Google Ads to BigQuery, without building a pipeline

Written by Uptimal, the team that builds and operates Upbuild and runs marketing analytics for multinational clients across the globe. Everything below reflects how Upbuild actually behaves in production.

Getting Google Ads into BigQuery has three usual answers: the Data Transfer Service, a paid connector priced per source, or a script somebody wrote and now maintains. All three work. All three are a project, and each one adds a place where the schema can drift from what your dashboards expect.

AI Data Exports, part of the Upbuild platform, offers a fourth: describe the export in the same chat where you ask the questions, and it runs on a schedule into a table you named, in a project you own.

How the export is defined

• You pick the destination. A project, a dataset and a table name, in your own Google Cloud. Upbuild never creates a dataset and never touches a table you did not name.
• You pick the cadence. Daily, weekly, or the days you choose, in your organisation’s time zone rather than in UTC.
• You pick the history. Ask for past months and the export backfills them.
• You pick the shape. The same fields you would ask for in a question: spend, clicks, conversions, campaign, ad group, date, whatever the report needs.

Two details that decide whether the table is trustworthy

Many accounts, one table. An agency running many Google Ads accounts does not want a table per account. One export can cover many accounts and land them together, with the account id and name on every row, so a dashboard can group by client without a union of many sources.

A restated day updates in place. Google Ads keeps revising recent days as conversions land late. Rows can be merged by key columns, typically date plus campaign plus account, so yesterday’s restatement updates the existing row instead of adding a second one. A naive append pipeline is exactly how a warehouse ends up double counting a week.

What this is not

It is not a full ELT platform, and it does not want to be. There is no transformation layer and no modelling tool on top. It moves the report you asked for into the table you chose on the clock you set. If your team already runs a warehouse practice, this is the cheap path for the ad data that never justified its own pipeline; if you have no data engineer at all, it is the whole answer.

Reading it back

Once the data lands, the same assistant can query it: BigQuery is one of the platforms the MCP server reads, so “compare last quarter in the warehouse to what Google Ads reports today” is one question, not an export and a spreadsheet. That loop, ask, schedule, ask again, is the point of doing it here rather than in a pipeline tool.

Frequently asked questions