Statistics and analytics terms in plain language — search
Median
Middle value robust to outliers: half the data above, half below. More honest than the mean for salaries/prices.
Percentile
P90 = the value below which 90% of data falls. 'P95 latency < 200ms' — 95% of requests are faster than 200ms.
Standard deviation
Typical spread around the mean. Small σ — dense data; large — scattered. ~68% within ±1σ (normal distribution).
Correlation
Joint movement of two variables: r from -1 to +1. Correlation ≠ causation: ice cream and drowning correlate via summer, not sugar.
Regression
Model 'y = a + bx': predicts a dependent variable from an independent one. Simple but powerful — forecasting base.
p-value
Probability of seeing this result with no real effect. <0.05 — statistically significant (but not necessarily practically important).
Sample vs population
Sample — the subset we measure; population — everyone. Sample findings generalize with a margin of error ±(depends on n).
Segmentation (cohort)
Cohort — a group united by an event-time (signed up in March). Cohort analysis is more honest than averages.
Confidence interval
Range where the true value lies with 95% probability. 42% ±3% — reality is somewhere in 39–45%.
ETL/ELT
Extract-Transform-Load: pull data from sources, clean, load into warehouse. ELT — load raw first, clean inside the warehouse.