[Project] Discount Strategy Analysis: When Discounts Start Destroying Profit (Python + Power BI)

Why I worked on this project

Discounts are one of the most common tools used in retail to increase sales.

Almost every company uses them.

But the question that interested me was simple:

At what point do discounts stop helping sales and start damaging profitability?

In many organisations, discount decisions are made quickly - often as part of marketing campaigns or short-term sales targets.

However, without analysing the relationship between discount levels and profit, it is easy to unintentionally reduce overall business performance.

This project was my way of exploring that relationship using data.

The business problem

Retail businesses frequently use discounts to stimulate demand and attract customers.

But excessive discounting can significantly reduce profit margins.

From a business perspective, several questions become important:

• Are higher discounts actually increasing profitability?

• At what level do discounts start generating loss-making orders?

• Is there a discount threshold that maximises total profit?

  
  

Answering these questions can support more effective pricing and promotional strategies.

How I approached the analysis

The analysis started with exploratory data analysis using Python.

I examined how different discount levels relate to profitability across thousands of transactions. This included identifying loss-making orders and analysing the overall distribution of discounts and profits.

To better understand the relationship between discount levels and profit outcomes, I used regression modelling and scenario simulations.

The simulation allowed me to test different maximum discount limits and estimate their potential impact on total profitability.

Once the analytical results were clear, I designed a Power BI dashboard to communicate the findings in a more accessible and business-friendly format.

The goal was not just analysis, but clear interpretation.

What the analysis revealed

Several patterns became visible during the analysis.

First, a surprisingly large proportion of transactions were actually loss-making.

Approximately 24.5% of orders generated negative profit, which immediately raised questions about discount strategy.

The data also showed a clear relationship between higher discount levels and decreasing profitability.

While small discounts had a limited impact, larger discounts were strongly associated with negative profit outcomes.

Using scenario simulation, I tested different discount cap strategies.

The results suggested that limiting discounts to around 20% could maximise overall profit.

Beyond that point, the increase in sales volume did not compensate for the loss in margins.

Why this matters from a business perspective

Without analysing discount behaviour, companies may unintentionally encourage strategies that increase revenue but reduce profitability.

This type of analysis helps businesses understand:

• where discounting becomes financially risky

• which promotional strategies actually improve results

• how pricing policies influence long-term profitability

  
  

In real business environments, these insights can support more balanced decisions between growth and margin protection.

Tools used

Python (Pandas, Seaborn, Statsmodels) for data exploration and modelling

Power BI for dashboard development and business-friendly visualisation

The focus of the project was not on complex modelling, but on understanding the relationship between pricing decisions and profit outcomes.

Final thought

What I found most interesting in this project is how easy it is to focus on sales volume without fully understanding what happens to profit underneath.

Discounts may increase revenue, but without careful analysis they can quietly erode margins.

This project reinforced something I keep seeing while learning data analysis:

good business questions are often more valuable than complicated models.

Want to see the technical side?

If you’d like to explore the full analysis, including the Python notebook, dataset and Power BI dashboard, the complete project is available on GitHub.

The repository includes the full analysis workflow, visualisations, and the Power BI dashboard used to communicate the results.