Someone tweeted a worry which involved the analysis process. It’s the same for most people and me, i.e. we started creating beautiful and visually appealing reports. Scrap off the beautiful for me as I was horrible at designs, so I always kept my report as simple as I could. His tweet is the genesis of this article.
Over the years, I have gone through unlearning and learning.
So what is my approach to analysis?
My approach has always been top-to-bottom, i.e. I start with an end goal in mind.
Let’s see this in a real-life scenario. To show this, I will be using the conventional SuperStore Data.
Before I think of the data, I define a clear objective. Why am I carrying out this analysis? What do I hope to achieve? Below is our objective
Aim: Revenue Growth
Next up is how I can use data to achieve my objective. To do this, you need to be able to raise questions your data will give you answers to.
An analyst's greatest tool is his/her mind. The data is there to help you answer questions and then help you raise more direct questions which would be answered with additional data
So what kind of questions can we ask the data that aid in meeting our objective?
I would say two things that affect revenue growth are;
Customer activity, i.e. how many monthly active customers, do we have, how many new customers are being generated by the marketing leads, do we have customers who buy a single product multiple times, do we have customers who purchase one and any other different product, etc.?
Product performance, i.e. which products are performing badly/well, and what is the year-on-year performance of those products?
If I could answer those questions with data, I’ve gone a step closer to achieving my objective because I would have been able to identify the bottlenecks.
Let’s go into PowerBI to answer these questions
We are working with a straightforward data model.
How many active customers are present?
To answer this, we will use the following DAX measures.
And then plot a line chart to see the performance.
A look at the chart above, we will notice most of our high points falls between Nov and December, i.e. there are more customers. Similarly, there is always a decline in the no. of active customers during October.
Remember earlier on; I said data lets you ask more refined questions. Just looking at the customer activity chart, questions I will note include.
Does the company have seasonal products?
Does the company have more inventory in those months?
Is a specific marketing effort being carried out during those two months that is not done in other months?
For October, is there always a decline in the company’s inventory? Hence, customers have fewer products to purchase
The questions above will help me further understand why customer activity is peaking between Nov and Dec, and that is what Diagnostic analysis is, i.e. understanding why something is happening. Understanding why the company has more customers in Nov and Dec which aid in giving better recommendations.
Note that additional data can help you answer the questions above.
How many new customers are being generated by the marketing leads?
The following DAX measure is used to know who the new customers are;
Note that an assumption is made that every customer that didn’t purchase a product for 90 days should be regarded as a new customer.
Above is the chart showing the new customers trend. We could see both a steady and sharp growth each year. Kudos to the Sales/marketing department; they are performing well in lead acquisition. But the chart also shows October as a problem; there is a sharp decline which would lead me to raise more questions.
Is marketing efforts relaxed during this time?
Still, on new customer activity, another set of questions to ask is what percentage they contribute to the total revenue.
To achieve this, we create the following set of measures.
Let’s see the numbers in a chart.
I prefer a % as it is way easier to picture in mind. Also the % of new customers. To get this, we use the following DAX Measure.
Let’s plot a chart to see the performance.
From the two charts above, new customers not only constitute the majority of the active customers, but they also contribute more to the total revenue.
Let’s all clap for the marketing department once more.
But for revenue, we not only need new customers. We also need returning customers. This leads to another set of questions;
What marketing strategy is currently being deployed, is it targeted at customer acquisition?
What channel is used to ensure there is no communication bridge between the company and customers?
Is there customer satisfaction/review data, as this will help us understand how customers feel about purchased products?
Customers who bought a product more than once
The DAX below will let us get an answer to that
Let’s see the numbers in a table.
From the table above, very few customers return to buy a product they have bought. The customer review data, if gotten, will help us further understand what customers think about the product they have bought.
Product Performance
Another metric that affects revenue is the performance of the performance. Because there is a very list of products, we will deal with the product category and sub-category.
Below is the required DAX
Let’s see the numbers.
From the table above, most of the products showed considerable appreciation in revenue. For those products with very little/no appreciation in income, one thing that can be done is to give discounts on products. But before I recommend discounts, two things I do is;
Take my analysis one step deeper to understand why revenue is low for that particular product. Check out how I do that here.
Carry out a scenario analysis. Giving discounts on the right product is as important as the discount itself. Check out how I do that here.
And that, guys, is my approach to EDA.
Comments