CardioGoodFitness — Customer Segmentation

Descriptive analytics that tells a treadmill company exactly who buys each of its three machines.

Analytics that moves decisions
Small-sample discipline
Findings in the audience language

Records: 180 purchases
Products: 3 treadmill tiers
Output: Segment profiles + actions
Method: Uni/bi/multivariate EDA

Overview

A classic retail-analytics brief executed cleanly: CardioGoodFitness sells three treadmill models at three price points, and the company wants to know who buys each one. From 180 customer purchase records — age, gender, education, marital status, income, expected usage, self-rated fitness — this analysis builds the customer profile for each product line and turns it into concrete targeting recommendations.

Problem

The three treadmills are not competitors; they are answers to different customers. But the sales data does not say that by itself. The task is descriptive statistics done properly: segment the buyers, characterise each segment, and say something a marketing team can act on — without overclaiming from 180 rows.

Goals

Build a demographic and behavioral profile for each treadmill model
Identify which variables actually separate the segments — income, usage expectations, fitness self-rating
Deliver recommendations phrased for a business audience, not a statistics one

Solution

The analysis works systematically through univariate views (who the customers are overall), bivariate views (how each attribute splits across the three products), and cross-tabulations of the interactions that matter, such as income against product tier and planned usage against fitness rating.

The picture that emerges is crisp: the entry model serves casual, budget-conscious buyers; the mid model looks demographically similar but skews to committed intenders; and the premium model belongs to a distinctly high-income, high-usage, self-rated-fit segment — closer to a different market than a different tier.

Findings land as recommendations: position the premium machine on performance to its niche, market the entry and mid tiers on accessibility, and use expected-usage questions at point of sale as a natural upsell signal.

Tooling

Tech stack

Python
pandas
seaborn
matplotlib
Jupyter

Friction

Challenges

Small sample discipline

With 180 records, subgroup cells get thin fast. The analysis stays at the resolution the data supports — product-level profiles — and resists slicing into unstable micro-segments.

Self-reported variables

Fitness ratings and usage plans are aspirations, not measurements. They still separate the segments well, but the write-up is explicit that they describe buyer psychology as much as buyer behavior.

Takeaways

Lessons learned

Descriptive analytics is underrated — a clean cross-tab that changes a marketing decision beats an unnecessary model
The premium-segment finding is the actionable one; most of the value of segmentation is finding the group that behaves differently
Writing conclusions in the audience's language is part of the analysis, not a garnish

Evidence

Dataset

CardioGoodFitness

Kaggle (retail case-study dataset)

180 treadmill purchase records across models TM195, TM498, and TM798 with demographics, income, expected usage, and self-rated fitness.

View dataset

Roadmap

Future improvements

Cluster analysis to test whether the data suggests segments beyond the product split
Price-sensitivity framing if transaction amounts become available
A one-page interactive summary for the browser

Continue

Data Science

Pima Indians Diabetes — Clinical Risk Modeling

Analytics

Automobile Data Visualization