Floemdata
Data & ETL

What is data science vs data analytics?

Data science versus data analytics, the four levels of analysis, and what has to be true about your data before any model is worth building on top of it.

Floemdata · · 5 min read

Data science is the field that extracts predictions and patterns from data using statistics and machine learning. The difference from data analytics lies in the question: analytics asks “what happened and why”, data science asks “what will happen next”.

Four levels, one order

  • Descriptive — what happened? (reporting)
  • Diagnostic — why did it happen? (data analytics)
  • Predictive — what will happen? (data science)
  • Prescriptive — what should I do? (decision support)

This is not a ladder you can skip steps on. Building a predictive model in a company whose descriptive layer is not settled is like roofing a building with no foundation: the model runs and produces numbers, but nobody can show that those numbers are right.

Why most data science projects stall on data, not models

Field experience is consistent: most of the time in a data science project goes into collecting, cleaning and reconciling data rather than choosing a model. With ERP data this is particularly laborious — cancelled documents, backdated corrections, multiple currencies, the same customer under two cards.

Questions to answer before modelling

  • What state was the training data in on a given date — can you reproduce history?
  • Are cancellations and returns removed, or are they inflating revenue?
  • If cost changed retroactively, does the old period still reconcile?
  • Does the number you produce agree with the accounting figure?

A realistic order

For most small and mid-sized companies the right sequence is: consolidate the data and reconcile it against the source; then use diagnostic analysis to see causes; then catch deviations with statistical anomaly detection; and only once those hold, move to predictive models. Floemdata builds the first two steps — what you put on top is your choice.

← All guides