Case Study in Marketing Analytics in Digital Marketing

A marketing analytics case study looks very different on paper than it does in real life, and I have spent 12 years sitting with students while they discover that gap. Most guides show you a clean formula and a tidy example. Real dissertations are messier. This is the story of one student, one spreadsheet, and one regression model that taught him more than any textbook chapter could.

Quick answer: A marketing analytics case study uses real business data, usually sales, pricing, and marketing spend, and runs it through a statistical method such as regression to show which factors actually influence outcomes and by how much. Below is a real dissertation example with actual regression numbers, including a low R-squared that ended up strengthening the dissertation rather than weakening it.

How I Approach Every Marketing Analytics Case Study

I have mentored many students over the years through their dissertation analysis chapters. In that time, I have formed a strong opinion about most content available online. Search for regression analysis in marketing and you will find dozens of guides walking through the same tidy textbook example. It is always a clean scatter plot, a suspiciously high R-squared, and a coefficient that behaves exactly the way economic theory predicts. My honest view, after years of reading these guides alongside actual student data, is that they teach the formula and then abandon the student. That happens the moment real data gets messy, which is almost always.

That gap is exactly where students get stuck. They know the maths. They do not know what to do when their own data refuses to behave.

My approach is simple. I do not hand a student a finished model. I sit with them while they build it, get confused by it, and eventually explain it back to me in their own words. That is when I know the learning has actually happened.

Meet James: A Marketing Analytics Case Study for Beginners

James is a composite drawn from students I have mentored over the years. Details have been changed to protect privacy, but the numbers and the teaching moments are true to how these conversations actually go.

James was a final-year marketing student at a UK university. His dissertation topic sounded straightforward: what drives sales for an online retailer. In practice, he was stuck with a spreadsheet full of numbers and a supervisor telling him to show “analytical rigour” without explaining what that meant in practical terms.

The Problem: Data Without Direction

James had collected quarterly data on three things:

• Product pricing
• Advertising spend
• Sales figures

This was existing company data rather than something he had gone out and collected through a survey. Before we touched the numbers, he needed to understand the difference between primary and secondary data to defend his methodology chapter properly. His figures were also tracked over several quarters rather than captured at one point in time. That made this a longitudinal dataset rather than a cross-sectional one, a distinction that mattered later when we checked his results for autocorrelation.

He had the raw material. He had no idea how to turn it into a story his examiners would respect.

The First Conversation: Finding the Research Question

I asked James one question first. What is he actually trying to find out? He said he wanted to know what makes customers buy from an online retailer. That single sentence gave us our dependent variable and pointed us toward our independent variables.

His dissertation was also structured around one retailer’s data rather than a broad multi-company survey. Understanding how a case study design differs from a survey design early on kept his methodology chapter consistent instead of mixing two research approaches together.

This step gets skipped too often. Students jump straight to running tests before they have pinned down what question the test is even answering.

Building the Regression Model: How to Use Regression Analysis in Marketing

Once James had his research question, I introduced him to regression analysis properly, not as a formula to memorise, but as a tool with a job to do. This is really how to analyze marketing data with regression in practice. You define an outcome, decide which factors might be driving it, and let the equation show you the relationship instead of guessing at it. Simple linear regression looks at one predictor and one outcome. Multiple regression, which is what James needed, looks at several predictors at once.

Choosing Dependent and Independent Variables

We defined the roles clearly:

• Dependent variable (what we are trying to explain): Sales
• Independent variables (what might be driving it): Price and Advertising Spend

The Regression Equation Explained Simply

Here is a definition worth remembering: regression analysis is a statistical method that measures how strongly one or more factors are associated with a particular outcome. For James, the equation looked like this:

Sales = a + b₁ × Price + b₂ × Ad Spend

Here, a is the baseline sales figure when both price and ad spend are zero. b₁ shows how sales shift when price moves by one unit, and b₂ shows how sales shift when ad spend moves by one unit.

James called it a recipe, where each variable is an ingredient contributing to the final result. That is one of the better ways I have heard a student describe it, and I still use that analogy with new students today.

Step by Step Marketing Analytics Case Study: Running the Analysis in Excel

Not every student has access to SPSS or R in their first year, and that is fine. This is also a practical example of how to use Excel for marketing analytics without needing paid software.

1. Open Excel and enable the Data Analysis ToolPak add-in
2. Enter Sales as the Y range and Price plus Ad Spend as the X range
3. Run the Regression tool from the Data Analysis menu
4. Locate three numbers in the output: R-squared, the coefficients, and the significance values
5. Note down each coefficient’s sign, positive or negative, before interpreting the size

If SPSS is available at your university, an SPSS data analysis tutor can walk you through the same output with cleaner formatting and built in diagnostic tests.

Interpreting the R-Squared Value

James’s model returned an R-squared of 0.0839. In plain language, price and advertising spend together explained only about 8.39 percent of the variation in sales.

James was disappointed. I was not. A low R-squared is not a failed model, it is a finding. It tells you that other factors, such as product quality, seasonality, or competitor activity, are also driving sales, and that becomes your limitations section.

Interpreting the Price Coefficient: A Real World Example of Regression Analysis in Marketing

The price coefficient came out at -7.71. This means that for every £1 increase in price, sales fell by roughly £7.71 million in James’s dataset. Readers working in dollars can apply the exact same logic to their own currency, the relationship is what matters, not the currency symbol.

That negative relationship is not a coincidence. It lines up with the economic principle of price elasticity of demand, a well established concept in economics teaching. That gave James something concrete to discuss with academic backing rather than guesswork.

Interpreting the Ad Spend Coefficient: How to Predict Sales Using Regression Analysis

The ad spend coefficient was 1.07, meaning every £1,000 in additional advertising was associated with roughly £1.07 million more in sales. James’s next question was sharp: why is this effect so small compared to how much was spent?

That question became his discussion chapter. We talked about diminishing returns, the quality of the advertising itself, and the possibility that advertising alone cannot carry sales without other supporting factors. This is also a genuine regression analysis for e-commerce sales, since James’s dataset came directly from an online retailer.

James’s regression results at a glance:

Metric	Value	What it told James
R-squared	0.0839 (8.39%)	Price and ad spend explain a small share of sales; other factors matter too
Price coefficient	-7.71	Higher prices were linked to lower sales, consistent with price elasticity
Ad spend coefficient	1.07	More advertising was linked to more sales, but the effect was modest

How to Interpret Regression Results in Marketing for Your Dissertation

Numbers on their own do not write a dissertation chapter. Interpretation does. Here is the structure I gave James, and it works for most business dissertations of this kind.

1. Introduction – state the research question and hypothesis clearly
2. Methodology – explain why you chose multiple regression, list your variables, and describe your data source
3. Analysis – present the equation, the R-squared, and each coefficient
4. Interpretation – explain what the numbers mean for the business, not just what they mean statistically
5. Limitations – be honest about what the model does not explain, and suggest other variables worth exploring

If your model involves more than two predictors, it is also worth checking for multicollinearity using a variance inflation factor test before you finalise your coefficients. Students who skip this step sometimes get coefficients that look strange for reasons that have nothing to do with their actual research question.

The Result: From Spreadsheet to Story

James submitted a strong 2:1 dissertation. What mattered more to me was that he could explain, without notes, what regression analysis was, why his R-squared was low, and why that was not a weakness in his work. He told me afterwards that I had turned a spreadsheet full of numbers into a story he could tell. That line has stayed with me because it captures exactly what dissertation mentoring should do.

What This Marketing Analytics Project Example Teaches Every Student

If you take one thing from James’s marketing analytics project example, take this: the goal is never a perfect model. Real business data is noisy, and a good analyst explains the noise honestly instead of hiding it.

A data analyst case study for students is nothing more than this: a real dataset, a real business question, and an honest, documented process for answering it. Those are the same three ingredients professional analysts use on client work, years into their career. The skill that survives is knowing how to read a result and say what it actually means, not just what you hoped it would mean.

Before you run your regression, it also helps to check if your variables move together in the first place. A quick Pearson vs Spearman correlation check on your data can save you from building a regression model on variables that were never meaningfully related to begin with.

And if your residuals look patterned rather than random, a Durbin-Watson test will tell you whether autocorrelation is quietly distorting your results. This is something students analysing time-based sales data like James’s run into more often than they expect.

Frequently Asked Questions

Siddharth Gupta is the founder of Statssy and Director and Lead Engineer at IntalliaTech24 Innovate Pvt. Ltd. He holds an MBA in Finance and an M.Tech, and has spent over 20 years in analytics consulting across R, Python, SPSS, Stata, Power BI, and SQL. That includes hands-on dissertation mentoring for marketing and business analytics students. Connect with him on LinkedIn.

If your own dataset is staring back at you the way James’s did, you do not have to figure it out alone. Explore Statssy’s statistical consulting services and let us turn your numbers into a story you can defend with confidence.