How embedded analytics pricing actually
works in 2026.

1. Why this is messy

Every major BI vendor has two price points: one they put on the marketing site, and one that emerges when you actually try to run a multi-tenant SaaS workload. They aren’t the same number.

The marketing price tells you what it costs to authenticate one user. The real price tells you what it costs when the user you signed up is your own service account, embedding dashboards for thousands of unlicensed viewers on the public internet, with RLS, branding, and a reasonable p95 response time.

The difference between those two numbers is where every embedded-analytics budget gets surprised. What follows is the 2026 reality of each vendor, with the math laid out. We sell an alternative (Embedportal, which embeds all four of these), so take the recommendations with a grain of salt — but the underlying numbers are verifiable from each vendor’s public pricing pages.

2. Tableau Embedded

Tableau is the oldest and most widely embedded BI tool. In 2026 it ships two embedding surfaces: the legacy Trusted Authentication path and Connected Apps (JWT direct trust). All new deployments should use Connected Apps.

Sticker price

• Tableau Cloud: ~$70/user/month (Creator), ~$42 (Explorer), ~$15 (Viewer). Embedding is available on any of these.
• Tableau Embedded Analytics: a specific SKU priced per-user-view at roughly $5 per Viewer-equivalent, usage-based.

Real price for multi-tenant SaaS

Tableau Connected Apps don’t require a Viewer licence per embedded end-user. That’s the big win: you can serve ten thousand viewers without buying ten thousand Viewer seats.

What you do need:

• A Creator or Explorer seat for whoever builds the dashboard.
• Tableau Cloud at a plan tier that includes Connected Apps (most paid tiers do; verify).
• Enough data-source compute to survive the view load — this is the hidden line. If viewers query against a live database, that database is the real scaling bottleneck.

For a typical B2B SaaS with ~100 tenants and a handful of dashboards per tenant, Tableau Cloud Creator + Viewer for the team (say 5-10 seats) is the floor — maybe $400-700/month. The embedded workload on top adds compute cost on the underlying warehouse but not much on Tableau itself.

3. Power BI Embedded

Power BI prices embedding in two fundamentally different ways depending on scale: per-user-licence for small deployments, and dedicated capacity (Premium / Fabric) for larger ones.

Sticker price

• Power BI Pro: ~$14/user/month — required for everyone who publishes.
• Power BI Premium Per User (PPU): ~$24/user/month — higher quotas, AI features.
• Power BI Embedded (Azure capacity, formerly “A SKUs”): from ~$750/month for the smallest A1 node up to five figures for larger nodes. Now mostly rolled into Microsoft Fabric capacities.
• Microsoft Fabric F SKU: pay-as-you-go from ~$250/month for F2 up to ~$8,000/month for F64 and beyond. Power BI embed workloads share this capacity.

Real price for multi-tenant SaaS

For production-grade multi-tenant embedding, Microsoft’s guidance is clear: don’t try to run it on Pro licences. You’ll hit rate limits and the user-identity model doesn’t scale past a few tenants. You need a Fabric capacity.

The capacity is shared across your entire Power BI workload — if you already run Power BI internally on a Premium capacity, embedding adds marginal load. If you don’t, the capacity is a new line item.

A realistic floor for a production SaaS using Power BI Embedded in 2026 is $700-1,500/month for the capacity alone, plus a handful of Pro or PPU seats for your developers. Bigger workloads scale into the thousands.

4. Amazon QuickSight

QuickSight has the most usage-aligned pricing model of the four. If your SaaS already runs on AWS, QuickSight is often the cheapest path.

Sticker price

• Authors: $24/user/month (or $18 annual).
• Readers: $5/user/month or $0.30 per 30-minute session, capped at $5. — this is the headline feature for embedding.
• Capacity pricing: bulk session/viewer packs from $250/month for 500 sessions up through enterprise levels.
• Q (natural-language querying, if you need it): add $250/month.

Real price for multi-tenant SaaS

The per-30-minute-session pricing is the killer feature for embedded analytics. If your viewer opens a dashboard for 5 minutes and closes it, you pay 30 cents. If nobody looks, you pay nothing.

For a B2B SaaS with bursty dashboard usage (viewers spike during the workday, quiet at night), this comes out far cheaper than a capacity-based model. A product with 1,000 active viewers averaging 2 dashboard sessions per week lands around $2,400/month for sessions, plus maybe $50/month for Authors.

Trade-off: QuickSight’s author experience is good but not Tableau-class, and you’re locked into AWS. If you’re already on AWS and your analysts are comfortable with QuickSight, it’s the economic winner.

5. Metabase

Metabase is the only one of the four with a genuinely free self-hosted edition. That makes its economics special.

Sticker price

• Metabase Open Source: free, self-hosted. Static embedding is included. You pay for the VM it runs on.
• Metabase Cloud Starter: ~$85/month for up to 5 users, more for bigger plans.
• Metabase Pro: from ~$500/month Cloud (or self-hosted), adds SSO, audit, advanced permissions.
• Metabase Enterprise: custom — adds interactive embedding and premium support.

Real price for multi-tenant SaaS

Static embedding — the flavour Embedportal and most SaaS teams use — is available in every tier including Open Source. That’s unusual in this market, and the reason Metabase is often the right choice for cost-sensitive startups.

For a small SaaS running Metabase Open Source on a $50/month VM and embedding statically, the embedded analytics line item is close to zero. You pay for a bigger VM as load grows.

For larger teams that want SSO, audit, and vendor support, Metabase Pro at ~$500/month is still competitive with Power BI Embedded and cheaper than a dedicated Tableau deployment.

Trade-off: Metabase’s expressiveness for analysts doesn’t match Tableau. For rich dashboarding power users, it’s a step down. For 80% of SaaS use cases (KPI cards, time-series, breakdowns, usage funnels), it’s plenty.

6. Side-by-side

7. Hidden costs

Three line items every embedded analytics project underestimates:

Engineering time to multi-tenant

None of these vendors ship a multi-tenant embedding layer. You build it. Token signing, RLS attribute forwarding, per-tenant branding, audit logs, the frontend shell. Budget 2-6 engineer-months for a production-grade version, depending on vendor and ambition. This is the single biggest line item for most teams, and it’s invisible until the project is half-done.

Underlying data warehouse load

Dashboards with live connections hit your warehouse on every viewer interaction. An embedded workload with 1,000 active viewers can easily 10× your BigQuery / Snowflake / Redshift bill. Caching at the BI layer helps, but you need to model for it.

Ongoing operations

Someone on your team will own the relationship with whichever BI vendor you pick — version upgrades, capacity tuning, access reviews for the service principal or Connected App, key rotation, SOC 2 evidence collection when the auditor asks. Budget 10-20% of one engineer ongoing.

8. A decision framework

• Zero budget, fast prototype — Metabase OSS. Self-host on a small VM, use static embedding. You can validate the whole embedded analytics idea with your customers before spending a dollar.
• You live on AWS — QuickSight. Session-based pricing wins on bursty workloads, RLS is clean via session tags, IAM role lets you keep data in your account.
• You live on Azure, or use Excel as a primary analytical surface — Power BI Embedded. The Fabric capacity model is expensive but predictable, and the integration with the Microsoft stack is hard to beat.
• Analyst experience matters most — Tableau. Your analysts will thank you. Pay the licence, use Connected Apps, and layer multi-tenancy on top.
• You want optionality across all of the above — Embedportal. One integration layer, every vendor. Swap later without a frontend rewrite.

Whatever you pick, the cost of the embedded analytics layer itself is usually dwarfed by the engineering time to make it multi-tenant and the warehouse compute to serve queries. Budget for all three, not just the sticker price.