A data clean room is a secure environment where two or more companies match and study their data without showing each other the raw records. A grocer and a brand can learn which ad drove which sale. Neither sees the other’s customer list. For a loyalty program, the room is how member data earns revenue from brand partners.
What is a data clean room?
The IAB defines a data clean room as a secure place where two or more parties work with shared data. The uses are agreed in advance. Strict limits on access mean no party sees the other’s customer data (IAB). The IAB Tech Lab’s guidance, released as a draft in February 2023, lists four building blocks (IAB Tech Lab):
- Data stays apart. No one can read the raw data unless every party agrees.
- Privacy tools. These include encryption and added noise in results.
- Privacy controls. These cap the number of queries, their type and their time window.
- Access controls. They set who can run which action, for what purpose and how long.
Three roles appear in most setups. A contributor supplies data. A provider supplies the app. A user runs queries and takes the outputs (IAB and Orrick, 2025).
How does a data clean room work?
Collaboration runs in seven steps. The parties connect the data, then transform, stage and prepare it. Next comes an interface, then the computation, then the output (IAB and Orrick, 2025). The output is often an aggregate, such as an overlap count or a lift report. A clean room can also return a record-level list for planning a campaign.
The IAB Legal Affairs Council sorted setups by two questions. Does the data move into the provider’s environment? Where does the application run? The answers give three models (IAB and Orrick, 2025):
| Model | Where the data sits | Where the application runs |
|---|---|---|
| A | Each contributor keeps its data at home | The provider’s environment |
| B | Copies sit in separate instances at the provider | The provider’s environment |
| C | Each contributor keeps its data at home | Inside one contributor’s environment |
Four uses are common. They are insights, ad measurement, richer audience profiles and campaign plans (IAB and Orrick, 2025). Retail media networks use clean rooms to share shopper data with brands (IAB).
Does a data clean room make data anonymous?
No. An IAB-sponsored legal paper from April 2025 calls it a myth that clean rooms turn personal data into deidentified data. The authors found no clean room that handles only deidentified data (IAB and Orrick, 2025). Most clean rooms turn records into unique identifiers. Those still count as personal information.
EU law lands in the same place. Data that went through pseudonymisation can still be linked back to a person with extra information. It stays information on an identifiable person (GDPR Recital 26). The EDPB agrees (EDPB Guidelines 01/2025).
The FTC says the same about hashing. Data is anonymous only when it can never be tied back to a person. A hash of an email address is not anonymous, and a company should not claim it is (FTC).
Group results get different treatment. California counts group data as aggregate when every identity is removed and nothing links back to a person or home (CCPA § 1798.140(b)). A lift rate can fit. A record-level list cannot.
California’s test for deidentified data asks if the information can be linked to a consumer. The business must also promise publicly not to reidentify it (CCPA § 1798.140(m)). A clean room cuts exposure. It does not take a program outside the privacy laws.
What are the limits of a data clean room?
The IAB Tech Lab lists limits that a buyer should know (IAB Tech Lab):
- The parties still have to trust one another, the provider and the cloud host.
- Common ID. The data sets need a shared identifier. An identity partner can supply one.
- Match rates. They can differ from what another provider reports. The key and the method both matter.
- Cleaning data is harder when neither side sees all of it.
- Attribution. Credit across many touches is limited. So is adding new fields to raw data.
Clean rooms can also be attacked. In a dictionary attack, one party loads every possible phone number or IP address in place of its real data. That could let it profile most of the other side’s records, beyond the overlap (IAB Tech Lab). A buyer should ask which attacks a provider defends against.
The IAB Tech Lab suggests four checks when choosing a provider (IAB Tech Lab):
- Scale. How many data sets, of what size, can it run? Which activation channels does it support?
- Speed. How fast can you set it up and run a query? Does it handle data that keeps changing?
- Privacy. Does it offer tools that protect each person, such as encryption or added noise?
- Defenses. Which attacks does it block? Ask for the list.
How do loyalty programs use data clean rooms?
In IAB research from 2023, 64% of companies using privacy tools were using clean rooms (IAB Tech Lab).
A loyalty program holds what a brand wants. That is purchases tied to a known member. Kroger’s 84.51° unit launched a shared cloud in 2021. It uses anonymized transaction data from 60 million US households (Grocery Dive, November 8, 2021).
For illustration, a grocer with 4,000,000 members and a brand with 1,000,000 customer emails match inside a clean room. They find 250,000 shared people. The brand shows an ad to 100,000 of them and holds back 150,000. In the following month 8% of the exposed group buy, against 6% of the held-back group. The room reports only the two rates and the lift: 2 points, or about 2,000 extra buyers among the 100,000 exposed. No one saw a customer list.
A clean room is one way to use second-party data. That is a partner’s first-party data, shared under agreement, as in travel alliances and retail media (first-party data page).
Share only the fields the task needs. The IAB Tech Lab defines data minimization as restricting access to only the data that is absolutely necessary for a specific task (IAB Tech Lab). A brand lift test needs a member ID, an exposure flag and a purchase flag. It does not need a home address.
Three numbers show whether a clean room program works: the match rate, the lift, and the revenue each partner campaign brings in. A fourth, the share of members who consented, sets the ceiling on the first.
Consent shapes who goes into the room. Members should be told how their data is used, and a withdrawal should remove them from the next match. Our post on first-party data strategy says a clean room lets two companies measure a shared campaign without trading their lists.
How GRAVTY handles data clean rooms
Loyalty Juggernaut’s retail solution describes a governed audience layer for CPG brands. Brands buy access to retailer audiences without touching raw personal data. The page calls it clean-room style: shared, governed and opt-in. The CPG sees lift, and the retailer keeps the customer. It adds consent-graph-aware activation and per-CPG settlement and revenue share. FEMSA runs it with 20 CPGs in Spin Premia, according to the page.
Frequently asked questions
What is a data clean room used for?
It runs four jobs. These are insights, ad measurement, richer audience profiles and campaign plans. They are the common uses the IAB lists (IAB and Orrick).
Is data in a clean room anonymous?
No. Most clean rooms work on pseudonymous identifiers, and those still count as personal data. The EDPB agrees (EDPB).
What do you need to use a data clean room?
You need at least two data sets and a common identifier to match them. The IAB Tech Lab says an identity partner can help create the match key when none exists (IAB Tech Lab).
Who owns the data in a clean room?
Each contributor keeps its own data and stays responsible for it. In most cases the contributors act as the business or controller. The provider is a service provider (IAB and Orrick).
Which technologies do clean rooms use?
Clean rooms use encryption, hashing and noise added to results. Some also use secure multi-party computation (IAB and Orrick).
