From Legacy Data Warehouse to Snowflake: The Bridge Enabling Marketing, Sales, and Service to Work Directly

Part 2 of 2 of our "Data is the Key for AI" series — and conclusion of the series. Part 1 showed why AI projects fail due to the data layer (Data is the key for AI: Why AI projects fail at the data layer). This part shows what the technical bridge between legacy systems and a modern data platform actually looks like.

The Spaghetti Architecture as a Bottleneck

Many banks know the problem from their own experience: data flows that have grown over years without ever having a consistent target picture behind them. McKinsey accurately describes this pattern in its analysis of banking data architecture as "spaghetti architecture" — characterized by fragmented data warehouses and data lakes, as well as the parallel operation of old and new platforms. The consequences are concretely measurable: delayed decisions, increased security and compliance risks, and rising costs for maintaining multiple complex environments simultaneously (McKinsey, Getting the data architecture right in banking).

The economic leverage of choosing the right architecture is significant. Banks that select the appropriate data architecture archetype can, according to McKinsey, cut implementation time in half and reduce costs by 20 percent. Five identified best practices bring an additional 20 percent lower platform construction costs, 30 percent faster time-to-market, and 30 percent lower change costs (McKinsey). This aligns with the insight described in Part 1: The data layer determines the success or failure of AI and digitalization projects — here additionally marked with a clear price tag.

Why Cloud Data Platforms are Becoming the Standard in the Financial Sector

The fact that this realignment is not just a theoretical recommendation is demonstrated by the market penetration of cloud data platforms in the financial sector. Already in 2021, around 57 percent of the financial services companies on the Fortune 500 list used Snowflake, with reference customers including Allianz, AXA, BlackRock, Capital One, NYSE, Refinitiv, Square, State Street, Western Union, and Wise (Snowflake, Launch of the Financial Services Data Cloud). Snowflake itself explicitly highlights the addressed legacy issue: outdated processes of duplicating data and sending files back and forth between systems.

As of January 31, 2026, Snowflake has more than 13,900 customers worldwide, including 790 from the Forbes Global 2000 list — an increase of 5 percent compared to the previous year. Product revenue in the fourth quarter was 1.23 billion US dollars, an increase of 30 percent compared to the prior-year quarter; in fiscal year 2026, 4.47 billion US dollars in revenue was achieved, an increase of 29 percent. Snowflake explicitly names "highly-regulated markets such as financial services" as a growth market and lists Capital One as a reference customer (Snowflake, Q4/FY2026 Financial Results). Quantified customer results from the financial sector underscore the effect: State Street reports, according to the provider, 25 times higher productivity of data operations teams and 87 percent fewer false data error alerts; Chicago Trading Company achieved a cost savings of 54 percent after replacing managed Spark (according to provider Snowflake, AI Data Cloud for Financial Services).

The Bridge Instead of the Big Bang Migration

The decisive difference to the classic migration strategy lies in the approach: not replacement, but integration. A big-bang migration, where core banking systems, card processing, and CRM are migrated to a new target architecture in a single step, is neither financially nor operationally viable for most institutions — the risks of parallel system landscapes described by McKinsey would likely worsen rather than resolve during the transition phase. The more practical way is the bridge: legacy data warehouses and data lakes remain in place for now, but are connected via connectors to a modern cloud data platform like Snowflake, enabling real-time aggregation from CRM, core banking, and card processing without immediately replacing existing systems.

Technically, Snowflake's Secure Data Sharing feature makes this particularly attractive: according to the provider, with zero-copy sharing, no actual copy of the data is transferred between accounts. Sharing runs through the services layer and the metadata store of the platform; shared data does not consume additional storage space for the recipient and generates no storage costs there — billing is based only on the actual computing power used. New or updated objects in a share are instantly available to all recipients, and access remains revocable at any time (Snowflake Docs, About Secure Data Sharing). For banks, this means that the repeatedly criticized practice of sending files back and forth between departments and systems is eliminated, without legacy systems having to be replaced immediately.

From Aggregation to Activation in Marketing

A consolidated data basis in the cloud is the prerequisite, but not yet the activation. The final step of the bridge leads the aggregated data from the warehouse back into the systems in which marketing, sales, and service work daily — via Reverse ETL. Providers like Hightouch read data directly from the Snowflake warehouse and synchronize only the actually changed records into downstream target systems, connected for example via Snowflake Partner Connect. From a scale of more than 100,000 rows, the provider recommends a sync engine with change data capture directly in its own Snowflake schemas for this (Hightouch Docs, Snowflake Source).

Specifically, the activation path into a customer engagement platform looks like this: Hightouch synchronizes a warehouse model or an audience into a Braze Cohort; this creates a segment in Braze that can be used as a filter for Campaigns or Canvas. Important to note: no new user profiles are created; instead, existing Braze profiles are only added to or removed from the cohort (Braze Docs, Hightouch Cohort Import). This warehouse-centric architecture has long been mainstream: 71 percent of surveyed martech and marketing operations professionals have a cloud data warehouse or data lake in their stack, with 47 percent feeding data one-way into the warehouse and 29.2 percent synchronizing data one-way from the warehouse to martech applications (chiefmartec/MartechTribe, „Martech for 2025").

What this Means for Banks in the DACH Region

For institutions that continue to work with Finastra, Temenos, or SAP Banking-based core systems, the bridge architecture is the more realistic path compared to a complete re-implementation. The acceleraid platform offers corresponding connectors to Finastra, Temenos, and SAP Banking as well as a connection to Snowflake, so that existing core banking and card processing systems can be connected to the modern data layer without requiring a big-bang replacement (acceleraid Platform). This allows the governed data foundation described in Part 1 — System of Record, Consent, Lineage, PII protection — to be realized even if an institution is not yet ready to completely replace its core banking system (acceleraid Banking).

This bridge logic also applies when connecting existing martech systems: how platforms like Braze, Adobe, or Salesforce Marketing Cloud can be integrated into the existing system landscape without rip-and-replace was described in the first part of our "Future-Ready Martech" series (No Rip and Replace: Seamless Integration with Braze, Adobe, Salesforce). Together with the data governance foundation described in Part 1 of this series, a complete picture emerges: governance and architecture are two sides of the same prerequisite for AI to actually become productive in banking — not as a one-off migration project, but as a continuously evolved bridge between existing systems and new capabilities.

Illustration: AI-generated. AI-supported content: In creating our posts, we use AI technologies and automated agents, including from Microsoft, Google, OpenAI, Anthropic, and other providers. Topics, editorial direction, and final approval lie with our team.