Insights in a Silo: Why Banks Have Enough Data but Do Too Little With It

Hypothetical example: For three months, a customer transfers large sums to an external savings account and opens the fixed-term deposit calculator twice in his banking app. On Thursday evening, his advisor calls him — not about his liquidity, but to pitch an instalment loan from a campaign spreadsheet. Payments, app analytics and the campaign tool had each recorded the signal, yet no system connected them. That is where tackling data silos in banks breaks down in practice: the raw data exists, but the bridge from insight to action is missing.

Many banks have invested in data lakes and reporting for years without seeing better decisions at the customer touchpoint. The bottleneck almost never lies in storage, but in missing activation across departmental boundaries.

At a glance

• Data silos in banks rarely stem from a lack of data; they arise because payments, product, channel and consent data sit in separate systems without a shared activation layer.

• Even in strictly supervised risk data aggregation under BCBS 239, the Bank for International Settlements found that only 2 of 31 global systemically important banks were fully compliant.

• A unified customer profile does not require a multi-year core banking replacement, but a logical data layer with clear data lineage, consent checks and direct channel integration.

Where insights get stuck in banks

Insights in banks usually get stuck between analytics and operational channels because reports explain the past without triggering a verified action in the target system. A report shows last month, whereas the customer's decision happens right now.

Data activation is the process of translating an analytical signal from account activity or channel usage — subject to consent and business rules — directly into an operational workflow. In many banks, however, the data team's work ends with a dashboard or a monthly CSV export. By the time that list reaches the advisor's screen, the customer's need has already passed.

Banking supervision shows how deep structural data silos in banks run. In its progress report on the BCBS 239 principles for risk data aggregation, the Bank for International Settlements (BIS, November 2023) found that nearly ten years after the standard was issued, only 2 of 31 assessed global systemically important banks (G-SIBs) complied fully with all eleven principles. In the April 2020 BIS report, that figure stood at 0 of 30 banks after 3 of 29 had been rated fully compliant in 2017; not a single principle is fully implemented across all banks. Likewise, the European Central Bank (ECB, May 2024) cites fragmented IT landscapes, legacy systems and unclear data ownership in its Guide on effective risk data aggregation and risk reporting (RDARR). Although these findings assess risk data, it is a fair conclusion that the same legacy boundaries also hold back customer data.

Chart: Fully compliant G-SIBs under BCBS 239 (2017, 2019 and 2022 assessments) and structural root causes cited by the BIS (2020, 2023) and ECB (2024).

The four typical data silos in banks

The four most persistent data silos in banks run between core banking and payments, digital channels, sales and service, and legal and compliance. Each domain stores a vital piece of the customer relationship, but uses its own keys, cycles and formats.

A data silo is a data repository that is technically or organisationally isolated so that other teams cannot use it consistently and in time within their workflows. In everyday banking, four separate islands appear repeatedly:

How banks move from isolated spreadsheets to coordinated steering is covered in our articles on Next Best Action in banking and Customer Value Management for banks.

What a unified customer profile requires

A unified customer profile requires a shared business taxonomy, traceable data lineage, built-in consent verification and interfaces for batch and real time. It does not replace the core banking system, but connects existing sources through a logical layer.

A unified customer profile (Customer 360) is a consolidated, purpose-bound view of a person's master data, product usage, transaction signals, contact history and valid consents. In its January 2026 BCBS 239 newsletter, the Basel Committee notes that several banks now extend principles from risk data governance — clear data ownership and end-to-end data lineage — to enterprise-wide data management.

Data lineage is the complete traceability of a data attribute from its capture in the source system through every calculation to the final decision. Every attribute in the profile must show which source it came from, how fresh it is and which purpose it may serve under the GDPR. How this layer delivers signals across the app, web and advisory is explained in our article on personalisation as the activation layer.

With its Data Layer, ACCELERAID provides a banking data layer that brings together transaction, channel and CRM data, builds categories and attributes, and checks consents before any signal is passed on. The platform is SOC 2 compliant, built on GDPR by design and hosted in an ISO 27001 data centre in Frankfurt or on-premises on request; models such as OpenAI via Azure (Germany region) and Gemini via Google Cloud (Frankfurt) can be connected, while domain agents, orchestration and the control plane are in development.

Self-check in 7 questions

Seven practical questions allow chief data officers and business leads to check quickly whether their bank has a data problem or an activation problem. More than two negative answers point to a gap on the last mile:

1. Identity: Does one customer ID link accounts, cards, app usage and advisory notes without manual matching?

2. Freshness: Are key account events such as large outflows available by the next business day or in real time?

3. Consent: Is valid GDPR consent per channel checked automatically at the moment of delivery?

4. Data lineage: Can teams trace which source fields and rule generated every displayed customer prompt?

5. Contact suppression: Does a central contact memory stop promotions on the day of an open complaint?

6. Feedback loop: Do responses from branch, service and app flow back into the profile automatically?

7. Autonomy: Can business teams define new triggers without waiting months for an IT release?

Frequently asked questions

Why do banks still have data silos despite having a data warehouse?

A data warehouse or data lake usually stores data for retrospective reporting, financial controlling and regulatory filings. For operational customer interaction, it typically lacks real-time interfaces, automated per-channel marketing consent checks and a direct feedback loop into branch advisory, service workspaces and the mobile banking app.

How does a unified customer profile differ from a data lake?

A data lake collects raw data from many systems in its native format for analytics and data science. A unified customer profile, by contrast, combines verified, business-translated attributes, recent events, contact history and valid consents for an individual so that operational channels can act on them immediately.

Do banks have to replace their core systems to break down data silos?

No, replacing the core banking system is neither necessary nor economically sensible for this purpose. Proven practice is to introduce a logical integration and activation layer above existing systems that reads relevant events via defined interfaces, standardises them, verifies privacy rules and routes them to the right channels.

If you would like to explore how to turn existing bank data into actionable customer insights without rebuilding your core systems, get in touch with us.

Illustration: AI-generated. AI-assisted content: We use AI technologies and automated agents in the creation of our articles, including from Microsoft, Google, OpenAI, Anthropic and other providers. Topics, editorial direction and final approval remain with our team.