Data Foundations

Data Foundations is the largest category on the site, thirteen of the forty planned pieces, because an agent reasons over the data it is given. If the conversion definitions differ by platform, if the metric has three owners, or if a feed broke on Tuesday and nobody noticed, the agent acts on that with the same confidence as on good data.

The planned pieces run from the ground up. The first will ask whether the data model has to come before any agent, and what a team that skips it gets wrong. Two will ask why an agent cannot compare Google and Meta as the platforms report them, and what it takes to give each metric one governed definition. Others will take up the stack and what to own or rent, whether agents should read ad platform data through pipelines or live connectors, what a knowledge graph adds on top of a warehouse, and a scoring method for how ready each brand's data is for AI.

The rest is the unglamorous work the category will cover piece by piece: whether identity resolution has to be built in stages, server-side tracking and what it feeds the platforms' own models, tagging creative so performance can be explained, monitoring data quality when an agent rather than a person acts on the numbers, consent rules enforced in the data layer, and the system users, scopes, and tokens that decide whether agent access to an ad account is safe.

The reading order below follows how a team would build the foundation in practice, starting with the data model.