence techniques.
There are two categories of recommendation algorithms: collaborative filtering and content-based filtering. First one uses data collected from the user’s activities on the website in order to find similarity with other users’ activities. For example, customer is looking for a new phone. The recommendation system will propose to buy a phone cover too, because other users often buy phone and phone cover together.
Content-based filtering finds recommendations based on comparison between the content of the product and a user profile. For example, customer likes jazz music. So, the recommendation system on music store website proposes to user the hottest jazz album. There is also the combination of two above filtering types, called hybrid recommendation filtering. Usually, we can take full advantage of the hybrid technique when there is more information about users. In hybrid filtering, predictions of one technique are used as input for another technique.
Websites with recommendation engine have powerful advantage compared to that ones who haven’t: they can recommend their customers what customers didn’t know they wanted. Amazon says its recommendation system generates a 20-35% of the company's revenue.
Customer Lifetime Value (LTV) Modeling
In any business it is very important to focus marketing efforts on the right channels. Machine learning methods allow to predict customer lifetime value (LTV). LTV means the value that each customer will bring to the company’s revenue. Model uses information about customer’s needs, expenses, recent purchases in order to make an estimation of how long a person has remained a customer and to predict his future purchases. Such analysis helps e-commerce businesses in several ways:
· Identification and care for high-value customers.
· Estimation of future sales and defining growth.
· Adjusting campaign and advertisement.
Churn model is a commonly used model that helps to achieve customer retention. New customers engagement is good, but it is more expensive than maintaining relationships with existing customers. Customer churn prediction discover customers who are risky to leave and helps to understand reasons for that. Understanding of retention reasons helps in generating higher customer lifetime value. For example, the company can improve its services or change marketing strategy by analyzing retention reasons (that’s also helps in attracting new customers).
Machine learning algorithms are also used for pricing optimization. Optimal price is not simply consideration of the demand: when the demand is high the prices increase and decrease when the demand is low. In e-commerce it is important to account for many factors at once: competitors’ prices, time of day, attitude of customer, warehouse stock, season. Such an approach will allow the company to have dynamic prices in order to win sales every time. For example, you can offer the best price for some products or send coupons for some customers to achieve customers’ tolerance. For some customers, it could be an exciting game in which they can try to hunt the best prices. Dynamic pricing brings significant results in the short term perspective: 2-7% increase in business margins and a 200-350% average growth in ROI over a year.
One of the cheap possibilities to improve customer service is using of chatbots. Chatbots are based on AI. They use natural language processing (NLP) or speech recognition, in the case of phone calls, to interpret users’ questions and respond to them. They can communicate with a customer, identify an issue, and resolve it.
Customers sometimes are frustrated in site navigation: when they want to find a product page, FAQs, sales. Chatbots can help to find, choose and make an order of the product the customer is looking for. Also, chatbots carry out through transaction process. The advantages are obvious:
· Answering questions 24/7
· Providing realistic conversation
· Giving higher satisfaction to customers
E-commerce is a field with high scammers activity caused by close connection to payments. There are many fraud schemes in e-commerce, like chargeback, unauthorized discounts, unauthorized sale voiding, returns, and many others. Thus, AI looks like the best solution for fraud prevention.
Fraud detection system collect and analyze historical data of customers in order to learn normal customer behavior, such as typical customer’s devices, the time, the location. The abnormal account activity may indicate, for example, that account or credit card of a customer was stolen. Also, AI can detect subtle behavioral patterns of scammers and identify persons that abuse the refund policy. This is quite known kind of scam when fraudster orders a product, then returns a fake one.
Big data brings competitive advantages for e-commerce companies. Big e-commerce organizations like Amazon, eBay, Netflix have long felt a positive effect of using data science methods. Small companies find in data science cheap solutions to increase their management effectiveness.