From Business Analyst
to Data Analyst.

OLTP (e.g., Salesforce) databases handle live operations — each write is a fixed type at a fixed position. OLAP (analytical) databases store historical data optimized for aggregated queries. Always ask which one you're querying before you optimize anything; the two behave very differently.

Structured data lives in rows and columns — spreadsheets, SQL tables, CSV files. Every field has a fixed type and position. The easiest format to search and work with, and the one analysts do most of their analysis work in.

Semi-structured: flexible structure with key-value pairs. JSON (APIs, event logs), Parquet (the data-warehouse pivot format). Has fields and values, but the structure can vary between records — you cannot treat it like a CSV until you parse it.

Raw: free text, logs, images, PDFs. No fixed schema. Requires significant pre-processing — usually NLP or ML — before it becomes usable for analysis.

SELECT, FROM, WHERE, JOIN, GROUP BY, row aggregates — the core of analytics work. You cannot do this without it. Know it well and don't rely on someone else for every query.

Window functions calculate rankings, running totals, and moving averages without collapsing your data. CTEs (WITH clause) break complex queries into readable, sequential steps. Together they replace most subquery spaghetti.

Python specifically for loading and reshaping data. Not for ML pipelines. Pandas puts you one read_csv() away from a full DataFrame so you can quickly inspect it, then pass you to a much stronger position alongside familiar BI tooling.

Mean, median, distribution, correlation. SQL tells you what the data says; statistics helps you understand whether that answer is meaningful. Understanding variation prevents you from drawing too much insight from noise.

Knowing how to build a dashboard and tell the underlying data story. The goal is decision-making findings, not just charts. Pick one tool and go deep — breadth here matters less than depth.

A cloud data warehouse purpose-built for analytics. Separates storage and compute so teams can scale independently. The most common place BAs find themselves querying production data in enterprise environments.

A Lakehouse platform built on Apache Spark, purpose-built for data engineering and ML pipelines. Increasingly the engine behind enterprise data transformations — if your company has a large data science team, you will encounter it.

Google's serverless analytics warehouse. Pay-per-query pricing and native integration with GCP services. Often the entry point for teams already invested in Google Workspace or using Google Analytics data.

Normalization removes data redundancy by splitting tables; denormalization trades storage for query speed. Knowing when to apply which is a core design decision that determines how easy — or painful — your data is to query downstream.

A technique for organizing data warehouses into fact tables (measurements) and dimension tables (context). It is the standard mental model for analytical databases — understanding this unlocks how real enterprise data is structured and why.

The simplest dimensional model: one fact table in the centre, dimension tables radiating outward. Easy to query, easy to explain to stakeholders, and what most BI tools are optimized to consume.

An extension of the star schema where dimension tables are further normalized. More storage-efficient but more complex to query. Knowing when to choose one over the other separates junior data analysts from senior ones.

The de-facto tool for transforming raw data into analytics-ready models, in SQL. Version-controlled data transformations with built-in testing and documentation. If your team isn't using dbt, someone is spending time solving a problem it has already solved.

Microsoft's cloud ETL service for ingesting and moving data at scale. The standard connector between source systems and the warehouse in most enterprise Microsoft stacks. Knowing it means you can trace where data comes from without asking an engineer every time.

An orchestration platform that schedules and monitors data pipelines as code. When pipelines grow complex enough that "run this after that finishes" needs to be reliable and auditable, Airflow is the standard choice.

Master Data Management — the practice of creating a single, authoritative record for key business entities (customers, products, accounts). Without MDM, "customer count" returns a different number in every system, and every report becomes a negotiation.

Defining who can see what, how long sensitive data is retained, and how it is classified. GDPR, CCPA, and internal compliance policies all fall here. Non-negotiable in any regulated industry — and increasingly standard everywhere else.

The systematic process for measuring, monitoring, and improving data quality — completeness, accuracy, consistency, timeliness. Bad data is the most common reason analytics projects lose stakeholder trust long before the insight is ever acted on.