Open Source Embedded Analytics: Choosing a White-Label BI Platform

If you're building a product, customers eventually expect to see their own data inside it. Charts, dashboards, drill-downs, exports — the whole analytics surface, embedded directly in the application they already use. That's the brief for embedded analytics, and the question most product and data teams arrive at fairly quickly is: do we build it from scratch, license a commercial BI tool, or stand up an open source platform we can embed and white-label?

This guide focuses on the third path: open source embedded analytics platforms that let you put a fully customized, branded BI experience inside your product without paying per-seat for every customer who looks at a chart. We'll cover what "embedded" and "white-label" actually mean in practice, the platform capabilities that matter, and how the leading open source options — Apache Superset, Metabase, Lightdash, and Redash — compare against each other and against proprietary alternatives like Looker, Power BI, and Tableau.

What "embedded analytics" actually means

Embedded analytics is the practice of integrating data visualizations, dashboards, and self-service exploration inside another application — typically a SaaS product — instead of asking users to leave the product and log in to a separate BI tool. There are three broad integration patterns:

• iframe embedding. The fastest way to ship: render a dashboard from your BI platform inside an <iframe> on a customer page, with the BI tool handling auth via a guest token or signed URL. Quick to set up, harder to deeply customize.
• Component / SDK embedding. A JavaScript SDK exposes individual charts, dashboards, or query builders as components your front-end can drop into any layout. More flexible styling, deeper integration with your app's auth and routing, and the option to mix BI components with native UI.
• API-driven custom UI. You build the entire UI yourself and use the BI platform purely as a query/semantic layer over the warehouse, calling its REST or GraphQL API for data. Maximum control, maximum work.

Most teams start with iframes for speed and graduate to SDK or API embedding as their requirements deepen. The platforms worth evaluating support all three.

What "white-label" buys you

White-labeling is the difference between "powered by SomeBITool" and "this is just part of the product." In practice it means:

• Theming and branding — colors, fonts, spacing, logos applied consistently to every embedded surface so it matches your product's design system.
• No vendor branding — no third-party logos, "edit in" buttons, or sign-up prompts leaking into your customer's view.
• Custom domains — embedded dashboards served from analytics.yourproduct.com rather than a third-party domain, which keeps cookies, CSP, and trust signals coherent.
• Hidden tool surface area — your customer never knows (and never needs to know) which BI engine is rendering their chart.

Open source platforms have a structural advantage here: you control the source, so you can theme as deeply as you need without waiting for a vendor to expose a setting. Commercial vendors often gate the most important white-label features (custom domains, removing branding, full theming) behind their highest-priced tiers.

Multi-tenancy and security

If you're embedding analytics for your customers — not just internal users — the platform has to handle multi-tenancy correctly. The non-negotiables:

• Row-level security (RLS). Each customer should only see their own data, enforced at query time, regardless of which dashboard or query they hit.
• Per-tenant authentication. Guest tokens or signed URLs scoped to a single tenant, short-lived, and revocable — never a shared API key.
• Workspace isolation. Logical separation so one tenant's saved queries, datasets, and permissions can't leak into another's.
• Audit logging. Who saw what, when. Required for SOC 2 and most enterprise procurement reviews.

Apache Superset, Metabase, and Lightdash all support row-level security to varying degrees. The depth of multi-tenant isolation, audit logging, and SSO/SCIM integration is where the open source projects diverge from each other — and where managed offerings built on top of them (like Preset for Superset) typically add the operational guarantees enterprise customers expect.

Open source vs. proprietary: the real tradeoff

The pitch for proprietary embedded analytics — Looker, Power BI Embedded, Tableau Embedded — is that someone else operates the infrastructure and the SDK is polished. The cost is real: per-viewer licensing that scales with every end-user of your product, lock-in to a vendor's roadmap, and limits on how deeply you can customize the rendered output.

Open source flips the equation:

• No per-viewer licensing. Pricing scales with your infrastructure, not with how many of your customers happen to open a dashboard this month. For products with thousands of end-users this is the difference between a viable unit economics model and a non-starter.
• Source-level customization. When you need to render a chart type the platform doesn't ship, change query behavior, or hook into your product's auth, you can. With a closed platform you file a feature request.
• Warehouse-native. The leading open source projects connect directly to Snowflake, BigQuery, Redshift, Databricks, Postgres, and the rest of the modern warehouse stack — your data stays where you already store it, no replication or extracts.
• Portability. If you outgrow the project, you take your dashboards, datasets, and queries with you. With a proprietary tool, your work lives inside their format.

The tradeoff: you're now responsible for operating, scaling, and securing the platform — or for finding a managed offering that does it for you.

The open source shortlist

Four projects realistically belong in the conversation for embedded BI today:

Apache Superset

The most feature-complete open source BI platform, originally built at Airbnb and now an Apache Software Foundation top-level project. Strengths for embedded use cases:

• 40+ visualization types, including time-series, geospatial, pivot tables, and Sankey/network diagrams
• A semantic layer (datasets, metrics, virtual datasets) that lets you define business logic once and reuse it across charts
• Embedded SDK with guest-token auth and per-dashboard RLS
• Native connectors for every major warehouse, plus 40+ databases via SQLAlchemy
• An active contributor community and predictable release cadence

The tradeoff is operational: standing up production Superset with multi-tenant guarantees, SSO, audit logging, and the embedded SDK takes real engineering investment. This is exactly the gap that managed Superset offerings — including Preset — fill, with white-label embedding, viewer licensing, RLS, and SOC 2 / HIPAA-eligible deployments out of the box.

Metabase

The friendliest of the open source BI tools, with a no-code question builder that non-technical users learn quickly. Embedded analytics is supported via signed URLs (in the open source edition) and a more capable embedding SDK in the commercial Metabase Pro / Enterprise tier. White-labeling and per-viewer features are mostly behind the paid tiers.

Best fit when ease-of-use for end users is the priority and the embedding requirements are relatively simple.

Lightdash

A newer entrant built natively on top of dbt — your dbt models are your semantic layer. If your team already lives in dbt, Lightdash inherits that work directly. Embedded support exists but is less mature than Superset or Metabase, and the ecosystem of visualization types is narrower.

Best fit for dbt-first analytics teams that want their BI layer to track their dbt models without a translation step.

Redash

The original "SQL editor + dashboards" project, acquired by Databricks in 2020. Redash is excellent for SQL-savvy internal users and ad-hoc reporting. Embedded analytics support is limited compared to the others, and active development has slowed since the acquisition.

Best fit for internal data exploration; not the strongest choice if customer-facing embedded analytics is your primary use case.

Comparison at a glance

Capability	Apache Superset	Metabase	Lightdash	Redash
Visualization types	40+	~25	~15	~20
Semantic layer	Yes (datasets + metrics)	Limited	Yes (via dbt)	No
Embedded SDK	Yes (open source)	Paid tier	Yes	Limited
Row-level security	Yes	Yes (paid)	Yes	Limited
White-label / custom domain	Source-level	Paid tier	Yes	Limited
Warehouse coverage	40+ databases	25+ databases	dbt-supported	35+ databases
AI / natural language	Via managed offerings (e.g., Preset)	Yes (Metabase AI, paid)	Limited	No
Active contributor community	Very active	Active	Growing	Limited

How to choose

A short decision tree, based on what we've seen teams optimize for:

• You need broad visualization coverage and warehouse flexibility, with a path to enterprise-grade embedded analytics. Apache Superset, ideally via a managed offering like Preset if you don't want to operate it yourself.
• You need the simplest possible end-user experience and your embedding needs are basic. Metabase, with the understanding that the white-label and embedding features you'll likely need are in the paid tier.
• Your analytics team lives in dbt and you want zero translation between modeling and BI. Lightdash.
• You need an internal SQL-and-charts tool for technical users. Redash, but watch the development cadence.

For customer-facing embedded analytics — the white-label, multi-tenant, customer-pays-the-bill case — Apache Superset is the most complete open source answer today. The ecosystem around it (warehouses supported, visualization types, embedded SDK, AI features, managed offerings) is the deepest, and it doesn't put white-labeling behind a per-seat paywall the way most commercial alternatives do.

Where Preset fits

Preset is a managed Apache Superset platform built specifically for the kinds of teams this guide is aimed at. We handle the operational pieces — multi-region infrastructure, SOC 2 / HIPAA-eligible deployments, SSO + SCIM, audit logging, the Embedded SDK, viewer licensing — so your team can focus on building the product.

If you're evaluating open source embedded analytics for a customer-facing product, our Embedded Dashboards page walks through the Preset embedding model in detail, and the team is happy to talk through your specific architecture. Whether or not you end up on Preset, going in with a clear picture of what white-label, multi-tenant embedded analytics actually requires is the most useful thing you can do for your product.