Applying Predictive Analytics for E-commerce Apps

Problem Statement

Basic e-commerce web apps developed in open source technologies such as HTML, JavaScript, CSS, PHP and web frameworks are good for the development purpose although more needs to be done to make the sale. Additionally e-commerce application have a design,  layout and  information architecture as per standards. Newer techniques also ensure correctness of information for the buyer so there is no ambiguity and product purchase decision making is simplified. In addition to these decision aids in customer is more interested in reviewing and comparing ensuring the quality aspect, differentiation, uniqueness and more importantly the value proposition. Some customers will also demand the most efficient logistics and delivery options.

For most e-commerce web applications today a product that ensure best value for money, ready availability are key solution areas to work on. Its found that to further improve upon the sale depends on not only the product value but also the "better experience" of the e-commerce application sale process. For this purpose the development of e-commerce application need to apply advanced data techniques and analytics to improve the customer delight.

If the product is just cheaper on a e-commerce store but no assurance about its quality then the purchase experience is poor and value of the product delivered is much lower than what the manufacturer intended. Hence better purchase experience if can be achieved it also ensures better recommendation to other buyers and favorable discussion with friends and social media reviews.

The Solution

We solve the customer experience problem by applying techniques to understand and know the customer. This is done by collecting two types of data:

• Individual user usage data
• Success / Recommended usage data

Its possible to collect the user data to improve the users experience. This is done by adding some database tables that will help in collect user data (with user consent). In addition to the tables code need to be added to update these tables within the web application using JavaScript event handlers. 

Following type of data is collected in various data tables:

• Mouse clicks and Hoover data
• Click data
• Success data from sale that was successful.

As you see the schema take into account each page and possible the navigation where the customer has visited whether the product was purchased is not an important. By adding event handler hooks to all events all the data is now available for further analysis and understanding the pattern

Applying Predictive Model

As per the definition Predictive modeling is a process used in predictive analytics to create a statistical model of future behavior. Predictive analytics is the area of data mining concerned with forecasting probabilities and trends.

Refer to this link for more information. 

For this particular problem we apply the Logistic regression technique. Refer to this link for definition.  
Logistic regression helps us identify the relationship model. By using the  the page visits and the success pattern of page visits leading to the sale it can be dynamically estimated in probabilities what can be offered to customer by application of  the logistic function. Logistics function constants can be dynamically updated based on the different and changing Success Data. E.g. if a certain product sale is completed (starts trending) it can be highlighted to the user. It should be noted that there are many algorithms that can be applied hence it depends on the data. In addition based on the recent customer sale and pattern a specific product can be recommended. Many such features are possible and can result in a dynamic dashboard for e-commerce portal.

Based on the requirements and business analysis the collected data can now be used for validating strategies and further analysis for adding more value to specific customers and also to improve and add more features to the product. With these techniques its possible to improve more and more on customer experience. These technique as explained will also ensure repeat customer visits.