Background
Most organisations resolve customer identity overnight, in batch (L2). That is too late for anything that has to happen while the customer is still there — the anonymous session that would convert if the screen recognised them, the opt-out that must take effect before the next message goes out. Going real-time (L3) closes that gap, but not the way it’s usually imagined. It does not re-match or rebuild identity on the fly — that heavy work stays overnight. Instead it does event-to-profile stitching: when a live event arrives carrying an identifier, that identifier is looked up against the already-built profile index and the event is attached to the existing unified profile, in under 50 milliseconds. The real work — and what this playbook scopes — is deciding which use cases justify real-time, and making sure each event actually carries an identifier that can be looked up.
What Is This Workbook
This is a self-serve, platform-agnostic toolkit for taking customer identity from overnight batch (L2) to real-time (L3). It sets out the method: the two-plane model that separates the heavy identity-matching (which stays overnight) from the fast real-time lookup; the two filters that decide which use cases are worth doing in real time; and the engineering output — the data fields each use case needs on its event and the architecture to source them. Includes the methodology, worked examples, and a flow graphic.
Benefits
- Corrects the costly misconception that real-time means resolving identity on the fly — the two-plane model that keeps you from building the wrong architecture.
- Two filters that kill non-viable use cases before you build: value + real-time activation, then the deterministic-key check.
- Produces a concrete engineering deliverable — the fields per use case, split into ready-to-stitch and must-be-sourced, with the sourcing architecture.
- Keeps your proven L2 batch graph as the source of truth.
Who This Is For
Data architects, identity leads, and martech/marketing-ops owners who’ve deployed L2 (batch, multi-signal resolution) and want to act on identity in real time — in-session personalization, live suppression, on-connect servicing.