# Can we run Croct analytics without collecting personal data?

Asked by Ilse Brandt on 2026-04-22. Tags: privacy, analytics, gdpr.

Good morning. I coordinate data protection at a publisher and I am assessing
whether we can operate personalization and experiment analytics strictly on
behavioral data, with personal attributes never populated.

To be precise about categories: by personal data I mean directly identifying
attributes such as name, email address, or telephone number. Behavioral data
(pages viewed, session activity, campaign source) we treat separately in our
records of processing.

If we never call any identification function, what does a profile then actually
contain? And does targeting remain functional on behavioral data alone, or does
the product degrade without personal attributes

## 1 answer

### Answer from annika_w (2026-04-22)

This maps cleanly onto how profiles are structured. Croct profiles hold
statistical, behavioral, and personal data as separate categories, and the
personal category only exists if you put data there. Attributes like name,
email, and phone are populated exclusively through `croct.identify` and
`croct.user.edit()` calls that your own code makes. If you never call them,
profiles stay anonymous behavioral profiles, which is the default state, not a
degraded mode.

Targeting does not depend on the personal category at all. Audiences evaluate
on first-party signals: behavior, campaign source, location, device,
interests. Experiments assign and measure on the same basis. So a condition
like `session's stats' pageviews > 3 and campaign's medium is "cpc"` works
identically whether or not a single personal attribute exists anywhere in your
workspace.

For your records of processing, the retention side: profiles, sessions, and
events are kept within plan-bounded windows (1, 30, or 60 days depending on
plan), analytics is kept 90 days on all plans, and profiles inactive for 90
days are deleted. So the behavioral data you do process has documented,
bounded lifetimes you can cite directly.