Anatomy of an Engagement Stack
A vendor-neutral reference model. Click a layer to see the patterns, the decisions that actually matter, and the trade-offs hiding behind them.
6 layers + a feedback loop. Click any to explore.
-
Execution across email, push, in-app, SMS, web, direct mail, and paid, with deliverability and rendering reality.
Key decisions
- Native channel coverage versus best-of-breed per channel
- Deliverability, warm-up, and sender reputation ownership
- Content and localisation model across channels
The trade-off
Consolidation simplifies operations; best-of-breed wins on any single channel. The gap shows up in deliverability long before it shows up in a demo.
-
Turning decisions into coordinated journeys across channels, as control systems rather than static flowcharts.
Key decisions
- Event-triggered versus scheduled, and the default for new journeys
- Global frequency and priority across competing programmes
- How experiments and holdouts run without fragmenting measurement
The trade-off
A beautiful canvas is not the same as a system that can continuously decide and improve. Flexibility for marketers versus governability for the business.
-
The system of intelligence that decides the next best action, increasingly continuous and increasingly agentic.
Key decisions
- Rules, models, or a mix, and who can change them safely
- Where the decision is made: central engine versus per-channel logic
- Guardrails, fatigue limits, and how autonomy is bounded
The trade-off
Centralised decisioning is consistent but can become a bottleneck. Distributed logic is fast but quietly diverges into contradictions.
-
The single, governed view of the customer: attributes, computed traits, consent state, and history.
Key decisions
- Real-time profile versus periodically rebuilt profile
- Which computed traits live centrally versus inside each tool
- Data residency, retention, and the right-to-be-forgotten path
The trade-off
A central profile reduces drift but adds latency and a dependency everyone leans on. Duplicated traits are fast until two systems disagree.
-
Stitching anonymous and known signals into a stable person across devices, channels, and time.
Key decisions
- Deterministic versus probabilistic matching, and the confidence threshold
- Identity graph ownership: in the warehouse, the CDP, or the engagement platform
- How consent and suppression travel with identity
The trade-off
Aggressive matching lifts reach but risks merging the wrong people. The cost of a false merge is almost always higher than a missed one.
-
Capturing behaviour and events from web, app, store, service, and back-office systems as they happen.
Key decisions
- Streaming versus batch, and where each is acceptable
- Event schema and a shared naming taxonomy
- Server-side versus client-side collection and consent at capture
The trade-off
Richer client-side data versus resilience to ad blockers and privacy controls. The taxonomy you skip now becomes the migration you pay for later.
Closing the loop: outcomes flow back as signals so the system can learn rather than just report.
Key decisions
- Attribution approach and the incrementality you can actually prove
- Which outcomes are fed back into decisioning, and how quickly
- A shared metric layer versus per-tool reporting
The trade-off
Last-touch is easy and wrong; incrementality is honest and slow. Measurement that only reports is overhead; measurement that feeds back is leverage.