Skip to content
NEW: Martech, explained →

Martech, explained · Data foundations · 1:44

Data clean rooms, explained

Two companies want to know how many customers they share without handing over a list: how a clean room matches on hashed keys and lets only totals out

Embed this film

Paste this where the film should play. It links back here.

The short answer

A data clean room is a controlled space where two companies bring customer data but neither sees the other's rows. Identifiers go in as hashed keys, the keys are compared inside, and only totals come out: how many customers overlap and how they split. Each side sets the rules, small groups are held back, and consent still applies.

Questions this answers

What is a data clean room?

A controlled space where each side brings data but can't see the other's rows. Two companies learn what their customers have in common without either one handing over a customer list.

How do two companies match customers without sharing emails?

Each side converts its identifiers into hashed keys before anything goes in. An email becomes a long string, and the same email always gives the same string, so the two sides can match on it without passing the address.

What comes out of a data clean room?

Totals: how many customers overlap, and how they split by region or age. Enough to plan a joint campaign or to see whether an ad led to a sale. Nobody walks out with the names of who matched.

Who sets the rules in a clean room?

Each side does: which fields can be used, which questions can be asked, and what may leave. A group that's too small is held back, because a count of three can point at real people.

Does a clean room replace consent?

No. Each side still needs the right to use what it brings in. Both sides must also prepare their keys the same way, or real matches are missed.

Transcript

Every chapter as text. Click a time to play the film from there.

0:00 Two companies, one question

A clothing brand and a news site think they share a lot of customers.

Both want to know how many. But a customer list isn't something you hand over.

A data clean room lets them find out, without either side giving its customers away.

It's a controlled space where each side brings data, but can't see the other's rows.

0:23 Hashed keys

Before anything goes in, each side converts its identifiers into hashed keys.

An email becomes a long string, and the same email always gives the same string.

So the two sides can match on it, without passing the address.

Inside, the keys are compared. Where both have the same one, that's a shared customer.

But nobody walks out with the names of who matched.

0:49 Only totals come out

What comes out are totals: how many customers overlap, and how they split by region or age.

Enough to plan a joint campaign, or to see whether an ad led to a sale.

1:02 Each side sets the rules

Each side sets rules: which fields can be used, which questions asked, and what may leave.

And a group that's too small is held back. A count of three can point at real people.

It doesn't replace consent. Each side still needs the right to use what it brings in.

Both sides must also prepare their keys the same way, or real matches are missed.

1:27 The answer leaves the room

Back to the brand and the news site. They know how many customers they share.

The answer left the room. The customers never did.

Learn more at Martech Stack Builder, or build your own.