Customer Data Layer: Architecture and Technical Design

Overview The National Museum of Australia is seeking the design of its customer data layer, covering the architecture and technical design required to bring together customer data currently held across separate systems. Purpose The Museum holds rich data about its audiences but cannot see them as whole people. Customer information is split across systems that do not connect: Tessitura for memberships, donations and ticketing, Shopify for retail, website analytics for online behaviour, and email platforms for campaign responses. A single supporter who buys a ticket, joins as a member, gives a donation and shops online can appear as four separate records. As a result, the Museum cannot reliably identify its most valuable audiences, explain what drives repeat visits, or measure which campaigns lead to visits, memberships or gifts. Marketing is largely broadcast rather than targeted, and donor and member journeys cannot be traced from first contact through to long-term engagement. Objectives The objectives below are the capabilities the architecture must be designed to support: Design a single, auditable customer view: resolve each person to one record across the connected source systems, using deterministic matching on known identifiers, or fully governed probabilistic logic where argued. The design must be build-ready and include the matching logic, a data-rectification runbook, initial estimates of duplicate and partial records with a verification strategy, and confidence scoring Design identity resolution across touchpoints: clean, de-duplicate and stitch disparate identifiers, including emails, phone numbers, device identifiers and anonymous website identifiers, into a deterministic or governed probabilistic identity graph, so a person is recognised consistently across systems Design the engagement measurement architecture: every engagement number must trace to a definition and a source in the NMA data dictionary. The design must support the Museum's three-layer model of reach, engagement and commitment, fed by paid, earned, shared and owned media, so reported figures can be audited and stood behind Design for self-service segmentation and governed activation: once built, non-technical teams must be able to create and maintain audience segments without custom engineering, and push them to downstream tools under consent and preference controls Design governed consent, retention and compliance: centralise consent and communication preferences and enforce them across connected platforms, and automate data retention and deletion, including acting on requests to correct or delete personal information. The design must support the Privacy Act 1988 and the Australian Privacy Principles, and align with the Museum's obligations under the Archives Act 1983 Design for downstream personalisation and analytics: supply downstream experience platforms and analytics with governed, low-latency profile data, so later phases can add personalisation, next-best actions and suppression of converted audiences without rework Design out duplicated effort: remove the redundant data-stitching scripts and manual integration the Museum relies on today, giving every system one trusted source and reducing custom engineering for routine integration Scope The scope includes the following deliverables to be considered when preparing a response: A lakehouse design that can be implemented by a build vendor not involved in the original design, detailing how data is ingested, cleansed and served to downstream consumers An identity resolution design that automatically resolves individuals across the NMA's source systems into a single, connected record A data model built on the NMA's data dictionary definitions, carrying the entities that measurement, scoring and consent depend on An enterprise architecture for the NMA showing how the platform, its services and its data fit together, and how and where future workstreams plug in