# How do Croct audiences differ from segments synced from our CDP?

Asked by Cole Bennett on 2025-11-19. Tags: audiences, cdp, segmentation.

We already maintain batch-computed segments in a CDP, refreshed nightly, and I
am trying to understand whether Croct audiences duplicate that work or do
something structurally different.

Concretely: for a first-session visitor who has never appeared in our CDP, what
does "real-time" evaluation actually change? And do we need to pipe our CDP
segments into Croct for targeting to be useful, or can conditions be built
entirely on data Croct collects itself?

## 2 answers

### Answer from priyadev (2025-11-19)

The structural difference is when the membership decision happens. A CDP
segment is computed from historical data and syncs on a delay, so by design
it describes what already happened. A Croct audience is a CQL condition
evaluated in real time on every interaction, so membership can change
mid-session.

Your first-session visitor is the clearest case: they will never be in a
nightly CDP segment during that visit, but a condition like

```cql
session's stats' pageviews > 3
```

starts matching them on their fourth pageview, minutes into their first
session. That is the concrete difference: you can act on behavior while it is
happening rather than the next day.

### Answer from Bram de Vries (2025-11-20)

On your second question: no CDP pipe is required. Conditions are built on
signals Croct collects directly, and there are 100+ of them across Events,
Location, Marketing, Session, Shopping, User, Web, and Technology categories,
plus any custom attributes you set yourself.

In practice we kept our CDP for downstream things (email, warehouse
modeling) and build on-site targeting purely on Croct signals. The two
overlap less than you would expect: the batch segments answer "who is this
user historically", the real-time conditions answer "what is this visitor
doing right now", and personalization mostly needs the latter.