Enterprise BI with Streamlit in Snowflake: Use Cases, Architecture & Best Practices
Key takeaways:
- Traditional BI dashboards often stop at reporting, leaving business users without the ability to investigate trends, collaborate, or take action. Streamlit in Snowflake extends analytics beyond visualization by enabling interactive business applications that support operational decision-making.
- Enterprise data is frequently fragmented across CRM, ERP, lending, and operational systems, resulting in inconsistent metrics and delayed insights. By combining Snowflake, dbt, and Streamlit, organizations can build governed analytics applications on trusted data models with a single source of truth.
- Business teams need role-specific insights rather than one-size-fits-all dashboards. Interactive applications tailored for executives, operations, finance, risk, marketing, and customer service deliver faster decisions while maintaining centralized governance.
- Scaling analytics should not require additional infrastructure or complex application development. Streamlit in Snowflake allows organizations to build secure, scalable analytics experiences directly within the Snowflake ecosystem, reducing architectural complexity and accelerating time to value.
- Modern BI initiatives succeed when technology aligns with business outcomes. Sigma Infosolutions helps organizations design, engineer, and operationalize Snowflake-native analytics solutions—from data engineering and modern BI architecture to interactive analytics applications that drive measurable business impact.
Introduction
Business leaders today have more data than ever before, yet many still struggle to turn that data into timely, actionable decisions. Traditional BI platforms excel at delivering dashboards and reports, but they often fall short when teams need to investigate trends, collaborate across functions, or interact with data to support operational workflows. As organizations modernize their data platforms, the focus is shifting from static reporting to interactive analytics experiences that empower business users to make faster, better-informed decisions.
For enterprises using Snowflake as their cloud data platform, Streamlit in Snowflake (SiS) bridges this gap by enabling organizations to build secure, interactive analytics applications directly within their existing data environment. Instead of moving data between multiple tools or developing custom web applications from scratch, businesses can create governed experiences for portfolio monitoring, executive reporting, operational dashboards, customer insights, financial analysis, and industry-specific workflows—all while keeping data protected within Snowflake.
This approach transforms Snowflake from a data warehouse into a business decision platform, allowing organizations to combine trusted data, governed access, and interactive analytics in a single ecosystem. Whether the objective is improving operational visibility, empowering self-service analytics, or accelerating decision-making across departments, Streamlit extends the value of modern BI investments beyond visualization.
Organizations investing in modern data platforms increasingly need more than dashboards. They need interactive analytics applications that allow business users to explore data, monitor operations, and act on insights without moving data outside Snowflake. This guide explains how Streamlit in Snowflake enables organizations to build secure business applications for analytics, operational decision-making, and governed self-service reporting.
Build Smarter BI & Analytics Solutions
Transform your data into interactive, decision-ready analytics with Sigma Infosolutions’ BI & Analytics Development Services.
1. What is Streamlit in Snowflake?
Streamlit in Snowflake (SiS) is a native Snowflake capability that enables organizations to create interactive analytics applications directly within Snowflake, combining governed data access with business-friendly interfaces for reporting, operational workflows, and decision support.
| Aspect | Detail |
| What it is | Interactive web apps (filters, charts, forms, search) built with the Streamlit Python library |
| Where it runs | Inside Snowflake — no separate app server or hosting |
| How it connects to data | Active Snowflake session — no external connection strings |
| Access control | Snowflake RBAC on the STREAMLIT object |
| Code storage | App files on a Snowflake stage, referenced by a STREAMLIT object |
Why SiS for lending + HubSpot
- Blend origination, servicing, and CRM in one interactive UI
- Keep PII inside Snowflake with masking and row policies
- Build ops tools (queues, lookups, light write-back) that standard BI struggles with
- Accelerate delivery of business analytics applications without additional infrastructure
- Keep governed business data securely inside Snowflake
- Enable operations teams to move beyond static dashboards
- Reduce dependency on custom web application development
2. Turning Your Snowflake Data Platform into Business Applications
┌──────────────┐ ┌──────────────┐ ┌──────────────────────┐
│ Airflow │────▶│ dbt │────▶│ Snowflake Marts │
│ (orchestrate)│ │ (transform) │ │ (trusted datasets) │
└──────────────┘ └──────────────┘ └──────────┬───────────┘
│
▼
┌──────────────────────┐
│ Streamlit in Snowflake│
│ (apps & workflows) │
└──────────────────────┘
| Layer | Responsibility | SiS role |
| Airflow | Ingest HubSpot, LMS, origination data; schedule dbt | Feeds data; optional write-back processing |
| dbt | Business logic, tests, documentation, marts | Primary data source for all SiS apps |
| SiS | User-facing apps | Read marts; optional write to staging tables |
Principle: Keep transformations and metric definitions in dbt. Use SiS for interaction, visualization, and lightweight workflows.
Read the blog: Data Warehousing and BI Development: Building a Scalable Analytics Foundation
3. Runtime options
Snowflake offers two SiS runtime environments. Choose based on audience size, package needs, and usage pattern.
3.1 Warehouse Runtime
Best suited for executive dashboards, departmental analytics, and proof-of-concept business applications.
| Characteristic | Detail |
| Compute | Virtual warehouse |
| Instance model | One personal instance per viewer |
| Best for | POCs, small teams, simple apps |
| Packages | Primarily Snowflake Anaconda channel |
| Consideration | Warehouse suspend = cold start; cost scales with concurrent users |
3.2 Container runtime (GA 2026)
Recommended for enterprise-wide analytics applications, AI-powered assistants, and operational workflows supporting larger user communities.
| Characteristic | Detail |
| Compute | Compute pool (SPCS) + warehouse for SQL |
| Instance model | Shared long-running instance for all viewers |
| Best for | Production apps, many daily users, AI/chat apps |
| Packages | Broader PyPI support, newer Streamlit versions |
| Extras | Secrets via st.secrets, GPU option, better caching across sessions |
3.3 Compute separation (recommended)
| Resource | Purpose |
| WH_APP (small) | Run Streamlit app code |
| WH_QUERY (medium) | Execute SQL against loan/HubSpot marts |
Set via QUERY_WAREHOUSE when creating or altering the Streamlit object.
Read the blog: Predictive Analytics for Customer Retention: BI Strategies That Work in 2026
4. Domain data model (prerequisite marts)
SiS apps should query curated dbt marts, not raw landing tables.
| Mart | Description |
| dim_borrower | Unified borrower; identity keys |
| dim_loan | Loan master (product, term, status, dates) |
| fct_origination | Application → underwriting → approval → disbursement |
| fct_loan_snapshot_daily | Daily book position, outstanding balance |
| fct_payment | Payment transactions |
| fct_delinquency | DPD buckets, roll history |
| fct_hubspot_deal | CRM deals and stages |
| dim_campaign / dim_channel | Marketing attribution |
| bridge_contact_loan | HubSpot contact ↔ borrower ↔ loan |
| fct_borrower_timeline | Unified event timeline |
| fct_vintage_performance | Origination cohort vs subsequent performance |
| fct_fraud_alert | Fraud flags and reason codes |
HubSpot + lending integration patterns (dbt)
- Identity resolution — Link hubspot_contact_id to borrower_id via email, phone, hashed national ID
- Attribution — Tie funded loans to first/last-touch campaign, not just HubSpot “closed won”
- Funnel truth — Define “funded” as disbursement event, not CRM stage change
- Vintage spine — Every loan carries origination_month, channel, campaign, credit_grade
Read our success story: Modernizing a Lending Data Platform with Snowflake Native Capabilities, Governance, and AI Readiness
5. Use cases — Tier 1 (highest value)
5.1 Improve Loan Conversion Visibility Across the Origination Funnel
| Item | Detail |
| Users | Growth, product, lending operations |
| Data | HubSpot leads/deals, applications, approvals, disbursements |
| Features | Lead → application → approved → funded funnel; drop-off by stage; time-to-fund SLA; cohort by campaign/UTM/partner |
| Key question | “Which channels actually produce funded loans — and where do we lose applicants?” |
| dbt models | fct_origination_funnel, dim_marketing_channel, fct_hubspot_attribution |
5.2 Monitor Portfolio Risk and Delinquency in Real Time
| Item | Detail |
| Users | Collections, risk, loan management leadership |
| Data | Loan balances, DPD, payments, roll rates |
| Features | PAR 30/60/90; roll and cure rates; early warning (bucket movement last 7 days); segment by product, region, acquisition channel; exportable work lists |
| Key question | “Where is credit quality deteriorating — and which vintages or channels drive it?” |
| dbt models | fct_loan_snapshot_daily, fct_delinquency, fct_roll_rate |
5.3 Single borrower 360° view
| Item | Detail |
| Users | Customer service, collections, underwriting QA |
| Data | Borrower, all loans, payments, HubSpot activity, communications |
| Features | Search by phone / loan ID / HubSpot contact ID; unified timeline; current exposure and payment status |
| Key question | “What is the full story on this customer before I call or decide?” |
| dbt models | dim_borrower, fct_borrower_timeline, bridge_contact_loan |
| Governance | Dynamic masking on PII; role-based field visibility; audit access |
5.4 Underwriting & credit policy monitor
| Item | Detail |
| Users | Risk, credit policy, origination leadership |
| Data | Application attributes, bureau/decision outputs, approvals/declines, post-fund performance |
| Features | Approval rate by score/income/product; vintage performance (e.g. 90+ DPD at 6 months); policy drift alerts; decline reason breakdown |
| Key question | “Is our credit policy performing — and is approval behavior drifting?” |
| dbt models | fct_underwriting_decision, fct_vintage_performance, dim_credit_policy_version |
6. Use cases — Tier 2 (strong fits)
6.1 Measure Marketing ROI from Lead to Funded Loan
| Item | Detail |
| Users | Marketing, finance, growth |
| Data | HubSpot campaign cost, attributed originations, collections/LTV |
| Features | CAC by channel; funded principal vs HubSpot deal value; payback period by cohort |
| Key question | “Which campaigns acquire profitable loans — not just leads?” |
| dbt models | fct_marketing_spend, fct_loan_ltv_cohort, fct_hubspot_loan_match |
6.2 Partner / broker performance portal
| Item | Detail |
| Users | Partner managers, broker relationship teams |
| Data | Originations by partner, quality metrics, HubSpot partner deals |
| Features | Volume, approval rate, ticket size; early delinquency scorecard; commission preview (read-only) |
| Key question | “Which partners deliver volume and quality?” |
| Governance | Row access policy: partner_id = current_partner() |
6.3 Collections workbench (light workflow)
| Item | Detail |
| Users | Collections agents and team leads |
| Data | Delinquent loans, contact history, HubSpot tasks/notes |
| Features | Prioritized queue; disposition capture; promise-to-pay logging; agent notes |
| Write-back | App writes to stg_collection_action → Airflow/dbt processes into warehouse |
| Runtime | Container runtime recommended for all-day agent usage |
| dbt models | fct_delinquency, fct_collection_queue, fct_hubspot_engagement |
6.4 Fraud & anomaly review queue
| Item | Detail |
| Users | Fraud operations |
| Data | Application signals, velocity rules, identity graph, HubSpot duplicates |
| Features | Flagged applications with reason codes; linked identities; approve / escalate / false-positive actions |
| Write-back | Disposition to stg_fraud_review |
| dbt models | fct_fraud_alert, dim_identity_graph |
6.5 Regulatory & management reporting pack
| Item | Detail |
| Users | Compliance, executive leadership |
| Data | Book size, originations, PAR, write-offs, restructures |
| Features | Standard regulatory views; as-of date selector; CSV/PDF export; GL reconciliation check |
| Note | Less interactive than other apps; useful when BI access is limited |
7. Use cases — Tier 3 (advanced)
7.1 Cortex-powered lending assistant
| Item | Detail |
| Users | Ops analysts, team leads |
| Examples | “Which campaigns had highest 90-day DPD on personal loans?”; summarize borrower history before a call; classify HubSpot tickets |
| Runtime | Container runtime (streaming responses, richer libraries) |
| Data | Curated marts + optional text fields (notes, tickets) |
7.2 Early warning / pre-delinquency targeting
| Item | Detail |
| Users | Retention, collections strategy |
| Data | Payment behavior, HubSpot engagement drop, support tickets |
| Features | Not-yet-delinquent risk list; contact priority; campaign A/B results |
| dbt models | fct_behavioral_risk_score, fct_hubspot_engagement |
7.3 Product & pricing experiment dashboard
| Item | Detail |
| Users | Product, risk, finance |
| Data | Loan terms at origination vs performance |
| Features | Yield/IRR by variant; term/rate/fee experiments vs default; HubSpot product interest vs actual product taken |
8. Hero app concept — Lending Growth & Risk Control Tower
A single multi-tab app that serves marketing, risk, and operations.
| Tab | Audience | Primary question |
| Funnel | Growth, ops | Are HubSpot leads converting to funded loans? |
| Book | Collections, risk | What is PAR and roll rate by channel and product? |
| Vintage | Risk, finance | Are recent campaigns producing bad loan vintages? |
| Borrower | CS, collections | What is this customer’s full history? |
Recommended as the flagship POC — demonstrates cross-domain value in one deployment.
9. SiS vs BI tools
Streamlit in Snowflake complements—not replaces—traditional BI platforms. While Power BI, Tableau, and Looker remain ideal for standardized reporting, Streamlit enables organizations to build interactive business applications where users need to explore data, execute workflows, or capture operational inputs.
| Dimension | Power BI / Tableau / Looker | Streamlit in Snowflake |
| Primary audience | Broad business users | Internal power users, ops, specialists |
| Customization | Semantic model + templates | Full Python logic |
| Workflows | Limited write-back | Forms, queues, multi-step actions |
| Time to build niche tools | Slower | Faster for bespoke use cases |
| Best for | Standard enterprise reporting | Lookup tools, workbenches, control towers |
Recommendation: Use BI for board-level and standard reporting; use SiS for operational tools and cross-domain exploration.
10. Governance and security
10.1 Must-haves for lending data
| Control | Application |
| Dynamic masking | National ID, bank account, address, phone, email |
| Row access policies | Partner-scoped views; regional restrictions; agency-assigned collections |
| Secure views | Apps query views, not base tables |
| Role design | Separate app owner role vs end-user roles |
| Audit | Access history + query history; log sensitive lookups in Borrower 360 |
| No raw API calls in app | HubSpot data only via curated Snowflake marts |
10.2 Owner’s rights model
SiS apps run with the owner role’s privileges. Design so:
- Owner role has minimum required access
- End users are constrained by secure views, RLS, and masking
- PII exposure is policy-driven, not hard-coded in Python
11. Write-back pattern (when apps need to capture data)
For collections dispositions, fraud review, or manual overrides:
User action in SiS app
↓
INSERT into stg_* table (staging)
↓
Airflow job validates & promotes
↓
dbt model incorporates into curated layer
Rules:
- Never write directly to production fact tables from SiS
- Staging tables have clear ownership and validation rules
- Airflow provides audit trail and error handling
12. Implementation Considerations
12.1 Creation methods
| Method | Best for |
| Snowsight UI | Rapid POC; side-by-side editor and preview |
| SQL (`CREATE STREAMLIT`) | Infrastructure-as-code; repeatable deployments |
| Snowflake CLI | CI/CD from Git repository |
12.2 Example SQL (warehouse runtime)
CREATE STREAMLIT lending_control_tower
ROOT_LOCATION = ‘@analytics.apps.app_stage/lending_control_tower’
MAIN_FILE = ‘streamlit_app.py’
QUERY_WAREHOUSE = ‘WH_ANALYTICS’;
GRANT USAGE ON STREAMLIT lending_control_tower TO ROLE lending_analyst;
12.3 Example SQL (container runtime)
ALTER STREAMLIT lending_control_tower
COMPUTE_POOL = ‘STREAMLIT_POOL’
QUERY_WAREHOUSE = ‘WH_ANALYTICS’
RUNTIME_NAME = ‘SYSTEM$ST_CONTAINER_RUNTIME_PY3_11’;
12.4 Typical app code pattern
import streamlit as st
from snowflake.snowpark.context import get_active_session
session = get_active_session()
st.title(“Portfolio Health”)
product = st.selectbox(“Product”, [“All”, “Personal”, “SME”, “Mortgage”])
df = session.sql(“””
SELECT as_of_date, dpd_bucket, COUNT(*) AS loan_count
FROM analytics.marts.fct_delinquency
WHERE (:product = ‘All’ OR product = :product)
GROUP BY 1, 2
ORDER BY 1, 2
“””, params={“product”: product}).to_pandas()
st.bar_chart(df.pivot(index=”as_of_date”, columns=”dpd_bucket”, values=”loan_count”))
13. Business Value Roadmap
| Phase | Business Priority | Analytics Capability | Primary Business Users | Expected Business Outcome |
| Phase 1 | Improve Origination Visibility | Origination Funnel & Conversion Analytics | Growth, Lending Operations | Identify conversion bottlenecks, improve funding rates, and optimize acquisition channels. |
| Phase 2 | Strengthen Portfolio Monitoring | Portfolio Health & Delinquency Analytics | Risk, Collections, Executive Leadership | Gain real-time visibility into portfolio performance and detect emerging credit risks earlier. |
| Phase 3 | Create a Unified Customer View | Borrower 360° Analytics | Customer Service, Collections, Underwriting | Provide teams with a complete borrower history to improve servicing and decision-making. |
| Phase 4 | Optimize Marketing Performance | Marketing ROI & Customer Acquisition Analytics | Marketing, Finance | Measure campaign effectiveness, CAC, LTV, and channel profitability using trusted business metrics. |
| Phase 5 | Enhance Credit Strategy | Credit Policy & Portfolio Performance Analytics | Credit Risk, Underwriting Leadership | Monitor policy effectiveness, approval trends, and portfolio quality to refine lending strategies. |
| Phase 6 | Enable Operational Decisioning | Collections Operations Workbench | Collections Teams & Managers | Streamline collections workflows, improve agent productivity, and accelerate resolution through guided operational workflows. |
Organizations can deploy these capabilities incrementally based on business priorities. Sigma Infosolutions helps enterprises define the roadmap, build governed Snowflake-native analytics applications, and scale them across business functions while ensuring security, performance, and long-term maintainability.
Success criteria per phase
| Phase | Success metric |
| 1 | Daily active users in growth and collections teams |
| 2 | Reduction in ad hoc SQL requests to data team |
| 3 | Measurable time saved in collections disposition logging |
14. What not to build in SiS
| Capability | Better platform | Reason |
| Core loan servicing (payments, statements) | LMS / core banking | System of record |
| Heavy ETL or metric logic | dbt | Tested, versioned transformations |
| Sub-second real-time dashboards | Streaming stack + specialized UI | Streamlit reruns on interaction |
| Public internet SaaS | Dedicated web platform | SiS is for authenticated Snowflake users |
| Regulatory submission engine | Controlled batch pipeline | Audit and immutability requirements |
15. Cost considerations
| Factor | Guidance |
| Warehouse runtime | Cost per user session; many concurrent users = many instances |
| Container runtime | Steady compute pool cost; often cheaper at scale for popular apps |
| Query cost | Heavy scans bill the query warehouse — pre-aggregate in dbt/materalized views |
| Caching | Use @st.cache_data for expensive queries (especially container runtime) |
| Optimization | Point apps at aggregated marts, not raw event tables |
16. Architecture diagram
Architecture diagram (conceptual flow)
GROWTH –> FUNNEL
GROWTH –> CAC
RISK –> PAR
RISK –> TOWER
COLL –> PAR
COLL –> WORK
COLL –> B360
CS –> B360
FUNNEL –> FACT
PAR –> FACT
B360 –> DIM
B360 –> CRM
CAC –> CRM
WORK –> FACT
TOWER –> FACT
TOWER –> CRM
DIM –> MASK
FACT –> RLS
CRM –> MASK
sis –> AUDIT
Why Sigma Infosolutions for BI & Analytics Solutions
Modern BI initiatives require more than implementing dashboards—they demand a scalable data foundation, trusted business metrics, intuitive analytics experiences, and governance that supports enterprise-wide decision-making. Sigma Infosolutions partners with organizations to build end-to-end BI and analytics ecosystems that transform raw data into actionable business intelligence.
Whether you’re modernizing legacy reporting, building a Snowflake-native analytics platform, or enabling self-service analytics across business functions, our teams combine deep expertise in cloud data platforms, data engineering, business intelligence, and custom analytics application development to accelerate time to value.
Our BI & Analytics capabilities include:
- Modern Data Platform Engineering – Design and implement scalable data platforms using Snowflake, Azure, AWS, and modern cloud architectures.
- Data Integration & Engineering – Build reliable ETL/ELT pipelines, data lakes, and governed data warehouses using technologies such as dbt, Airflow, and cloud-native services.
- Business Intelligence & Executive Dashboards – Deliver interactive reporting and KPI dashboards with Power BI, Tableau, Looker, and Snowflake-native analytics.
- Custom Analytics Applications – Develop interactive business applications with Streamlit in Snowflake for operational analytics, executive decision support, and domain-specific workflows.
- AI-Driven Analytics – Enhance business intelligence with predictive analytics, anomaly detection, natural language querying, and AI-powered decision support.
- Data Governance & Security – Establish robust governance frameworks, role-based access controls, data quality processes, and compliance aligned with enterprise requirements.
- Analytics Modernization – Migrate legacy reporting environments to modern cloud-based analytics platforms while improving scalability, performance, and cost efficiency.
Business Outcomes We Help Deliver
Organizations partner with Sigma Infosolutions to:
- Build a trusted, enterprise-wide single source of truth.
- Reduce reporting cycles through automated, governed analytics.
- Empower business users with self-service insights and interactive analytics.
- Improve operational visibility across sales, finance, operations, and customer success.
- Scale analytics initiatives without increasing platform complexity.
- Accelerate data-driven decision-making with secure, cloud-native BI solutions.
This business-first approach enables organizations to move beyond static reporting and unlock the full value of their data investments through modern, scalable, and insight-driven analytics solutions.
Conclusion:
Streamlit in Snowflake enables organizations to transform trusted data into interactive business applications without introducing additional platforms or duplicating data. By combining governed analytics, operational workflows, and secure collaboration within Snowflake, enterprises can move beyond static reporting and empower teams to make faster, data-driven decisions while preserving security and governance.
Ready to Modernize Your Analytics?
Partner with Sigma Infosolutions to build secure, scalable, Snowflake-native BI solutions.
Frequently Asked Questions
1. What is Streamlit in Snowflake (SiS)?
Streamlit in Snowflake (SiS) is a native capability that enables organizations to build and run interactive analytics applications directly within Snowflake. It allows business users to explore data, monitor KPIs, and support operational workflows while keeping data secure and governed inside the Snowflake environment.
2. How is Streamlit in Snowflake different from traditional BI tools?
Traditional BI platforms like Power BI, Tableau, and Looker are designed primarily for dashboards and standardized reporting. Streamlit in Snowflake complements these tools by enabling interactive analytics applications, operational workbenches, search interfaces, and lightweight workflows that require user interaction beyond visualization.
3. What business use cases are best suited for Streamlit in Snowflake?
Organizations commonly use Streamlit in Snowflake for origination funnel analytics, portfolio health monitoring, borrower 360° views, marketing ROI analysis, credit policy monitoring, collections workbenches, fraud review, and AI-powered analytics assistants. These applications help teams make faster, data-driven decisions across business functions.
4. Is Streamlit in Snowflake suitable for enterprise BI and analytics?
Yes. Streamlit in Snowflake supports enterprise analytics by combining Snowflake’s security, governance, and scalability with interactive business applications. It enables organizations to deliver role-based analytics experiences without moving data outside the Snowflake platform.
5. Can Streamlit in Snowflake integrate with existing business systems?
Yes. Streamlit applications can leverage curated data from CRM, ERP, lending platforms, marketing systems, and other enterprise applications through Snowflake and modern data pipelines. This enables organizations to create unified analytics experiences using trusted business data.
6. How does Streamlit in Snowflake ensure data security and governance?
Applications inherit Snowflake’s security framework, including role-based access control, dynamic data masking, row-level security, secure views, and auditing capabilities. This allows organizations to build interactive analytics while maintaining enterprise governance and regulatory compliance.
7. When should organizations choose Streamlit in Snowflake over custom application development?
Streamlit in Snowflake is an excellent choice when organizations want to rapidly build secure analytics applications without managing separate infrastructure or moving governed data outside Snowflake. It significantly reduces development effort while accelerating time to value for business analytics initiatives.
8. How can Sigma Infosolutions help organizations implement BI and analytics solutions?
Sigma Infosolutions helps organizations design, build, and modernize end-to-end BI and analytics ecosystems—from data engineering and Snowflake implementation to interactive analytics applications, governance frameworks, AI-powered insights, and enterprise dashboard development. The focus is on delivering scalable, secure, and business-centric analytics solutions that drive measurable outcomes.
