Business Analytics

Units

1. 1

   Descriptive analytics and visualization

   4 lessons

   1. Data types and tidy tablesClassify each column of a business data set as categorical, ordinal, numeric or time, arrange the data as a tidy table with one row per record and one column per variable, and handle missing values without distorting a summary.
   2. Center, spread and outliersCompute the mean, median, sample standard deviation and quartiles of a business data set, and flag outliers with the 1.5 IQR rule on a box plot.
   3. Choosing the right chartMatch a business question to the chart that answers it (bars for categories, lines for trends over time, scatter plots for relationships, histograms for distributions), and tell a clear chart from a cluttered or misleading one.
   4. Pivot tables and SQL queriesSummarize a tidy table by group with a pivot table and with SQL SELECT, WHERE, GROUP BY and JOIN queries, and read the result as a smaller table.
2. 2

   Probability and decision trees

   4 lessons

   1. Probability rules and conditioningUse the addition rule P(A or B) = P(A) + P(B) − P(A and B), conditional probability P(A | B) = P(A and B) / P(B) and the test for independence, reading each from a Venn diagram and a table of counts.
   2. Bayes rule and base ratesUpdate a probability with P(A | B) = P(B | A) P(A) / P(B), using a frequency tree of 1000 people to see why a positive result from a good test can still mean a low chance of the condition when the condition is rare.
   3. Random variables and expected valueDefine a discrete random variable by its payoff distribution, compute the expected value E[X] = Σ x P(x) and the variance Var(X) = Σ (x − μ)² P(x), and read E[X] as the balance point and the standard deviation as the risk.
   4. Decision trees and expected valueDraw a decision tree with decision squares, chance circles, probabilities and payoffs, fold it back from the payoffs to the root by averaging at chance nodes and choosing the best branch at decision nodes, and find the probability at which the choice flips.
3. 3

   Sampling and A/B tests

   4 lessons

   1. Sampling distributions and the CLTExplains sampling error, the standard error s / sqrt(n) and the central limit theorem, shown as histograms of sample means narrowing and turning bell shaped as n grows.
   2. Confidence intervalsBuilds an interval estimate, mean plus or minus z times s / sqrt(n), and reads its 95 percent meaning, drawn as many intervals on a number line with a few missing the true mean.
   3. Hypothesis tests and p-valuesTeaches null and alternative hypotheses, test statistics, p-values and Type I and II errors, shown as a bell curve with the rejection tails shaded and the observed statistic marked.
   4. A/B testing in practiceDesigns a randomized website experiment, compares two conversion rates with a z test, and checks sample size and peeking, drawn as a split-traffic flow into variant A and B with a results table.
4. 4

   Regression and forecasting

   4 lessons

   1. Simple linear regressionFits y = b0 + b1 x by least squares, interprets slope and intercept, and R squared, shown as a scatter plot with the fitted line and vertical residual segments.
   2. Multiple regression and dummiesAdds several predictors and dummy variables, reads coefficients holding others fixed, and checks p-values, shown as a regression output table with the key coefficients highlighted.
   3. Checking a regression modelExamines residual plots for nonlinearity and unequal spread, multicollinearity and overfitting, shown as three residual plots, one healthy and two with problems. Fixes include transforming variables and dropping redundant predictors.
   4. Time series forecastingForecasts with moving averages, exponential smoothing F(t+1) = a A(t) + (1 - a) F(t), trend and seasonality, drawn as a line chart of actual demand with forecast lines and error bands.
5. 5

   Linear programming

   4 lessons

   1. Formulating a linear programTurns a product mix problem into decision variables, an objective such as maximize 40x + 30y, and constraints, shown as a word problem beside its algebraic model with each part labeled.
   2. Graphical solution and corner pointsSolves a two variable program by graphing constraints, shading the feasible region and testing corner points, drawn as a shaded polygon with the objective line sliding to the optimal vertex.
   3. Using Solver and sensitivity analysisSolves larger models in a spreadsheet Solver and reads shadow prices and allowable ranges, shown as a spreadsheet layout with changing cells, target cell and a sensitivity report.
   4. Integer and network modelsAdds yes or no decisions and transportation or assignment structure, such as facility location, drawn as a network of supply nodes and demand nodes with arcs labeled by cost and flow.
6. 6

   Simulation and Monte Carlo

   4 lessons

   1. Random inputs and distributionsModels uncertain inputs with uniform, normal and triangular distributions and generates draws with RAND, shown as three distribution curves and a column of simulated draws.
   2. Monte Carlo simulation of profitBuilds a spreadsheet model where uncertain demand and cost drive profit, repeats it thousands of times, drawn as a model flow diagram feeding a histogram of simulated profit.
   3. Reading risk from simulation outputReads mean, percentiles and the probability of a loss from simulated results, shown as a cumulative distribution curve with the 5th percentile and the zero profit line marked.
   4. Comparing decisions under uncertaintyRuns the same simulation for two alternatives and compares expected value against risk, drawn as two overlaid profit histograms with their means and downside tails marked.
7. 7

   Revenue management and inventory

   4 lessons

   1. The newsvendor modelChooses an order quantity with critical ratio Cu / (Cu + Co) against a demand distribution, shown as a normal demand curve with the optimal quantity cutting off the critical fractile area.
   2. Overbooking and capacity controlSets how many reservations to accept when customers may not show, balancing bumping cost against empty seats, drawn as a binomial show-up distribution with the optimal booking limit marked.
   3. Price optimization and demand curvesFits a linear demand curve q = a - b p, finds the revenue maximizing price, and shows it as a downward demand line beside a revenue parabola peaking at the best price.
   4. Fare classes and Littlewood's ruleProtects seats for high fare customers using P(D2 > y) = r1 / r2 style reasoning, drawn as a seat bar split into protected and discount sections with the nested booking limits.
8. 8

   Intro to predictive classification

   4 lessons

   1. Logistic regressionPredicts a yes or no outcome like churn with p = 1 / (1 + e to the minus z), shown as an S shaped probability curve over a predictor with a 0.5 cutoff line.
   2. Confusion matrices and metricsCounts true and false positives and negatives and computes accuracy, precision and recall, shown as a two by two confusion matrix with each cell labeled and formulas beside it.
   3. Training, testing and overfittingSplits data into training and test sets and shows why a flexible model overfits, drawn as a curve of training and test error with the best complexity at the test minimum.
   4. Decision tree classifiersBuilds a classification tree that splits on the best predictor by impurity and reads rules to target customers, drawn as a tree of yes or no splits ending in labeled leaves.