Multi-Tenant SaaS Analytics

A complete working example showing how multiple tenants share the Akili platform with full data isolation. Each tenant has its own data products, connections, quality SLAs, and governance policies — all running on shared infrastructure with per-tenant resource quotas.

%%{init: {'flowchart': {'curve': 'basis'}}}%%
flowchart TB
    subgraph Tenant_A["Tenant: acme-corp"]
        A_SRC[Raw Orders\nCDC from PostgreSQL] --> A_AGG[Daily Revenue\nSQL Aggregation]
        A_AGG --> A_SERVE[Analytics Dashboard]
    end

    subgraph Tenant_B["Tenant: globex-inc"]
        B_SRC[Raw Transactions\nIncremental Sync] --> B_AGG[Weekly Summary\nSQL Aggregation]
        B_AGG --> B_SERVE[Lookup API]
    end

    subgraph Platform["Shared Platform Infrastructure"]
        PG[(PostgreSQL\nRLS per tenant)]
        SR[(StarRocks\nCatalog per tenant)]
        CEPH[(Ceph RGW\nPrefix per tenant)]
        RP[(Redpanda\nTopics per tenant)]
    end

    A_SERVE --> SR
    B_SERVE --> PG
    A_SRC --> CEPH
    B_SRC --> CEPH
    A_AGG --> RP
    B_AGG --> RP

Scenario

Two SaaS customers (tenants) use the same Akili deployment:

Tenant	Slug	Use Case	Classification	SLA
Acme Corporation	`acme-corp`	E-commerce order analytics	`confidential`	24h freshness
Globex Inc	`globex-inc`	Financial transaction summaries	`restricted`	1h freshness

Each tenant has independent:

Data products and connections
Quality checks and SLAs
Governance classification policies
Resource quotas
StarRocks analytics catalogs

Tenant Setup

Creating the Tenants

# Admin creates both tenants
akili tenant create acme-corp --display-name "Acme Corporation"
akili tenant create globex-inc --display-name "Globex Inc"

# Verify both are active
akili tenant list

Setting Up Connections (Per Tenant)

Each tenant registers its own connections. Connections are tenant-scoped and invisible to other tenants.

Acme Corporation:

# Switch to acme-corp profile
akili connection create \
  --name acme-orders-db \
  --connector-type postgres \
  --config '{
    "host": "orders.acme-corp.com",
    "port": 5432,
    "database": "production",
    "ssl_mode": "require"
  }'

akili connection test acme-orders-db

Globex Inc:

# Switch to globex-inc profile
akili connection create \
  --name globex-finance-db \
  --connector-type postgres \
  --config '{
    "host": "finance.globex-inc.com",
    "port": 5432,
    "database": "transactions",
    "ssl_mode": "require"
  }'

akili connection test globex-finance-db

Acme Corporation: Daily Revenue Pipeline

product.yaml

product:
  name: daily_revenue
  namespace: sales
  version: "1.0.0"
  archetype: source
  description: >
    Daily revenue aggregation from production order database.
    Classified as confidential due to financial data.
  owner: data-team@acme-corp.com
  tags:
    - revenue
    - daily
    - financial

inputs.yaml

inputs:
  - name: orders
    kind: source
    connection: acme-orders-db
    resource: public.orders
    strategy: cdc
    cursor_field: updated_at
    schedule: "0 */4 * * *"
    columns:
      - order_id
      - customer_id
      - product_id
      - amount
      - currency
      - region
      - order_date
      - created_at

  - name: products
    kind: source
    connection: acme-orders-db
    resource: public.products
    strategy: incremental
    cursor_field: updated_at
    schedule: "0 6 * * *"
    columns:
      - product_id
      - product_name
      - category
      - unit_cost

transform.sql

SELECT
    DATE_TRUNC('day', o.order_date) AS date,
    o.region,
    p.category AS product_category,
    COUNT(DISTINCT o.order_id) AS order_count,
    SUM(o.amount) AS gross_revenue,
    SUM(o.amount - p.unit_cost) AS net_revenue,
    AVG(o.amount) AS avg_order_value
FROM {{ ref('orders') }} o
LEFT JOIN {{ ref('products') }} p ON o.product_id = p.product_id
GROUP BY 1, 2, 3

quality.yaml

quality:
  checks:
    - name: no_null_dates
      sql: "SELECT COUNT(*) FROM {{ this }} WHERE date IS NULL"
      severity: critical
      threshold: 0
      description: Date must never be null

    - name: positive_revenue
      sql: "SELECT COUNT(*) FROM {{ this }} WHERE gross_revenue < 0"
      severity: critical
      threshold: 0
      description: Gross revenue must be non-negative

    - name: freshness
      sql: "SELECT DATEDIFF('hour', MAX(date), CURRENT_DATE()) FROM {{ this }}"
      severity: critical
      threshold: 24
      description: Data must be no more than 24 hours stale

    - name: net_lte_gross
      sql: "SELECT COUNT(*) FROM {{ this }} WHERE net_revenue > gross_revenue"
      severity: warning
      threshold: 0
      description: Net revenue should not exceed gross revenue

serving.yaml

serving:
  modes:
    - intent: analytics
      config:
        engine: starrocks
        materialized: true
        refresh_interval: "1h"
        indexes:
          - columns: [date, region]
            type: sort_key

governance.yaml

apiVersion: akili/v1
kind: Governance
classification: confidential
access:
  - team: sales-analytics
    role: viewer
  - team: data-engineering
    role: owner
steward: cfo@acme-corp.com
compliance:
  - sox

compute.yaml

compute:
  engine: dagster
  schedule: "0 6 * * *"
  resources:
    cpu: "1"
    memory: 2Gi
  timeout: 1800
  retries: 2
  tags:
    - daily
    - financial

Globex Inc: Weekly Transaction Summary

product.yaml

product:
  name: weekly_transaction_summary
  namespace: finance
  version: "1.0.0"
  archetype: source
  description: >
    Weekly aggregation of financial transactions.
    Classified as restricted -- contains regulated financial data.
  owner: compliance-team@globex-inc.com
  tags:
    - financial
    - weekly
    - regulated

inputs.yaml

inputs:
  - name: transactions
    kind: source
    connection: globex-finance-db
    resource: public.transactions
    strategy: incremental
    cursor_field: processed_at
    schedule: "0 */2 * * *"
    columns:
      - transaction_id
      - account_id
      - amount
      - currency
      - transaction_type
      - counterparty
      - processed_at
      - status

transform.sql

SELECT
    DATE_TRUNC('week', t.processed_at) AS week_start,
    t.transaction_type,
    t.currency,
    COUNT(*) AS transaction_count,
    SUM(CASE WHEN t.amount > 0 THEN t.amount ELSE 0 END) AS total_credits,
    SUM(CASE WHEN t.amount < 0 THEN ABS(t.amount) ELSE 0 END) AS total_debits,
    SUM(t.amount) AS net_flow,
    COUNT(DISTINCT t.account_id) AS unique_accounts,
    COUNT(DISTINCT t.counterparty) AS unique_counterparties
FROM {{ ref('transactions') }} t
WHERE t.status = 'completed'
GROUP BY 1, 2, 3

quality.yaml

quality:
  checks:
    - name: no_null_weeks
      sql: "SELECT COUNT(*) FROM {{ this }} WHERE week_start IS NULL"
      severity: critical
      threshold: 0
      description: Week start must never be null

    - name: balanced_flow
      sql: >
        SELECT CASE
          WHEN ABS(SUM(net_flow)) > 1000000000
          THEN 1 ELSE 0 END
        FROM {{ this }}
        WHERE week_start >= DATE_TRUNC('week', CURRENT_DATE())
      severity: warning
      threshold: 0
      description: Extreme net flow may indicate data quality issue

    - name: freshness
      sql: "SELECT DATEDIFF('hour', MAX(week_start), CURRENT_DATE()) FROM {{ this }}"
      severity: critical
      threshold: 168
      description: Data must be no more than 1 week stale

    - name: completed_only
      sql: "SELECT 0"
      severity: critical
      threshold: 0
      description: Transform filters to completed status -- should always pass

serving.yaml

serving:
  modes:
    - intent: lookup
      config:
        engine: postgresql
        cache_ttl: "10m"

governance.yaml

apiVersion: akili/v1
kind: Governance
classification: restricted
access:
  - team: compliance
    role: viewer
  - team: finance-engineering
    role: owner
steward: ciso@globex-inc.com
compliance:
  - sox
  - gdpr
  - pci-dss

compute.yaml

compute:
  engine: dagster
  schedule: "0 2 * * 1"
  resources:
    cpu: "2"
    memory: 4Gi
  timeout: 3600
  retries: 3
  tags:
    - weekly
    - regulated
    - high-priority

Tenant Isolation in Action

Storage Isolation

Each tenant’s data lands in its own S3 prefix:

s3://akili-data/tenant-{acme-id}/products/daily_revenue/data/
s3://akili-data/tenant-{globex-id}/products/weekly_transaction_summary/data/

Event Bus Isolation

Each tenant gets its own topic namespace:

acme-corp.daily_revenue.data.available
acme-corp.daily_revenue.execution.events
globex-inc.weekly_transaction_summary.data.available
globex-inc.weekly_transaction_summary.execution.events

Analytics Isolation

Each tenant has a separate StarRocks catalog:

-- Acme queries through their catalog
SET CATALOG tenant_acme_corp_catalog;
SELECT * FROM sales.daily_revenue WHERE date >= '2026-03-01';

-- Globex queries through their catalog
SET CATALOG tenant_globex_inc_catalog;
SELECT * FROM finance.weekly_transaction_summary WHERE week_start >= '2026-03-01';

Governance Isolation

Classification and access policies are tenant-scoped:

Aspect	Acme Corporation	Globex Inc
Classification	`confidential`	`restricted`
Access teams	sales-analytics, data-engineering	compliance, finance-engineering
Compliance	SOX	SOX, GDPR, PCI-DSS
Steward	CFO	CISO

A user authenticated to Acme’s tenant cannot see, query, or reference Globex’s data products, and vice versa. This is enforced at the JWT claims level — the tenant_id in the token determines which tenant’s data is accessible.

Deploy Both Tenants

Acme Corporation (profile: acme):

akili product create --name daily_revenue --namespace sales --profile acme
akili product deploy daily_revenue --version 1.0.0 --profile acme

Globex Inc (profile: globex):

akili product create --name weekly_transaction_summary --namespace finance --profile globex
akili product deploy weekly_transaction_summary --version 1.0.0 --profile globex

Monitoring Per Tenant

Each tenant monitors their own products independently:

# Acme checks their SLA
akili governance sla daily_revenue --profile acme

# Globex checks their quality
akili governance quality weekly_transaction_summary --profile globex

# Neither can see the other's data
akili product list --profile acme    # Only sees Acme's products
akili product list --profile globex  # Only sees Globex's products

Multi-Tenancy — isolation architecture
Multi-Tenant Setup — tenant creation and configuration
Governance — classification and access control
Security Architecture — authentication and authorization

Multi-Tenant SaaS Analytics

Scenario

Tenant Setup

Creating the Tenants

Setting Up Connections (Per Tenant)

Acme Corporation: Daily Revenue Pipeline

product.yaml

inputs.yaml

transform.sql

quality.yaml

serving.yaml

governance.yaml

compute.yaml

Globex Inc: Weekly Transaction Summary

product.yaml

inputs.yaml

transform.sql

quality.yaml

serving.yaml

governance.yaml

compute.yaml

Tenant Isolation in Action

Storage Isolation

Event Bus Isolation

Analytics Isolation

Governance Isolation

Deploy Both Tenants

Monitoring Per Tenant

Related