Data Analysis and Digital Business
Data literacy is now a management competence. Know descriptive versus inferential statistics and when each applies. The mean is sensitive to outliers; the median and mode are not. Correlation does not establish causation - look for confounders, reverse causality and selection bias. On sampling, understand random versus convenience samples, survivorship bias, and Simpson's paradox, where a trend in subgroups reverses when data are aggregated. In hypothesis testing, a p-value is the probability of data this extreme if the null hypothesis were true - not the probability that the null is true. Statistical significance is not business significance. In experimentation, A/B tests need adequate sample size and power, a pre-registered primary metric, and awareness of peeking.
Digital business: funnel metrics (traffic, conversion, average order value, retention), cohort analysis, RFM segmentation, unit economics, and the gap between vanity metrics and actionable KPIs. Data governance covers quality, ownership, privacy such as GDPR, security and access control. AI and automation follow garbage in, garbage out: models need monitoring for drift and bias.
Traps: quoting an average for a skewed distribution; declaring a 0.3 percentage point lift a win without significance; ignoring seasonality; treating a dashboard as a strategy.
Worked example: 1,000 visitors convert at 10% in version A and 12% in version B. At that sample size the two-point gap is near the edge of significance, so extend the test rather than shipping immediately.
Rule of thumb: always ask compared with what, over what period, and who is missing from the data?