Business Analytics with R and Python: A Practical Textbook on Data-Driven Decision Making for Commerce Majors

Learn the analytics that commerce actually pays for.

Most analytics courses teach methods and hope a decision appears at the end. This textbook runs the other way. Every chapter opens with a choice a manager has to make, identifies the one quantity that would settle it, builds the model that estimates it, and converts the estimate back into money and a recommendation.

Both languages, side by side. Every substantial technique appears in R and Python, so you can join any employer's stack. Where the two disagree - factor handling, one-hot encoding, degrees of freedom in a variance, the regularization scikit-learn applies by default and glm does not - the difference is named outright, because silent default mismatches are the most common reason two analysts reach two answers from one file.

One company, followed all the way through. Aurora Commerce is an omnichannel retailer with a subscription program and three distribution centers. Its dataset carries you from data cleaning to a funded retention program, supplying the churn, supply-chain and A/B testing cases throughout.

What you will be able to do

• Frame a decision before touching data, and compute what information is worth before commissioning a study.
• Clean and join without the errors that quietly destroy results - the grain mistake that inflated a subscription total 17.7-fold, and the missing values that overstate satisfaction because unhappy customers do not answer surveys.
• Design experiments that can answer the question: randomization units, minimum detectable effect, guardrails, and the reason checking results twenty times turns a 5 percent error rate into 25 percent.
• Build models that hold up: regression diagnostics and robust standard errors, logistic regression and calibration, trees and boosting, honest cross-validation - and a demonstration of the leakage that lifts a score from 0.681 to 0.953 while making the model worthless.
• Price the decision, not just the model. Moving a churn threshold from the software default to the economically correct value raises expected campaign profit from $10,850 to $41,767 on the same model and the same customers.
• Handle customers, operations and money: survival analysis and lifetime value, segmentation, demand forecasting with honest backtesting, bottleneck analysis and the newsvendor model, and linear programming with shadow prices you can take to a budget meeting.
• Communicate so people act: the encoding hierarchy, chart redesign, dashboards, and the one-page memo that wins funding.

Built for teaching and for self-study

• 22 chapters organized in five parts, from decision framing to a complete worked program.
• 48 original figures - decision trees, Lorenz curves, power and ROC curves, calibration plots, survival curves, queueing curves, feasible regions and more.
• 46 comparison tables and 77 derived equations, with assumptions stated.
• 88 questions with fully worked answers that show the reasoning, not just the result.
• A laboratory in every chapter, plus a companion archive with the dataset generator and setup scripts.
• Every code listing runs. All of them were executed against the accompanying data before printing.

Who it is for

Commerce, business, marketing, finance and operations undergraduates from their second year onward; MBA and master's students needing a first rigorous analytics course; and working analysts who want the reasoning behind the tools. Introductory statistics is assumed. No prior programming experience is required.

By the last page you will have followed one decision from a blank page to a funded program, and you will know how to name the assumption that would prove you wrong.