Snowflake: choosing the right multi-tenant strategy

by | May 27, 2025

Introduction

When developing modern SaaS applications, it's not uncommon to want to offer the same Snowflake software platform to several different organizations, be they corporate customers, subsidiaries of the same group, or partners. Each of these entities wishes to access its own data, benefit from a certain level of isolation, and sometimes even customize certain functionalities. This need to pool infrastructure while guaranteeing data isolation and security naturally leads to the adoption of a multi-tenant Snowflake architecture.

In this context, it is crucial to choose the right architecture to meet the specific needs of each customer, while optimizing costs, security and scalability. Snowflake offers three main architecture models for multi-tenant applications: Multi-Tenant Table (MTT), Object Per Tenant (OPT) and Account Per Tenant (APT). This blog explores these three models, their advantages, disadvantages and selection criteria.

 

1. Snowflake Mutli-Tenant Table (MTT)

Concept: Consolidation of tenants in a single table.

The logic diagram shown in Figure 1 illustrates a fairly common application configuration:

  • Firstly, application users access tenant data via secure views in a service database (highlighted in red).
  • Next, a table of rights controls which Snowflake users or roles have access to which tenants.
  • What's more, secure views ensure that application users only see their tenant's lines.
  • Finally, all tables are grouped by a tenant_id column.

To ensure that users can only see their tenant's rows, tenants query via secure views, which join the base tables to the tenant_id rights table. Common tables, where everyone can see all rows, use regular views pointing to the base tables.

Figure 1: A service database with secure views ensures that application users only see the rows of their tenant.

  • Benefits:
    • Simple architecture.
    • High scalability.
    • Low cost.
  • Disadvantages:
    • Rigid data model.
    • Complex security management.

Features:

  • Data model: Must follow the same structure across different clients. Automatic MERGE, UPDATE and CLUSTERING operations can be a challenge on very large tables.
  • Storage cost management: Difficult to determine storage costs per tenant in a multi-tenant table.
  • Warehouse cost management: Dedicated warehouse per tenant (easy to determine cost per customer but can increase overall cost) or shared between tenants (optimizes overall cost but difficult to determine cost per customer).
  • Security management: Requires advanced security management (tenant reference table, Secure views, RBAC (Role-Based Access Control) and RLS (Row-Level Security) push) to isolate access between clients. High risk of error.
  • Scalability: Relatively easy, with few new objects to create.

 

2. Snowflake Object Per Tenant (OPT)

Concept: Isolate tenants into separate objects (tables, schemas or databases).

OPT lets you isolate tenant data by database, schema and table, and use RBAC to control which user or role can view or query an object. Separating clients into their own databases is the most common practice (see Figure 2), as it's the simplest and cleanest level of isolation.

Figure 2: Isolating customers in their own databases is the most common OPT model.

  • Benefits:
    • Flexible data model.
    • Enhanced security.
  • Disadvantages:
    • Limited scalability.
    • Potentially higher costs.

Features:

  • Data model: may be unique or similar for each customer.
  • Storage cost management: partitioned storage for each customer in separate objects (table, schema, DB).
  • Warehouse cost management: Dedicated warehouse per tenant (easy to determine cost per customer but can increase overall cost) or shared between tenants (optimizes overall cost but difficult to determine cost per customer).
  • Security management: less complex than MTT, mainly based on RBAC.
  • Scalability: Creating objects in Snowflake is easy, but maintaining a consistent state across many similar objects is difficult.

 

3. Snowflake Account Per Tenant (APT)

Concept: Isolation of Snowflake accounts.

With the APT model, there is usually one Snowflake account, one warehouse and one database per tenant.

Figure 3: Account Per Tenant (APT) model

  • Benefits:
    • Maximum safety.
    • Strict insulation.
  • Disadvantages:
    • Complex management.
    • Limited scalability.
    • High costs.

Features:

  • Data model: may be unique or similar for each customer.
  • Storage cost management: partitioned storage for each customer in separate accounts.
  • Warehouse cost management: Dedicated warehouse per tenant (easy to determine cost per customer, but can increase overall cost).
  • Security management: The most secure option for partitioning customer data. Allows you to add account-level security options to further strengthen access to each customer's data (e.g. customer-specific encryption).
  • Scalability: Creating objects in Snowflake is easy, but maintaining a consistent state across many similar objects is difficult.

 

4. Detailed comparison of the MTT, OPT and APT approaches

Criteria MTT OPT APT
Data model Unique Flexible Flexible
Storage costs Shared, difficult to ventilate Partitioned Partitioned
Cost Warehouse Shared or dedicated Shared or dedicated Always dedicated
Security Complex to manage Easier Maximal
Scalability Very good Average Average

 

Questions to help you choose the right architecture:

What is the nature of each tenant's data?

  • Each customer has specific data modeling needs -> OPT
  • Data modeling requirements follow a standardized model ("Core Model") for all customers -> MTT

What level of security is required?

  • Need for maximum security and strict data isolation -> APT
  • Safety is important, but strict isolation is not necessary -> OPT

How many customers?

  • Large number of customers with similar data requirements -> MTT
  • Smaller number of customers with varied data needs -> OPT or APT

What's your budget?

  • OPT and APT architectures can lead to higher costs due to resource isolation.

What are your technical capabilities?

  • Setting up and managing OPT and APT architectures can require significant technical expertise. If you have limited technical resources, the MTT architecture may be easier to manage.

Conclusion

Choosing the right Snowflake multi-tenant model depends on balancing security, flexibility, cost and operational complexity. There is no one-size-fits-all solution - each customer context deserves a detailed assessment.

To find out more: Snowlake Multi Tenant.

You too can become a Snowflake expert with our tailor-made training courses: Snowflake Bilink training.

You can contact us here.

The following two tabs change content below.

Ayoub Arhiate

"At Bilink, we support our customers in implementing Snowflake architectures tailored to their business and regulatory challenges." Ayoub Arhiate, Senior data cloud engineer

Latest articles by Ayoub Arhiate (view all)

Share This