Data Analytics

Short answer: Data analytics covers the work of turning raw data into answers. A widely used framework divides it into four types. Descriptive analytics asks what happened. Diagnostic analytics asks why it happened. Predictive analytics asks what is likely to happen. Prescriptive analytics asks what action to take. Real projects often mix the types.

People often draw the four types as a ladder of rising difficulty and value. But a company does not need to finish one step before starting the next. Descriptive reporting stays useful at every stage. People describe the link between data analytics and business intelligence in different ways. Many treat business intelligence as regular reporting on past and current performance. They see analytics as the wider field that adds statistics, forecasting and modeling. Others use the two terms as if they mean the same thing. Data analytics also differs from data science. Data science usually means more technical work, such as building models and putting them into use, although these lines also vary.

Analysis usually starts with a specific business question. The analyst gathers the relevant data, checks and cleans it, and explores it with summaries and charts. Depending on the question, the analysis might compare groups, calculate a trend, run a statistical test or build a model. The analyst then checks whether the result holds up, for example on a different period or a part of the data. They explain the result in business terms and share it along with its limits. The final step is a decision or an action. Ideally, someone checks later whether it worked.

A subscription business with 10,000 customers sees monthly cancellations rise from 400 to 600. That is the descriptive finding. Diagnostic analysis then compares the two months. In both months, about 1 in 6 customers with a late delivery cancel. Only about 1 in 30 customers with on-time deliveries cancel. What changed is the number of customers with late deliveries, which rose from 500 to 2,000. Last month, 500 late customers at 1 in 6 give about 83 cancellations, and 9,500 on-time customers at 1 in 30 give about 317, or 400 in total. This month, 2,000 late customers give about 333 and 8,000 on-time customers give about 267, or 600 in total. So the rise in late deliveries explains the whole increase in cancellations. A predictive model could then flag customers at risk.

Most growing companies already have more data than they use. Analytics helps leaders move from knowing that a figure changed to understanding why and what to do next. It helps with decisions on pricing, staffing, product range and keeping customers. In those areas, gut feel alone can miss patterns across large numbers of transactions.

Analysis can find patterns that are pure coincidence. A link between two figures does not prove that one causes the other. In the example above, late deliveries might cluster in one region that also has a stronger competitor. Results also depend on the analyst's choices, such as which period, filters and comparisons to use. Analysts and managers can look, without meaning to, for evidence that supports what they already believe. Small samples produce unstable results.