Retail availability platform: Azure SQL Database to Snowflake migration

One platform, zero disruption.

We migrated a business-critical retail availability platform from Azure SQL Database and two external engines onto Snowflake, without breaking a single downstream integration.

Retail / Consumer Goods
Data engineering
Software development
Project

1platform instead of three systems

0integrations broken

100%consumers moved in one cutover

5phase, validation-led delivery

The Challenge

What stood in the way.

The legacy platform worked, but it had not kept pace with growth. It relied on Azure SQL Database plus two separate engines, and every hand-off between systems added cost, delay and risk.

Fragmented logic

Rules and analytics were spread across three systems, which made changes slow and risky.

Maintenance overhead

Keeping schemas, jobs and contracts in sync consumed engineering capacity.

Scalability limits

SQL Server was nearing cost-effective scale as data volumes and file counts grew.

Inconsistent ingestion

Many file formats forced bespoke SSIS handling throughout the pipeline.

Our Solution

How we answered it.

We re-platformed the system onto Snowflake, replaced the external legacy sales engine with a native Availability Analytics Engine (AAE), standardised ingestion with a serverless File Preprocessor and served all consumers through a single API.

ProblemT-SQL behaves differently on Snowflake.
Our answerWe built a shared conversion catalogue, mapping patterns such as temporary tables, scalar UDFs, MERGE and procedural T-SQL to Snowflake-native equivalents
ProblemThe new engine had to match legacy results exactly.
Our answerWe validated the AAE against legacy outputs in a parallel run on real data, reconciling record counts, Lost, Baseline and Expected Sales, alerts and intervention outcomes.
ProblemMany inconsistent source file formats.
Our answerThe File Preprocessor normalises everything before it reaches Snowflake.
ProblemZero tolerance for disruption.
Our answerContract-first design kept every integration interface fixed, followed by a single coordinated cutover.

How It Works

The parts that made the difference.

Availability Analytics Engine

Native calculation of Lost Sales Value, Baseline and Expected Sales inside Snowflake, replacing the external legacy engine.

Smart alerting

Business rules flag at-risk store-product combinations for field and head-office teams.

File Preprocessor

An Azure Function validates and standardises every source format before it reaches Snowflake.

Layered data quality

Validation, cleaning and quality gates run before data is promoted to the Curated layer.

Results

What changed for the client.

The client now runs a simpler, scalable platform that turns raw retail data into measured field impact. Azure SQL Database, the SSIS pipelines and the legacy sales engine were retired, and every field