Customer Analytics in Python

Learn segmentation, targeting, and positioning in customer analytics, using demographic, geographic, and psychographic data to tailor offerings. Apply these steps within the fmcg context and marketing mix.

Explore how the marketing mix: product, price, promotion, and place, drives positioning and purchase decisions, detailing product features, branding, packaging, pricing strategies, promotions, and distribution.

Compare physical and online retailers through the four questions—who the customers are, purchase probability by category, brand choice, and units bought—highlighting data, locations, returns, and ratings as key factors.

Learn how price elasticity measures how purchases respond to price changes, including own and cross elasticity. Apply this to purchase probability, brand choice, and quantity using Python.

Install Anaconda, Python 3, and Jupyter Notebook to set up the coding environment for customer analytics in Python, and learn how to install the relevant packages.

This course explains why Python and Jupyter are widely used in data science: open source, cross-platform, general purpose, with easy syntax and rich packages for interactive notebooks.

Installing Anaconda guides you to install Python 3 and the Jupyter Notebook with data science packages on Windows, Mac, or Linux, and to launch the Jupyter dashboard.

Take a tour of the Jupiter dashboard, learn file and directory management with checkboxes, upload notebooks, and create or rename notebooks, folders, and text files in a Python notebook workflow.

Explore how to code in Jupyter notebooks using cells, input and output fields, and keyboard shortcuts, and learn to mix markdown and code for efficient, piecewise execution.

Install the relevant Python libraries for this course, including numpy, scipy, pandas, scikit-learn, TensorFlow, seaborn, and pickle, via the Anaconda prompt.

Explore the segmentation dataset of 2000 individuals, pre-processed with numeric encoding and missing-value handling, using Excel for initial eyeballing before moving to Python and reviewing the data dictionary.

Import and load segmentation data with numpy, pandas, and scipy; inspect with head and describe, then explore the Pearson correlation and visualize via a heatmap and age against income scatter.

Standardize segmentation data in Python using scikit-learn's standard scaler to equalize feature ranges, enabling accurate age and income-based customer segmentation models.

Learn hierarchical clustering for customer segmentation, compare divisive and agglomerative approaches, and apply distance measures like euclidean, squared euclidean, and manhattan with the ward method to form distinct groups.

Explore hierarchical clustering in Python using SciPy, including dendrogram visualization and Ward linkage, to identify four clusters and decide cut points for segmentation.

Explore the basics of k-means clustering: choose k, initialize seeds, assign points by squared Euclidean distance, update centroids, and iterate to convergence while noting outliers, k-median clustering, and spherical shapes.

Apply k-means clustering with scikit-learn to partition data, evaluate cluster quality using within-cluster sum of squares, and determine the optimal number of clusters via the elbow method.

Explore K means clustering results by attaching segment labels, computing mean features per cluster, and naming segments to reveal distinct customer profiles for analytics in Python.

Apply dimensionality reduction to simplify high-dimensional data using principal component analysis, transforming three or more variables into two or more principal components, and prepare for improved clustering with K-means.

Learn to perform principal component analysis with sklearn for customer analytics in Python, import PCA, reduce seven features, interpret explained variance ratios, and select components by cumulative variance.

Learn how PCA reduces seven features to three components, interprets loadings, and uses a seaborn heatmap to visualize component relationships and generate PCA scores for k-means clustering.

Segment data with k-means using principal component scores as features, identify four clusters via elbow method, and visualize within-cluster sum of squares to validate the segmentation.

Analyze four-cluster results from K-means PCA by building a data frame with PCA component scores, labeling clusters, computing mean features by cluster, and visualizing separation on a two-component scatter plot.

Export your segmentation models using pickle, save the scaler, PCA, and k-means with pickle files, and prepare for future preprocessing and purchase analytics.

Explore purchase analytics using linear and logistic regression to answer category purchase, brand choice, and quantity, as part of positioning in customer analytics.

Explore the purchase dataset through purchase analytics, using 500 individuals' daily chocolate candy bar purchases over two years to test segmentation models and promotions.

Import numpy and pandas, load the purchase data CSV, inspect with head, confirm no missing values thanks to preprocessing, and apply a segmentation model to group transactions by segments.

Apply the segmentation model by standardizing seven features, reducing to three principal components with PCA, and assigning new customers to one of four clusters using a K-means model.

Learn descriptives in customer analytics in python by grouping purchases by individuals and segments, calculating visits, purchases, and average purchases, and visualizing segment proportions with a pie chart.

analyze how purchase occasions and purchase incidences vary by customer segments using mean and standard deviation. visualize results with bar charts to compare segments and highlight homogeneous versus heterogeneous groups.

Analyze brand choice across segments by constructing brand dummies, computing average brand selection per customer, and visualizing with a heatmap to reveal price-related preferences.

Explore how to compute segment-wise revenues across multiple brands, identify top revenue drivers like the career-focused segment, and use grouped data to compare brand performance.

Apply binomial logistic regression to estimate customer purchase probability at each shopping trip, interpret probability outputs, assess price elasticity under different conditions, and classify purchases with a 0.5 threshold.

Prepare the dataset for logistic regression in a new Jupyter notebook, load the scaler, PCA, and k-means with PCA, and create the Dfpa data frame for predictive purchase analytics.

Calculate price elasticity of purchase probability using a logistic regression coefficient, across a price range from 0.5 to 3.5 with 0.01 steps, and create a price elasticities dataframe.

Analyze the price elasticity of purchase probability results using a plotted elasticity curve to show negative elasticities and a switch from inelastic to elastic around 1.25, guiding pricing decisions.

model purchase probability by segment using logistic regression on segment-specific data, compare price elasticities for career focused and fewer opportunities, identify tipping points and inelastic ranges.

Fit a logistic regression model using price and promotion to estimate purchase probability. Discover how price lowers purchase probability while promotion raises it, with elasticity to be explored next.

Explore price elasticity and promotion elasticity using logistic regression, creating a price elasticity promotion data frame with promotion set to one, and compare elasticities for scenarios with and without promotions.

Compare price elasticities with and without promotion using curves that show promotion raises purchase probability elasticity from 1.27 to 1.46 across the price range, suggesting higher prices with promotions.