Customer 360 Integration: Architecture and Platform Choices

Architecture sheets map systems to customer profile

Customer 360 integration is the engineering work that connects your CRM, billing, service, web, mobile and back-office systems so that every customer resolves to one trusted profile. The tool you buy matters less than the integration layer beneath it: how data is ingested, how identities are matched, who owns the definitions and how fast changes travel. This guide sets out that architecture for CEOs, CTOs and integration managers who need a unified customer view that holds up in production.

Polarising
polarising.com
Connect Customer Data Across Systems
Polarising tackles complex integration challenges, building AI systems that connect with business processes and account for governance from the outset.

Explore AI engineering

Key Takeaways

  • A Customer 360 is an integration and identity problem first. Dashboards and AI features only work when the data underneath is connected, matched and governed.
  • Most enterprises need both a CDP and master data management (MDM): MDM supplies governed golden records, the CDP activates them with behavioural data.
  • Use batch ingestion for history and reference data, and event-driven or streaming patterns for the changes that must be reflected quickly, such as consent updates.
  • Start with a few high-value use cases, assign data stewards early and agree one set of customer definitions before building pipelines.
  • Integration platforms such as Boomi, MuleSoft, TIBCO and Solace each cover a different part of the connective tissue, and the right mix depends on your estate.

Table of Contents

What a Customer 360 actually is

A Customer 360 gives the business a complete view of each customer across every touchpoint and interaction. Salesforce describes it as bringing unified customer data from many sources into one platform, so that teams can personalise interactions, improve service and spot cross-sell and upsell opportunities.

For an enterprise, that definition hides the hard part. The customer exists in a core banking or policy system, a CRM, a billing engine, a contact-centre tool, an e-commerce platform and probably a handful of spreadsheets. Each has its own identifier, its own data model and its own owner. “Customer 360” is the outcome. Integration is how you get there.

The results are mixed in practice. A Gartner-linked figure quoted in AWS guidance on Customer 360 says only 14% of organisations have successfully implemented one. The reasons given are a lack of consensus on what a 360-degree view means, data quality problems and weak cross-functional governance. None of those is a software gap. All three are decisions about integration, ownership and definitions.

How customer data comes together

The reference architectures from the large cloud and CRM vendors converge on the same shape: collect, cleanse, resolve identity, then activate.

Four stages of Customer 360 data integration

Sources you will need to connect

AWS lists typical sources for a customer data platform: contact centres, email, web and mobile events, point-of-sale transactions and CRM systems. Its Customer 360 guidance also names advertising and social media events, plus transactional data from systems of record and third-party datasets.

In a regulated enterprise the list is usually longer and older. Policy administration, claims, core banking, network provisioning, patient or order management systems all hold customer facts. Many have no modern API. This is where integration engineering earns its fee: exposing legacy data safely, without a rewrite, so it can join the unified view.

Batch and streaming ingestion

Salesforce’s Data 360 integration patterns separate two layers that are useful for any platform:

  • Batch ingestion is the foundation for onboarding large historical datasets from cloud storage and from CRM clouds, harmonising them into the unified layer for identity resolution, segmentation and analytics.
  • Streaming and real-time ingestion covers event-driven ingestion through an API, real-time web and mobile behaviour for immediate personalisation, and near-real-time CRM synchronisation so that customer attributes and consent updates show up quickly.

The practical rule: batch what is large and slow-moving, stream what is time-sensitive or legally sensitive. A consent withdrawal that reaches your marketing platform tomorrow morning is a compliance problem. A product catalogue that refreshes overnight is not.

AWS’s composable CDP guidance shows the same split in its own terms: batch data lands through connectors into object storage with transformation and quality checks, while real-time data arrives through streaming services. Different vendors, same pattern.

Identity resolution is the hard centre

One practitioner view holds that a Customer 360 is an identity resolution and data unification problem, not merely aggregation, producing an ID graph that maps each system-specific identifier to a canonical account and contact. The same source warns that merging records that should stay separate, or leaving apart records that belong together, can corrupt downstream workflows.

Salesforce’s Data 360 architecture describes a sound way to do this. Match rules can be exact, fuzzy or normalised. Reconciliation rules then pick the best-quality values to build a “gold standard” profile. Link tables tie source records to the unified individual while the original records are preserved. Keeping the originals matters: you can always explain why a profile looks the way it does, and unpick a bad merge.

AWS adds that identity processing also needs to identify anonymous logins, link to external data and group customers into domains of interest.

CDP vs MDM: which do you need?

This is the question leaders ask most often, and the answer is usually both, with clear boundaries.

Customer data platform (CDP) Master data management (MDM)
Main purpose Unify behavioural, transactional and interaction data for marketing activation Govern authoritative reference data across the enterprise
Scope Customer-centric Several domains: customers, products, suppliers, locations
Typical users Marketing and customer experience teams IT and data governance teams
Processing Real-time activation Mostly batch, some near-real-time
Strength Personalisation and consistent experience across channels Data quality, stewardship, regulatory compliance

The table draws on several explainers, including Dataversity’s comparison, which recommends MDM when the priority is governance, data quality and compliance, and a CDP when the priority is marketing and customer experience. A Treasure Data analysis describes MDM as the trusted foundation of governed records and global IDs, with the CDP enriching those records with behavioural data and activating the unified view.

The same analysis offers advice worth keeping: choose technology for the use cases you need to solve, not for what your existing vendor happens to sell. CDP.com makes the trigger conditions concrete. You need MDM when several enterprise systems must share consistent reference data and regulation demands auditable stewardship. You need a CDP when marketing needs real-time behavioural profiles.

For a bank, insurer, hospital group or telecoms operator, MDM often comes first because the golden record is also a compliance record. If you are weighing that route, our guide to master data management for enterprise leaders goes deeper. Either way, both products need feeds from the same operational systems, which is why the integration layer should be designed once and shared.

CRM data integration: where the single view starts

CRM is rarely the whole customer, but it is usually the anchor. Salesforce’s single-view guide says its Customer 360 Data Manager connects Salesforce apps and other customer data to create a single global ID and profile per customer. It also says the MuleSoft Anypoint Platform bridges Customer 360 to on-premises, cloud and legacy systems through APIs, unlocking ERP and database data to synchronise orders and automate workflows.

The lesson holds whichever CRM you run. CRM integration should not be a point-to-point tangle. Expose customer, order and entitlement data as reusable APIs and events, and let the CRM, the CDP and the MDM hub consume the same interfaces. When a new channel arrives, you connect it once rather than rebuilding links to six systems.

That is also the foundation of connected channels. If your goal is a consistent experience from branch to app to contact centre, see our piece on the enterprise architecture behind omnichannel integration. For the commercial side of the same goal, this overview of omnichannel marketing and customer experience shows what a unified record makes possible once the plumbing is in place.

CRM data integration: where the single view starts

Architecture and governance that keep it reliable

A unified view that nobody trusts is worse than none. Four practices recur across the vendor guidance and the field experience.

  1. Governed pipelines. Salesforce’s integration patterns call for schema consistency, lineage tracking and mastering rules so the unified layer and external platforms stay consistent and timely.
  2. Raw first, then modelled. Land data in its original schema, then map it to a common customer model. Salesforce’s architecture does this with data lake objects mapped to data model objects, supporting batch and streaming transformations such as trimming and standardising fields.
  3. Centralise, don’t copy. Snowflake’s maturity model has four stages, from siloed data to an intelligent Customer 360, and advises organisations in the first two stages to centralise on a single governed platform that avoids data copies.
  4. Classify and mask. Tagging, masking and classification policies belong in the design from the start, particularly for health and financial data. Governance bolted on later is expensive and rarely complete.

The vendor sources above describe their own platforms, so treat them as patterns to adapt rather than prescriptions. Your estate, regulators and data residency requirements decide the final design.

The challenges that stall Customer 360 programmes

The obstacles are consistent across sources. Informatica lists data silos, proliferating formats, restrictions from privacy and security processes, and inaccurate data that creates rework. Databricks adds siloed ownership, inconsistent definitions and scope creep. Workato points to the sheer volume of integrations, with few people able to build them, which stretches timelines.

Here is how we suggest leaders respond:

  • Agree what “360” means for your business before any build. Name the three or four decisions the view must support.
  • Assign data stewards early, across business and IT, and keep metric definitions in one shared catalogue.
  • Clean and deduplicate at source or in the first pipeline stage, not in the dashboard.
  • Start with a few high-impact use cases and expand in increments, as the Databricks guidance recommends.
  • Plan integration capacity. The volume of connections is a delivery risk, so decide whether you build, augment or bring in a specialist team.

Making the view ready for AI

Unified customer data is what makes AI useful in service, sales and risk. It is also what makes AI risky if the controls are missing. Salesforce notes that Data 360 governance includes tagging, masking and classification policies, with an Einstein Trust Layer for data protection. Whatever the platform, the same questions apply. Which attributes may a model see? Is consent current? Can you trace an output back to the source records?

Those answers come from the integration layer: lineage, consent propagation and access control. If you are connecting language models to this data, our guide to LLM integration patterns for enterprise architects covers how to avoid rebuilds later.

Choosing integration platforms: Boomi, MuleSoft, TIBCO and Solace

No single product connects an enterprise. We work across four integration platforms, Boomi, MuleSoft, TIBCO and Solace, and choose between them by the job in front of us rather than by habit.

In broad terms, they map to different layers of a Customer 360:

  • API-led connectivity exposes customer, order and account data from CRM, ERP and legacy systems as reusable services. MuleSoft is the platform most often associated with this approach, as the Salesforce guidance above shows.
  • Cloud and application integration moves data between SaaS applications, databases and the unified layer, with managed connectors and low-code flows. Boomi is a common choice here.
  • Messaging and enterprise integration suits estates with established message-based or hybrid architectures, where TIBCO is often already present.
  • Event streaming distributes changes, such as an address update or a consent withdrawal, to every subscribing system as they happen. This is Solace’s territory.

A single customer programme can use more than one. Batch loads and APIs feed the master record, events keep downstream systems current, and the choice follows what you already run, your latency needs and your team’s skills. That is why we start with an assessment of the existing estate before recommending a platform. Our enterprise integration services and data integration services pages describe how we scope that work.

A phased route to a unified customer view

  1. Define outcomes. Pick two or three use cases, such as a single service view, churn signals or consent-aware marketing, and the metrics they move.
  2. Map the estate. List every system holding customer facts, its owner, its identifier and its integration options.
  3. Set the canonical model and identity rules. Decide match logic, survivorship rules and who resolves disputes.
  4. Build the integration layer. Batch for history, APIs for lookups, events for change. Add lineage and monitoring from day one.
  5. Stand up the unified layer. MDM, CDP or a governed data platform, depending on your use cases.
  6. Activate and extend. Feed service, marketing and AI use cases, then add sources and domains in increments.

This order reflects the approach most guidance converges on: governance and identity before activation, small scope before wide scope.

Our perspective

We have spent over 20 years building and connecting enterprise systems, and our view is simple: Customer 360 succeeds or fails in the integration layer. The business case is rarely the problem. The problem is a core system that exposes nothing useful, two customer identifiers that never match, or a consent flag that lives in one application and nowhere else.

That is why we design governance and accountability in from the outset, rather than adding them once the data is flowing. Our approach is to treat identity, lineage and access as engineering requirements, not documentation tasks.

We also believe senior people matter more here than in most projects. Legacy-heavy estates in finance, healthcare, telecommunications and logistics reward engineers who have seen the pattern before. Our senior nearshore teams in the UK and Portugal are built for that, and our engagement models let you adjust team size and seniority as the programme moves from discovery to scale.

Choose the use cases first, connect the systems once and reuse the interfaces everywhere. That is how a unified customer view becomes a platform for AI rather than another silo.

Polarising Team

How Polarising can help

If your customer data sits across legacy platforms, CRM and cloud services, we can assess the estate, design the integration layer and deliver it with senior nearshore engineers. Our work covers API and event platform integration, legacy modernisation and migrations, with experience in regulated, high-volume environments across finance, healthcare, telecommunications, logistics and retail.

Polarising

Start with our enterprise integration services, see how we apply AI engineering from pilot to production, or look at the industries we serve. To talk through your own estate, reach us through Polarising.

FAQ

What is a Customer 360 system?

It is a set of integrated systems and data pipelines that bring customer data from every touchpoint into one unified profile per customer. Teams use it to personalise interactions, improve service and find cross-sell opportunities. The profile depends on integration and identity resolution beneath it.

What is the difference between a CDP and MDM?

A CDP unifies behavioural, transactional and interaction data so marketing can activate it in real time. MDM is an IT-led discipline that governs authoritative records across several domains, such as customers, products and suppliers. Large enterprises often use both, with MDM providing the governed records and the CDP adding behaviour.

How do you unify customer data from many systems?

Connect each source through batch loads, APIs or events, standardise it into a common model, then resolve identities with match and reconciliation rules. Keep the original records linked to the unified profile so merges can be explained and reversed. Governance and ownership should be agreed before the pipelines are built.

Why do Customer 360 projects fail?

The usual causes are siloed ownership, poor data quality, inconsistent definitions and scope creep. Integration volume is another factor, because few people have the skills to build the connections. Starting with a few high-impact use cases and naming data stewards early reduces these risks.

Sources

Content strategy and optimisation for Google and AI-powered search platforms by Velveteen SaaS — specialists in online visibility and customer acquisition.