Two thousand years of data, one evening your audience won’t forget.

The Roman Census

… and what we can learn from it today

In the sixth century BC a Roman king, said to have been born a slave, ordered every Roman citizen onto the Field of Mars. Name, father, age, wife, children and every possession, declared in public — and anyone who stayed away could lose their property and be sold into slavery. With that, Rome built the first database that ran a state: one register that decided who paid tax, who fought in the front line and whose vote counted.

In this talk I treat the Roman census as what it really was: a data system. We follow it from its origin to the censors who kept it, from the burning of the archive on the Capitol in 83 BC to the bronze and wax records that replaced it. And we look east to Han China, which counted 59.6 million people in AD 2 with a household register whose descendant is still in use today.

Every data problem your organisation struggles with today, Rome had first. Some of its solutions were better than ours.

The Impact of Invention

How the inventions of the 1970s still shape the way we handle data today

In 1970 Ted Codd proposed something radical: describe data by what it means, not by where it sits on disk. Within a few years there was a formal query language to go with it, and a clean split between the logical model and the physical storage underneath. Programmers stopped chasing pointers and started talking about data. That is the best idea our profession ever had.

Fifty years later, most data platforms have quietly lost it. The spreadsheet arrived in 1979 and never left. Reporting databases grew into heavy central warehouses. Data lakes gave data scientists freedom without governance. And today’s lakehouses are often so over-architected that the tooling buries the data it was meant to serve.

This talk follows the line from SQL to the OLAP cube, to the pandas dataframe, to the autonomous data warehouse — and shows how much of that line was shaped by accident and by a Darwinian struggle between inventions rather than by design. Knowing which of your choices are inherited accidents is the first step to making better ones.

The best idea in data management is fifty years old. Most platforms have forgotten it.