The pricing model matters more than the price #
A SaaS team picks an embedded analytics vendor at 500 users. Cost is fine. They hit PMF, scale to 10,000 users, and suddenly their analytics bill rivals AWS. The vendor's base price didn't change — the pricing model made growth expensive.
Most vendor pricing pages say "Contact Sales." The ones that publish numbers use units you can't compare: one charges per monthly active user, another per dashboard session, another per editor seat plus a viewer surcharge.
We built Gigacatalyst on a per-tenant model because we watched this destroy budgets at Series B companies. Their success — more customers, more engagement — was punishing them financially. Per-tenant isn't universally right, but it solved a real pattern.
This guide covers every pricing model, maps real costs where available, and helps you pick the structure that fits your growth — not just today's numbers.
How each model works #
Per-Tenant (Gigacatalyst) #
Flat fee per SaaS customer (tenant). End-users within each tenant are unlimited. 50 customers using dashboards = you pay for 50 tenants, whether each has 10 users or 1,000.
Predictability: High. Cost scales with customer count, which correlates with your revenue.
Risk: Few tenants with massive user bases = great value. Thousands of tenants with one user each = potentially more expensive than per-MAU.
Per-MAU (Luzmo) #
You pay based on monthly active users — unique individuals who load a dashboard in a given month. Luzmo starts at EUR 995/month (Scale) and goes to EUR 2,495+/month (Enterprise), with MAU limits at each tier.
Predictability: Medium. Stable if engagement is steady. Spikes when you onboard a large customer and their 500 users all activate in month one.
Risk: Success punishes you. A great dashboard experience that drives daily usage doesn't change your bill — but a single large onboarding can jump you a tier overnight.
Per-Session (Embeddable) #
Each time a user loads a dashboard counts as a session. Embeddable uses this model but doesn't publish rates — you need a sales call.
Predictability: Low. A power user checking dashboards 20 times a day costs 20x more than a monthly viewer.
Risk: Hard to budget. The better your dashboard experience, the more it costs.
Per-Seat + Per-Viewer (Metabase) #
Metabase Cloud Pro charges $85/user/month for editors. Self-hosted is free (open source), but you lose cloud management, SSO, and governance.
Predictability: High for editors. Variable for viewers if you embed for customers.
Risk: Free self-hosted is genuinely free if you have the team to maintain it. But customer-facing embedding requires their Enterprise plan — custom pricing.
Enterprise Contract (GoodData, Sisense/ThoughtSpot) #
Annual contracts negotiated with sales. Starting points: $50,000 to $200,000+/year. GoodData offers a growth tier around $30,000/year; Sisense skews higher.
Predictability: High short-term (fixed annual cost), unpredictable at renewal. Vendors have leverage once you're integrated.
Risk: Lock-in. Twelve months in, switching costs are enormous. Renewals often come with 15-30% increases.
Per-Query (Newer tools) #
Charges based on queries executed. A dashboard with 12 widgets fires 12 queries per load.
Predictability: Low. Complex dashboards get expensive fast. Users avoid refreshing, which defeats the purpose.
Open Source (Metabase CE, Apache Superset) #
Free to self-host. You pay for infrastructure and engineering time to deploy, maintain, and customize.
Predictability: Infra costs are steady. Engineering time is not. Security patches, upgrades, and embedding work add up.
Risk: "$0 licensing" can easily become $100k+/year in engineering salaries. No SLA, no vendor support during outages.
Pricing comparison table #
| Platform | Model | Starting Price | Embedded Use | End-User Cost | Published Pricing? |
|---|---|---|---|---|---|
| Gigacatalyst | Per-tenant | Custom (based on tenant count) | Built for embedding | Unlimited per tenant | Transparent on call |
| Luzmo | Per-MAU | EUR 995/mo (Scale) | Yes | Charged per MAU | Yes |
| Embeddable | Per-session | Custom | Yes (core use case) | Per session | No |
| Metabase | Per-seat / Self-hosted free | $85/user/mo (Pro Cloud) | Enterprise plan required | Viewer tier or unlimited (self-hosted) | Partially |
| GoodData | Enterprise contract | ~$30k/year (Growth) | Yes | Included in contract | No |
| Sisense (ThoughtSpot) | Enterprise contract | ~$50k-200k+/year | Yes | Included in contract | No |
| Apache Superset | Free (self-hosted) | $0 licensing | DIY embedding | N/A | Yes (it's free) |
Which model fits your business #
By tenant count vs. end-users #
50 tenants, 200 users each (10,000 total end-users):
- Per-tenant (Gigacatalyst): Cost scales with 50 tenants. Predictable.
- Per-MAU (Luzmo): 5,000 active MAUs puts you at EUR 2,495+/month.
- Per-session: 75,000 sessions/month at unknown rates. Hard to forecast.
- Per-seat (Metabase): 10,000 end-users requires Enterprise pricing. Contact sales.
500 tenants, 5 users each (2,500 total end-users):
- Per-tenant: Higher (500 tenants), but predictable.
- Per-MAU: More affordable if only 1,000 are active monthly.
- Per-seat: Still requires Enterprise for embedding.
The pattern: Per-tenant favors fewer, larger customers. Per-MAU favors many small customers where most users are inactive.
By growth stage #
Early (fewer than 20 customers): Self-hosted Metabase or Superset = zero licensing cost. Manageable engineering investment when small. Gigacatalyst's per-tenant model also works — low tenant count means low cost.
Growth (20-200 customers, scaling fast): Per-MAU becomes unpredictable as onboarding accelerates. Per-tenant stays linear. Enterprise contracts work if you lock in terms before usage spikes.
Scale (200+ customers): Enterprise contracts offer bulk discounts but require annual minimums. Per-tenant remains linear. Per-MAU at this scale triggers enterprise pricing regardless.
By usage patterns #
Heavy daily usage: Avoid per-session. Per-MAU is fine if user count is capped. Per-tenant ignores usage intensity entirely.
Occasional reporting: Per-session and per-MAU are both cheap when usage is light. Per-tenant may be overkill.
Unpredictable bursts: Per-tenant and enterprise contracts handle spikes without surprises. Per-MAU and per-session will spike with usage.
Scenario calculations #
Scenario A: Mid-Market B2B SaaS (50 tenants, 200 users each) #
| Model | Estimated Monthly Cost | Notes |
|---|---|---|
| Per-tenant (Gigacatalyst) | Predictable, scales with 50 tenants | No per-user charges |
| Per-MAU (Luzmo) | EUR 2,495+/mo | Assuming 5,000+ MAUs |
| Enterprise (GoodData) | ~$4,000-8,000/mo | Annualized from $50-100k contract |
| Self-hosted (Metabase CE) | $0 licensing + ~$3-5k/mo engineering | DevOps time, infrastructure |
Scenario B: SMB-Focused SaaS (500 tenants, 5 users each) #
| Model | Estimated Monthly Cost | Notes |
|---|---|---|
| Per-tenant (Gigacatalyst) | Higher (500 tenants) | Volume discounts likely available |
| Per-MAU (Luzmo) | EUR 995-2,495/mo | If only 500-2,000 are active |
| Enterprise (Sisense) | ~$5,000-15,000/mo | Annualized from larger contract |
| Self-hosted (Superset) | $0 licensing + $5-8k/mo engineering | More tenants = more isolation work |
Scenario C: Enterprise SaaS (10 tenants, 5,000 users each) #
| Model | Estimated Monthly Cost | Notes |
|---|---|---|
| Per-tenant (Gigacatalyst) | Low (only 10 tenants) | Best value scenario |
| Per-MAU (Luzmo) | Enterprise custom pricing | 25,000+ MAUs likely |
| Enterprise (GoodData) | $8,000-16,000/mo | Large user base triggers higher tier |
| Per-seat (Metabase Enterprise) | Custom | Viewer licensing at scale |
Hidden costs nobody mentions #
Beyond the sticker price:
Integration engineering. Every platform takes 2-8 weeks of engineering to embed properly. Security, SSO, theming — none of this is plug-and-play.
Ongoing maintenance. SDK upgrades, API changes, browser issues. Budget 10-20% of initial integration cost annually.
Support burden. When embedded analytics breaks, customers file tickets with you — not your vendor. You're first-line support.
Data pipeline. Most tools require you to pipe data into their system. That's separate infrastructure with its own cost.
With Gigacatalyst, most of this shrinks. Our AI builder connects directly to your existing APIs — no separate pipeline. Customers build their own reports, reducing support load. Apps match your design system automatically.
Building in-house #
The build vs. buy decision is tempting when vendor pricing feels high. Quick math:
- 2 senior engineers for 6 months: ~$300,000 in salary
- Ongoing maintenance: 1 engineer full-time (~$150,000/year)
- Infrastructure: $2,000-10,000/month
- Year-one total: ~$500,000+
Building makes sense only if analytics is your core product. If you're a CRM or ERP adding analytics as a feature, the math favors buying — even at enterprise pricing.
Frequently asked questions #
How do I know which model is cheapest for my case?
Map your tenant count, active user count, and average sessions per user. Multiply by 12 months, add 50% for growth. The model where that growth changes your bill the least is the right fit. For B2B SaaS with distinct customer accounts, per-tenant almost always wins.
Can I switch models later?
Technically yes. Practically, embedded analytics becomes deeply integrated within 3-6 months. Migration means rebuilding dashboards, re-implementing security, and disrupting customers. Choose for your 3-year trajectory, not today.
Is open source actually free for embedding?
For internal use, yes. For customer-facing embedding, you need multi-tenancy, security, and branding — features open source doesn't include. Metabase allows embedding but without SSO, audit logging, or sandboxed queries. Budget $100-200k/year in engineering for a production-grade setup.
What does "per-tenant" mean with sub-accounts?
A tenant is your direct customer — the company that signed your contract. Divisions or sub-accounts within that company usually count as one tenant. Gigacatalyst counts the SaaS customer, not their internal org structure.
How does AI change the pricing equation?
Traditional analytics charges you for dashboards your team builds and maintains. With AI-powered builders like Gigacatalyst, customers build their own on demand. This eliminates the cost of pre-building reports and reduces the dashboard objects you're paying to host.
What should I budget for year one?
Platform licensing + 2-3 months of engineering salary for integration + infrastructure + 30% contingency. For most mid-market SaaS, budget $50,000-150,000 all-in regardless of vendor. Licensing portion varies from $0 (open source) to $100,000+ (enterprise contracts).
The bottom line #
Pricing models aren't interchangeable. A platform cheap at 1,000 MAUs becomes painful at 50,000. Per-session works for occasional reporting but punishes you when users love your dashboards. Enterprise contracts look stable until renewal leverage kicks in.
The real question: which model won't punish you for growing?
Per-tenant aligns cost with revenue. More customers = more revenue = proportionally more analytics spend. No surprises from engagement spikes. No penalties for power users. No awkward finance conversation about why your analytics bill tripled during a great quarter.
That's why we built Gigacatalyst on per-tenant — and combined it with AI-generated dashboards that eliminate the engineering cost of building reports entirely. Customers build their own. Cost stays predictable. Engineers stay on the roadmap.
Want to model it for your case? Book a call and we'll walk through the numbers with your actual projections.
Pricing reflects publicly available data as of August 2026. Verify current rates directly before making decisions.