Real systems and common myths

The same split, every industry

Industry	OLTP: running it	OLAP: understanding it
Banking	Transfers, deposits, card payments, limit checks	Spending patterns, loan portfolio performance, unusual activity
E-commerce	Cart, checkout, discounts, payment status, refunds	Monthly revenue, top sellers, campaign comparison, repeat purchases
Healthcare	Patient registration, appointments, prescriptions, billing	Appointment use, waiting times, outcomes across groups
Education	Enrolment, attendance, marks, fee payments	Completion rates, results across semesters, retention
Social media	Posts, likes, comments, follows, messages	Daily active users, engagement, retention by signup cohort

Banks often add a third piece: fraud checks that score a card payment in real time, on streaming data and low-latency analytical stores. In healthcare, the analytics side also needs strict privacy controls and access rules, because the data is about patients. At very large scale, platforms combine transactional databases, event streams, search indexes, caches and analytical storage.

Myths to drop

Myth	Reality
OLTP is always normalized, OLAP always denormalized	Strong tendencies, not rules. Warehouses also use snowflake schemas; apps denormalize hot paths.
OLTP is always row-oriented, OLAP always columnar	Usually true, but hybrid engines exist. The access pattern decides.
OLTP databases can't run analytics	They can. The question is whether it hurts the transactions.
Warehouses can't update data	They load, update and delete all the time. OLAP describes the main workload, not a ban on writes.
A data warehouse is OLAP	OLAP is a kind of processing. A warehouse is a place that integrates data for it.
A data lake is OLAP	A lake stores files in any format. You still need an engine to analyse them.
OLAP and BI are the same	BI is the wider practice: reports, dashboards, metrics, decisions. OLAP is one part of it.

Real systems and common myths

The same split, every industry

Myths to drop

Revision

Further reading