Symphony of Variables Impact

Nothing at work is more exciting to me than to bring data driven insights that can drive clarity into business opportunities and strategy. Last week I had a unique opportunity to consult with an amazing company in the valley. During that process, I found myself asking how to better understand changes in predictor variables and how they will impact the dependent variable. It confusing, at least to me, to make sense of the relationships; especially when there are more than five statistically significant variables. 

Something like this with a few more variables -

[0] next to variables represents low activity. The data provided ordinal measures for low [0] and high [1] account activity. 

Its pretty clear that predictors such as Offer and Card have a negative revenue impact on low activity accounts, meaning high Offer and Card activity will provide a positive impact on revenues. Similar logic can be attributed to the variable Check[0]. Insights into these relationships will help us drive clarity into the business needs to optimize the revenue stream. 

Answering questions such as “What will be the revenue impact if the card activity increases?” or “What is the impact of a promotional offer on revenues?" can become overwhelming. 

Translating these questions to tech marketing, questions came to my mind like: How many pages per session will optimize the conversions rate? What is the optimal average session duration to generate revenues? How many new users do we need for sustainable growth? Which channels should be best leveraged to drive traffic? What are the best assisting channels to drive the conversions? What metrics are important to track? What does this segment wants from the product?

All of these questions are driven by the fact that we are in the midst of a grand symphony with numerous variables all playing their chords. The question is how to be a grand chorister that can better understand the changes in predictor variables and their impacts on the dependent variable. Surely, an exhilarating trek!

To answer my original question "how to better understand changes in predictor variables and how they will impact the dependent variable”, I found that Prediction Profiler in JMP, with its Microsoft Excel addin, very useful to visualize the changes . Surely, there are many analytics tools that will provide something similar to this. 

Appendix - 

Here are the answers to our questions from this example - 

1. Impact of the offer? 

Answer: The promotional offer increased revenues from $11.08 to $12.75

2. Impact of the card activity?

Answer: With increased activity in the card, we are able to increase the revenues from $54.5 to $62.70.

The goal was to see the impact of changing predictor variables on the response. As always the immediate audience is myself.