Hi, I’m Abdul Saboor
a Generative AI Specialist ML Engineer BI Developer Product & Innovation Lead

The Situation: 

Adventure Works is a fictional global manufacturing company that produces cycling equipment and accessories, with activities stretching across three continents (North America, Europe, and Oceania). Our goal is to transform their raw data into meaningful insights and recommendations for management. More specifically, we need to:

• Track KPIs (sales, revenue, profit, returns)
• Compare regional performance
• Analyse product-level trends
• Identify high-value customers

The Data: 

We’ve been given a collection of raw data (CSV files), which contain information about transactions, returns, products, customers, and sales territories in a total of eight tables, spanning from the years 2020 to 2022.

The Task: We are tasked with using solely Microsoft Power BI to:

• Connect and transform/shape the data in Power BI’s back-end using Power Query
• Build a relational data model, linking the 8 fact and dimension tables
• Create calculated columns and measures with DAX
• Design a multi-page interactive dashboard to visualize the data in Power BI’s front-end

The Process:

1. Connecting and Shaping the Data

Firstly, we imported the data into the Power Query editor to transform and clean it. The next process involved:

Removing Duplicates: Duplicate entries were removed from the dataset to ensure accurate analysis.

Handling Null or Missing Values: For some columns, missing values were replaced with defaults or averages. Null values in “key” columns were removed using filters.

Data Type Conversion: Columns were converted to appropriate data types to ensure consistency. Dates were converted to Date type, numerical columns to Decimal or Whole Numbers, and text columns to Text.

Column Splitting and Merging: Several columns were split to separate concatenated information, or merged to create a unified name (such as Customer Full Name).

Standardising Date Formats: All date columns were formatted consistently to facilitate time-based analysis. This step was important for ensuring accurate time-series analysis in Power BI.

Removing Unnecessary Columns: Irrelevant columns were removed to streamline the dataset. This helped focus the analysis on relevant information, reducing memory usage and improving performance.

2. Building a Relational Data Model

Secondly, we modeled the data to create a snowflake schema. This process involved creating relationships between the dimension and fact tables, ensuring cardinalities were one-to-many relationships.

Enabling active or inactive relationships, creating hierarchies for fields such as Geography (Continent-Country-Region) and Date (Start of Year-Start of Month-Start of Week-Date), and finally hiding the foreign keys from report view to ease the data analysis and visualization steps and reduce errors.

3. Creating Calculated Columns and Measures

Next, we used Power BI’s front-end formula language, DAX, to analyze our relational data model and create several calculated columns (for filtering) and measures (for aggregation), that we could later reference and use when analyzing and visualizing the data.

We used calculated columns to determine whether a customer is a parent (Yes/No), a customer’s income level (Very High/High/Average/Low), a customer’s priority status (Priority/ Standard), and the customer’s educational level (High School/ Undergrad/ Graduate).

The list of calculated measures is available below and includes key information on revenue, profit, orders, returns, and more.

4. Visualising the Data

The final step of the project was creating a multi-page interactive dashboard, including a range of visuals and KPIs that could serve management and lead to informed decision-making. We used several visuals and tools to demonstrate and visualize the data across the 4 report pages, including KPI cards, line and bar charts, matrices, gauge charts, maps, donut charts, and slicers. We made sure the report was fully interactive and simple to navigate, with icons used to enable filters, cancel filters, and guide users to each report page with ease. Features such as drill-through, bookmarks, parameters, and tooltips were also used throughout the dashboard, further enhancing its usefulness and impact on management.

Executive Dashboard: The first report page provides a high-level view of Adventure Works’ overall performance. We used card visuals to present Key Performance Indicators such as overall revenue, profit margins, total orders, and return rates. We also included additional cards to compare current and previous month performances, providing insights into recent trends, a line chart to visualize the trending revenue from 2020-2022 and highlight long-term performance, and presented the number of orders by product category to aid in understanding product sales distribution, and used a further table to display the top 10 products based on key indicators (total orders, revenue, and return rate).

Map: The second report page consisted of a map visual, an interactive representation of sales volume across different geographical locations. This offered insight into Adventure Works’ global sales distribution and worldwide reach.

Product Detail: The third report page focuses on detailed product-level analysis. It displayed detailed product information for the selected top 10 products from the Executive Dashboard, using the drill-through feature. It also included gauge charts presenting actual performance vs target performance of monthly orders, revenue, and profit, and included an interactive line chart to visualize potential profit adjustments when manipulating the price of the product, aiding in strategic decision-making regarding pricing strategies. This report page also included a line chart including key weekly product information on total orders, revenue, profit, returns, and return rate.

Customer Detail: The fourth and final report page provided a deeper insight into customer behavior and value. It used donut charts to break down customer groups into income level and occupation categories vs. total orders, helping in customer segmentation tactics, and used a matrix aided by KPI cards to identify high-value customers based on order and revenue contributions, aiding in identifying high-value customers and sales opportunities.