
Helicone AI
Langfuse
Portkey
BareMetrics
ChartMogul
ProfitWell
MarginDash
Gross margin per customer for AI SaaS

Helicone AI
Langfuse
MarginDash
LangSmith
LLMCap
OpenAI
Corrath
Know your AI costs per customer, per feature, per tier. PerUnit attributes every dollar of AI provider spend to the customers and features driving it.

Which is more popular?
Website, pricing, platforms and company facts side by side.
|
|
|
|
|---|---|---|
| Website | marginward.com | perunit.ai |
| Pricing | ||
| Platforms | ||
| Company | Startup from France · 1 - 9 employees · 2026 | — |
| Listed in |
In their own words, as submitted to SaaSHub.


MarginWard shows gross margin per customer for AI SaaS. It joins your LLM costs (Langfuse, OpenRouter, or a simple ingest API) with your Stripe revenue, flags customers who are unprofitable, and alerts you the moment one turns red. Free calculator, no signup. Paid plans from $29/mo.
Your AI bill is growing — but which customers, features, and pricing tiers are driving it? OpenAI, Anthropic, and Google dashboards show totals. Totals don't help you decide who to charge more, what to gate, or where to cut. PerUnit breaks down AI spend by customer, feature, and pricing tier so...
What each product offers, as listed by its team.


An editorial look at what each product does well and who it suits.


Overall verdict
Why this product is good
Recommended for
Overall verdict
Why this product is good
Recommended for
Walkthroughs and reviews on video.
MarginWard — gross margin per customer for AI SaaS
No PerUnit videos yet. You could help us improve this page by suggesting one.
How often each product is chosen within a category, 0–100% relative to the other.


As answered by people managing MarginWard and PerUnit.
MarginWard's answer
MarginWard is the only tool that joins your LLM costs with your Stripe revenue to show gross margin per customer. Cost-tracking tools show spend; revenue analytics show MRR, neither tells you which customers cost more than they pay. MarginWard does, and alerts you the moment one turns unprofitable.
PerUnit's answer:
PerUnit is built for AI unit economics, not just cost tracking. We break down AI spend by:
Customer — see which accounts drive costs Feature — see which AI features burn through tokens Pricing tier — free vs paid user breakdown We connect AI costs to revenue via Stripe. Provider dashboards show totals; observability tools show tokens. PerUnit shows who's driving costs and whether they're profitable. No data warehouse or pipelines — direct API sync to OpenAI, Anthropic, and Google.
MarginWard's answer
Observability tools track your LLM spend but don't know your revenue; SaaS analytics track revenue but ignore token cost. MarginWard is built for the intersection, gross margin per customer, with alerts on unprofitable accounts. Read-only Stripe key, plugs into Langfuse/OpenRouter or a simple ingest API, set up in ~15 minutes. Flat pricing, never a percentage of your spend. There's also a free calculator, no signup.
PerUnit's answer:
PerUnit is built for product and pricing decisions, not observability. Competitors focus on tokens, latency, and traces. We focus on:
Cost per customer Cost per feature Cost per tier — connected to revenue If you need to know who to charge more, what to gate, or where to cut, PerUnit gives you that. No pipelines, no data engineering, no data warehouse.
MarginWard's answer
Founders and teams building AI SaaS on flat or subscription pricing, products where LLM tokens are a real per-customer cost. From solo founders to small product teams who need their true unit economics, not just their MRR.
PerUnit's answer:
Founders, product managers, and CFOs at companies building AI-powered products. Anyone who needs to answer:
Which customers cost the most? Are free users profitable? Should we gate AI behind paid plans? Teams that care about unit economics and margins, not just total spend.
MarginWard's answer
MarginWard was built by a solo founder running his own AI SaaS. One month he realised his biggest customers were also his least profitable, burning more in LLM tokens than they paid, and Stripe never told him. So he built the tool he wished existed: the real margin of an AI product, customer by customer.
PerUnit's answer:
We've shipped products with AI and watched margins disappear without knowing why. The AI bill was one number — we couldn't see which customers or features were driving it. Every product and pricing decision was a guess.
We built PerUnit so product teams can have the same clarity on AI costs that they have on revenue and headcount.
MarginWard's answer
Next.js, TypeScript, Supabase (PostgreSQL), Stripe, Vercel, Tailwind CSS, Resend, with Langfuse and OpenRouter integrations for LLM cost data.
PerUnit's answer:
Frontend: Next.js, React, TypeScript, Tailwind CSS Integrations: OpenAI, Anthropic, Google APIs, Stripe Deployment: Vercel
PerUnit's answer:
PerUnit is in early access. We're onboarding a small group of teams who need clarity on their AI unit economics. Our early users are founders and PMs at companies building AI features — from early-stage to growth — who want cost attribution before scaling further.
Share your experience with using MarginWard and PerUnit. For example, how are they different and which one is better?
When comparing MarginWard and PerUnit, you can also consider the following products.

Open-source LLM Observability for Developers
Compare Helicone AI to MarginWard or PerUnit:

Langfuse is an open-source LLM engineering platform that helps teams collaboratively debug, analyze, and iterate on their LLM applications.
Compare Langfuse to MarginWard or PerUnit:

Build production-grade & reliable AI apps with Portkey
Compare Portkey to MarginWard or PerUnit:

Track AI cost and margin per customer. Real-time profitability insights, Stripe revenue sync, budget alerts, and a cost simulator to find cheaper models without changing code.
Compare MarginDash to MarginWard or PerUnit:


Build and deploy LLM applications with confidence
Compare LangSmith to MarginWard or PerUnit: