Question 1

What can a ClickHouse model send to Google Ads?

Accepted Answer

Customer match audiences, where the rows your model returns become the members, and conversion events for conversions that completed away from the click. Audiences are where most teams start. Conversions matter when the outcome you care about happens deep in the product or in a billing system, which is exactly the data a ClickHouse model is usually built from.

Question 2

What happens if the query exceeds a memory limit?

Accepted Answer

It is retried rather than failing outright. ClickHouse raises a distinctive error when a query exceeds its memory allowance, and Polytomic recognises that specific condition and retries the read. That is a genuinely useful behaviour on a wide aggregation over a large table, though it is a safety net rather than a substitute for a sensible model. If you are hitting it regularly, narrowing the selected columns or adding a filter is the better fix.

Question 3

How do I avoid rescanning the table each run?

Accepted Answer

Give the model a tracking column. Polytomic persists the previous position and reads rows past it, bounded by a new high water mark, rather than reading everything. Use a column that genuinely advances whenever a row changes, and note that the sorting key on the table can also be used for deterministic chunking. Without a tracking column the model still syncs differentially, but the query itself still has to scan.

Question 4

What can incremental reads not see?

Accepted Answer

Two things worth designing around. An update that does not move the tracking column will not be picked up, so the column has to change whenever the row does. And source deletes are not captured at all, since a row that disappears simply stops being returned by the query. For an audience that needs to shrink, model the removals explicitly rather than expecting the incremental read to notice them.

Question 5

Why build ad audiences from event data?

Accepted Answer

Because the useful definitions are behavioural. People who hit a feature three times but never converted, or who churned after a specific pattern, cannot be expressed in an ad platform's own targeting. Those segments live in whatever holds your events, which for high volume products is often ClickHouse. Reverse ETL sends the result on a schedule so the audience keeps up with the behaviour rather than freezing on the day it was built.

Sync ClickHouse to Google Ads

Sync ClickHouse to Google Ads with Polytomic

Audiences and conversions

Memory errors retried

Incremental on a tracking column

Or sync Google Ads to ClickHouse

Trusted by businesses worldwide

Step-by-step integration guide

Connect ClickHouse in Polytomic

Connect Google Ads in Polytomic

Create a sync between ClickHouse and Google Ads

Map fields and enable the sync

Frequently asked questions

The sync and ETL platform that does it all

Powerful data sync

Advanced analytics

Enterprise security

Easy setup

Ready to connect ClickHouse with Google Ads?