Analytics
494 bites tagged Analytics — interview questions with model answers, and 60-second explainers.
Holt-Winters Exponential Smoothing
Forecast time-series data by giving more weight to recent events. Holt-Winters smoothing is used to analyze metrics and make predictions by accounting for user-defined assumptions like seasonality.
Exponential Smoothing: Weighting Recent Data More Heavily
Exponential smoothing creates forecasts from a weighted average of past data, where weights decay exponentially for older points. It's used for quick, reliable predictions like inventory demand. The footgun: simple versions ignore trends and seasonality.
Stationarity: A Time Series's Stable Personality
A time series is stationary if its statistical personality, like its average and spread, remains constant over time. Many forecasting models require this stability to work correctly.
Moving Average: Smoothing Out the Noise in Data
A moving average smooths out noisy data by replacing each point with the average of its neighbors, revealing the underlying trend. It's used to track stock prices or server metrics without overreacting to short spikes.
Propensity Score Matching: Mimicking an A/B Test
Propensity Score Matching (PSM) mimics a randomized trial with observational data by finding a "statistical twin" for each subject. It's used to estimate a feature's impact when a true A/B test isn't possible. The footgun is assuming it removes all bias.
Instrumental Variables: Isolating True Cause and Effect
Instrumental Variables (IV) isolate true causality when a key variable is tainted by hidden factors. An IV is a "clean" substitute that influences your cause without directly touching your effect, letting you measure the true relationship, free from…
Regression Discontinuity Design (RDD)
RDD finds a natural experiment at a cutoff point, like a test score threshold for a scholarship. By comparing people just above and below the score, you can estimate the program's effect. The footgun is assuming this local effect applies to everyone.
Isolating Impact with Difference-in-Differences (DiD)
Difference-in-Differences (DiD) isolates an intervention's true effect by comparing a treatment group's change over time to a control group's. This reveals if a new feature truly boosted engagement, not just rode a general upward trend.
Sample Ratio Mismatch (SRM): When Your A/B Test Is Broken
Sample Ratio Mismatch (SRM) means your A/B test's traffic split is broken, violating random assignment. For example, a 50/50 split results in a statistically significant imbalance.
Average Treatment Effect (ATE): Isolating the Impact of a Change
The Average Treatment Effect (ATE) isolates an intervention's true impact by comparing the average outcome of a treated group to a control group. It's used in A/B tests and policy evaluations. The footgun is assuming causation without true randomization.
The Counterfactual Framework for Causal Inference
The Counterfactual Framework models causality by imagining two parallel universes for each person: one with a treatment, one without. It's the basis for A/B tests and analyzing observational data.
Twyman's Law: Interesting Data is Usually Wrong
Twyman's Law states that any data point that looks interesting is probably wrong. Before celebrating a sudden 10x spike in user engagement, first suspect a bug in your analytics pipeline or a bot attack.
Selection Bias: When Your Sample Skews Your Results
Selection bias occurs when your data sample isn't random, leading to flawed conclusions. This happens when surveying only volunteers or analyzing a non-representative group. The footgun is assuming your data reflects the whole population when it doesn't.
Bootstrapping: Quantifying Uncertainty with Resampling
Bootstrapping estimates uncertainty by resampling your own data. It's used to find confidence intervals for complex stats like medians where no simple formula exists. The footgun: it can't fix a biased sample, only reveal the uncertainty within it.
Statistical Power: Detecting Real Effects in Your Tests
Think of statistical power as your experiment's sensitivity. It's the probability of detecting a real effect, like a true lift in an A/B test. The main footgun is running a low-power test, which will likely miss a real improvement and lead you to discard good.
ANOVA: Comparing Group Averages by Analyzing Spread
ANOVA checks if group averages are different by comparing the spread *between* groups to the spread *within* them. It's used to see if three ad campaigns yield different click-through rates.
Standard Error: Gauging Your Measurement's Precision
Standard error measures the precision of a sample statistic, like the mean. It answers: "If I ran this experiment again, how much would my result change?" It's key for building confidence intervals and A/B testing. Don't confuse it with standard deviation.
Type I vs. Type II Errors: False Alarms vs. Missed Detections
A Type I error is a false alarm (a smoke alarm with no fire), while a Type II error is a missed detection (a fire with no alarm). This trade-off is crucial in A/B testing and medical diagnostics.
A/B Testing: Making Decisions with Data, Not Guesses
A/B testing is a controlled experiment pitting two versions of a product against each other with real users. It's used to see if a change, like a new button color, improves a metric like clicks.
Law of Large Numbers: More Data, Truer Average
The more you repeat an experiment, the closer your average result gets to the true, underlying average. This is why A/B tests need sufficient traffic and casinos can reliably predict earnings. The footgun is mistaking it for the 'law of averages' fallacy.
Sampling: Estimating the Whole from a Small Part
Sampling lets you understand a large group by studying a small, representative piece. Think of it like tasting a spoonful of soup to season the whole pot. It's used in A/B tests and polls, but the main footgun is a biased sample.
Probability Distribution: Mapping What's Possible
A probability distribution maps every possible outcome of a random event to its likelihood. It's used in A/B testing to model conversions or in monitoring to predict server load. The footgun is assuming every distribution is a bell curve; many are not.
Time Series Decomposition: Separating Signal from Noise
Time series decomposition breaks a metric into its core parts: long-term trend, repeating seasonal patterns, and random noise. This helps you understand *why* a metric changed—was it a real shift or just the usual holiday rush?
Simpson's Paradox: When Averages Mislead
Simpson's Paradox is when a trend seen in separate groups reverses when you combine them. This happens when a hidden variable, like user experience level, skews the results, making a bad feature look good overall. Always segment your data to avoid this trap.
Get Analytics bites daily.
Five a day, five minutes, offline. With quizzes so it sticks.
Open testing — you’ll join as an early tester.