Screen Recording 2023-01-22 at 6.48.31 PM.gif

Introduction

Crux intelligence is a Business Intelligence tool that helps its customers make better decisions. The biggest way people consume data is through pictorial representation aka data visualisations. So, I was super excited to get my hands and own the data visualisations at Crux intelligence. As a product we allowed users to analyse any sort of number and be able to find the wheres and whys of said number.

What is the problem?

Let’s say you’re a company that sells water bottles. And a specific blue water bottle you sell has completely gone down📉 in sales this month. As somebody that is consuming this information. Just knowing what is happening is not enough. You also want to why it’s happening so you can do something abut it right?

To aid the user do that, Crux had Drivers

Drivers 1.0- Correlation (the existing solution)

Untitled

What are drivers?

Drivers helps you understand what are the factors due to which your sales went up or down. And Crux uses “correlation” to help you figure out why your sales went up or down. The way it does this is by telling the user how a “factor” is related to your sales. It assigns each factor a score from -1 to +1. The closer the score is to 0, the less related they are. The closer the score is to 1, the more related they are. And the “+” means they’re directly related and “-” means they’re inversely related.

For instance, in this example inflation has a medium negative relation with Sales. Meaning, Sales go down as inflation goes up. And Instagram ads has a high positive relation with Sales. Meaning, Sales go up as you spend more money on Instagram ads

Untitled

Here’s a graph to explain it better.

But there are some problems with this :

  1. Correlation and the score is nearly not as intuitive as we thought it was. Users during our User research rarely ever looked at the score. Even when they did- they didn’t get what it meant. Or worse - they understood it and didn’t know what to do with the information. Sigh.
  2. Correlation sometimes didn’t mean anything when the graph fluctuated too much. Why does this matter? Because a parameter can still have an impact on your sales even if it’s not highly correlated. And it didn’t turn up in correlation at all.

So the solution - Chuck correlation and bring something simpler

Untitled

Our PMs and data scientists got together and came up with Key driver analysis.

Drivers v2.0

What is Key drivers?