21 Feb Part – II: Quantifying Digital Experience using Machine Learning
The author Dr.Srinivas Kilambi is the Founder and CEO of Sriya DXI LLC
Last week we talked about how current analytics techniques to measure DX are very subjective, inconclusive, non quantitative and need data scientists for interpretation and rarely have any correlations with sales, ROI or other growth metrics (https://www.linkedin.com/pulse/quantifying-digital-experience-dx-index-dxi-using-machine-kilambi?trk=prof-post). This week we will present an alternate methodology, “DXi”
Ideally, we should consider only the data generated during an individual’s visit to a website, social media site, or when they use an app. And this should not be aggregated, but kept in individual records to allow correlation of specific experience parameters with others of the same individual. From this “small data”, we stand a much better chance of getting results that are more definite rather, and which can be understood by a CEO, CMO or a CIO/CFO without long explanations by data scientists.
We need a quantitative technique which will convert this very subjective “DX” into an “objective number (DXi)“. This allows the compaction of zillions of big data, 100s of input DX parameters into 1 output number DXi. DXi is easily usable, comparable, transportable and understandable by one and all. A score of 120.13 this week vs 136 last week shows an approx, 13% drop which can be understood by CEO, CMO, CIO alike. This will allow for some analysis leading to action items for improving the score. DXi can also be correlated to sales or its surrogate and hence can be a measure of company’s success. “Higher the DXi, higher could be sales and hence profitability”.
This unique method not only gives a broad index scores of user DX on a web site but also sub scores namely Visual, Transaction, Behavior and KPI DXis and hence we can identify where to improve. For ex in the image above the behavior sub score was 93 compared to 145 for visual. So focus on parameters which affect behavior of an user on the web site like Total Page Views, Page views per visit, total time on all pages, total time per session and 1 derived parameter namely Bounce rate. This methodology also gives the weight of each parameter into the score and hence improve on those parameters with the most weight-ageas they have the maximum impact on the final score.
So how do we do this? By creating action items which when implemented can improve the high weight-age parameter score and hence the sub score and finally the overall DXi. Increase in overall DXi should result in higher sales and higher profitability. These action items focus on improving the top parameters selected by ML algorithms and decision trees.
“I will write about the need for decision trees and their benefits in the next post”