DXi: Optimizing Sales Conversion Rate

We have a solution that can handle data directly from Google Analytics, Adobe Analytics, or structured data provided by client systems. Initially, we cleanse and normalize the data, removing outliers. The cleansed data is then processed using a series of 10 machine learning algorithms to reduce the number of metrics that comprise it while making sure that the importance of the information remains intact. Our solution uses a proprietary algorithm to compute the final DXi score by dynamically attaching different weights to the individual input parameters based on the activity of each individual user on the website or the mobile app. The algorithm considers things that will be very familiar—such as time spent on the website, the number of products searched, the number of items in their purchase cart, and of course, whether the individual visitor actually made a purchase. Once the DXi score is calculated for every visitor, we compute the overall average of the entire website. We then feed the final DXi score into advanced machine learning algorithms such as random forests, blending, decision tree classifiers, data boosting, and Deep Learning algorithms to garner deeper insights about the client’s website and their customers that would not be apparent using standard analytical methods to look at the data. From these insights, we provide prioritized steps (prescriptions) to improve DXi scores and conversion rates. By using a single quantified score and a variety of algorithms, we easily identify the key aspects that are influencing the business outcomes.

For example, a reduction in the client’s DXi score over a certain period of time would suggest lesser sales due to low engagement with the visitors. But, if the DXi score goes up over a period of time and sales still don’t improve, we can conclude that there are other issues. These issues can be anything from low quality of products, to returns, to shipping delays that allow customers to be engaged by the website but not make a purchase. This could also indicate increased market competition. Using one single quantified score, we easily identify the key aspects that are influencing the business outcomes.

Can you share with us a customer success story?

One of our Canadian customers who sell industrial equipment wanted to revamp its local sales. To do so, they offered a rebate of 25 percent or more on their products. They were on a mission to increase local sales by 0.3 percent. We setup DXi in their system to correlate our DXi score computing process with their conversion rates. We analyzed their website and found that their customers were not looking for discounts; rather they wanted to learn more about the products. Having analyzed their DXi score, we asked them to optimize their website and make it more informative and then offer variable discounts reduced to 15 percent instead of 25 percent. Over three months, we found out that despite lower discounts, their sales and conversion rates had doubled.

In another instance, for a large online shoe company with close to one million website visitors in a month, we were able to identify the reason for low conversion rate. We correlated their DXi score with the functionalities of their website and found that the search function on their website was not easily accessible, and this was interfering with sales.

What does the future hold for DXi?

We are aiming to make our systems compatible with unstructured data sets like audio, video, and images. We are also taking initiatives to integrate our technologies with our solutions for social media channels to improve the process of DXi scoring.